Stratechery Today

Amazon Earnings, CapEx Concerns, Commodity AI

Amazon's massive CapEx increase makes me much more nervous than Google's, but it is understandable.

Business Cloud Finance

▲ 0

0 views

Stratechery Yesterday

Google Earnings, Google Cloud Crushes, Search Advertising and LLMs

Google announced a massive increase in CapEx that blew away expectations; the companies earnings results explain why the increase is justified.

AI Cloud

▲ 0

0 views

Stratechery 4 days ago

2026.06: SaaSmageddon and the Super Bowl

Welcome back to This Week in Stratechery! As a reminder, each week, every Friday, we’re sending out this overview of content in the Stratechery bundle; highlighted links are free for everyone . Additionally, you have complete control over what we send to you. If you don’t want to receive This Week in Stratechery emails (there is no podcast), please uncheck the box in your delivery settings . On that note, here were a few of our favorites this week. This week’s Stratechery video is on TSMC Risk . Is Software Dead? Software stocks have been in a free-fall all week, up-t0-and-including the biggest software company of them all: Microsoft. It’s tempting to say that everyone is over-reacting to the threat of AI — and they are, in the short run — but history shows that fundamentally changing in industry’s inputs transforms that industry in the long run, to the detriment of incumbents: just look at what the Internet did to content. Given that, Microsoft’s urgency in building out its own AI products, even if that meant missing on Azure numbers, is the right choice . Oh, and did I mention that tech is facing a massive compute supply crisis ? — Ben Thompson SaaSmageddon and Super Bowl Ads. Building on that Microsoft article, Ben and I discussed the future of Saas companies on this week’s Sharp Tech , including a more than half-trillion dollar collapse of the Nasdaq 100 this week. Is the market’s skepticism fair? We dive into why software companies have more moats than their skeptics acknowledge, but nevertheless face a variety of headwinds that are likely to spur painful corrections to the valuation of these companies, consolidation, and substantial layoffs. Additionally, we had a great time talking through deceptive Anthropic Super Bowl ads — a series of broadsides at OpenAI’s nascent advertising play — that Ben hated, why Sam Altman’s response was spot on, and who their real audience is.  — Andrew Sharp Madness in Basketball and Football. Speaking of Sunday… I don’t have a Seahawks-Patriots preview for you, on Sharp Text, but I did celebrated the occasion with a tribute to the madness and sneaky depth of Any Given Sunday . Elsewhere in the Stratechery sports universe, the NBA Trade Deadline came and went on Thursday this week, and Greatest of All Talk covered a very busy week of transactions across the association . Come to hear my anxious and unconvincing endorsement of my Wizards’ move to add Anthony Davis, and stay for thoughts on a topsy turvy deadline where the worst teams were buyers, the Celtics look like evil geniuses, and Giannis Antetokounmpo is staying in Milwaukee for at least few more months. — AS Microsoft and Software Survival — Microsoft got hammered on Wall Street for capacity allocation decisions that were the right ones: the software that wins will use AI to usurp other software. Apple Earnings, Supply Chain Speculation, China and Industrial Design — Apple’s earnings could have been higher but the company couldn’t get enough chips; then, once again a new design meant higher sales in China. An Interview with Benedict Evans About AI and Software — An interview with Benedict Evans about the crisis facing software, the future of the corporation, OpenAI, and the struggle to define the LLM paradigm. What ‘Any Given Sunday’ Gets Right — ‘ Any Given Sunday’ is a product of its time, and its treatment of modern pro football is both more alive and more poignant than just about any sports movie to emerge since. Apple Earnings and OpenClaw Silicon Valley Thinks TSMC is Braking the AI Boom Invasion of the Microplastics The PLA Purges One Week Later; World Leaders Flock to Beijing; A Trump-Xi Phone Call; Panama Canal Resolution? Deadline Notes and All-Star Announcements, The Scintillating Charlotte Hornets, Flagg, Dybantsa and Darryn Peterson Trade Deadline 2026: AD to DC?, A Topsy Turvy Week, Pacers Bet Big on Big Zu, Jazz and JJJ, Evil Celtics, and Lots More SaaSmageddon and the Future, Microsoft After a Market Correction, Anthropic’s Super Bowl Lies

AI Business

▲ 0

0 views

Stratechery 5 days ago

An Interview with Benedict Evans About AI and Software

An interview with Benedict Evans about the crisis facing software, the future of the corporation, OpenAI, and the struggle to define the LLM paradigm.

Business AI

▲ 0

1 views

Stratechery 6 days ago

Apple Earnings, Supply Chain Speculation, China and Industrial Design

Apple's earnings could have been higher but the company couldn't get enough chips; then, once again a new design meant higher sales in China.

Business Hardware Finance

▲ 0

0 views

Stratechery 1 weeks ago

Microsoft and Software Survival

Listen to this post : One way to track the AI era, starting with the November 2022 launch of ChatGPT, is by which Big Tech company was, at a particular point in time, thought to be most threatened. At the beginning everyone — including yours truly — was concerned about Google and the potential disruption of Search. Then, early last year , it was Apple’s turn , as its more intelligent Siri stumbled so badly it didn’t even launch. By the fall it was Meta’s in the crosshairs , as the company completely relaunched its AI efforts as Llama hit a wall. Now it’s Microsoft’s turn, which is a bit of a full circle moment, given that the company was thought to be the biggest winner from ChatGPT in particular, thanks to its partnership with OpenAI. I wrote in early 2023 in AI and the Big Five : Microsoft, meanwhile, seems the best placed of all. Like AWS it has a cloud service that sells GPUs; it is also the exclusive cloud provider for OpenAI. Yes, that is incredibly expensive, but given that OpenAI appears to have the inside track to being the AI epoch’s addition to this list of top tech companies, that means that Microsoft is investing in the infrastructure of that epoch. Bing, meanwhile, is like the Mac on the eve of the iPhone: yes it contributes a fair bit of revenue, but a fraction of the dominant player, and a relatively immaterial amount in the context of Microsoft as a whole. If incorporating ChatGPT-like results into Bing risks the business model for the opportunity to gain massive market share, that is a bet well worth making. The latest report from The Information , meanwhile, is that GPT is eventually coming to Microsoft’s productivity apps. The trick will be to imitate the success of AI-coding tool GitHub Copilot (which is built on GPT), which figured out how to be a help instead of a nuisance (i.e. don’t be Clippy!). What is important is that adding on new functionality — perhaps for a fee — fits perfectly with Microsoft’s subscription business model. It is notable that the company once thought of as a poster child for victims of disruption will, in the full recounting, not just be born of disruption, but be well-placed to reach greater heights because of it. I do, I must admit, post that excerpt somewhat sheepishly, as much of it seems woefully shortsighted: All of these factors — plus the fact that Azure growth came in a percentage point lower than expected — contributed to one of the worst days in stock market history. From Bloomberg last week: Microsoft Corp. shares got caught up in a selloff Thursday that wiped out $357 billion in value, second-largest for a single session in stock market history. The software giant’s stock closed down 10%, its biggest plunge since March 2020, following Microsoft’s earnings after the bell Wednesday, which showed record spending on artificial intelligence as growth at its key cloud unit slowed. The only bigger one-day valuation destruction was Nvidia Corp.’s $593 billion rout last year after the launch of DeepSeek’s low-cost AI model. Microsoft’s move is larger than the market capitalizations of more than 90% of S&P 500 Index members, according to data compiled by Bloomberg… The selloff comes amid heightened skepticism from investors that the hundreds of billions of dollars Big Tech is spending on AI will eventually pay off. Microsoft’s results showed a 66% rise in capital expenditures in its most recent quarter to a record $37.5 billion, while growth at its closely tracked Azure cloud-computing unit slowed from the prior quarter. I laid out my base case for Big Tech back in 2020 in The End of the Beginning , arguing that the big tech companies would be the foundation on which future paradigms would be built; is Microsoft the one that might crack? It can, when it comes to vibe coding, be difficult to parse the hype on X from the reality on the ground; what is clear is the trajectory. I have talked to experienced software engineers who will spend 10 minutes complaining about the hype and all of the shortcomings of Claude Code or OpenAI Codex, only to conclude by admitting that AI just helped them write a new feature or app that they never would have otherwise, or would have taken far longer to do than it actually did. The beauty of AI writing code is that it is a nearly perfect match of probabilistic inputs and deterministic outputs: the code needs to actually run, and that running code can be tested and debugged. Given this match I do think it is only a matter of time before the vast majority of software is written by AI, even if the role of the software architect remains important for a bit longer. That, then, raises the most obvious bear case for any software company: why pay for software when you can just ask AI to write your own application, perfectly suited to your needs? Is software going to be a total commodity and a non-viable business model in the future? I’m skeptical, for a number of reasons. First, companies — particularly American ones — are very good at focusing on their core competency, and for most companies in the world, that isn’t software. There is a reason most companies pay other companies for software, and the most fundamental reason to do so won’t change with AI. Second, writing the original app is just the beginning: there is maintenance, there are security patches, there are new features, there are changing standards — writing an app is a commitment to a never-ending journey — a journey, to return to point one, that has nothing to do with the company’s core competency. Third, selling software isn’t just about selling code. There is support, there is compliance, there are integrations with other software, the list of what is actually valuable goes far beyond code. This is why companies don’t run purely open source software: they don’t want code, they want a product, with everything that entails. Still, that doesn’t mean the code isn’t being written by AI: it’s the software companies themselves that will be the biggest beneficiaries of and users of AI for writing code. In other words, on this narrow question of AI-written code, I would contend that software companies are not losers, but rather winners: they will be able to write more code more efficiently and quickly. When the Internet first came along it seemed, at first glance, a tremendous opportunity for publishers: suddenly their addressable market wasn’t just the geographic area they could deliver newspapers to, but rather the entire world! In fact, the nature of the opportunity was the exact opposite; from 2014’s Economic Power in the Age of Abundance : One of the great paradoxes for newspapers today is that their financial prospects are inversely correlated to their addressable market. Even as advertising revenues have fallen off a cliff — adjusted for inflation, ad revenues are at the same level as the 1950s — newspapers are able to reach audiences not just in their hometowns but literally all over the world. The problem for publishers, though, is that the free distribution provided by the Internet is not an exclusive. It’s available to every other newspaper as well. Moreover, it’s also available to publishers of any type, even bloggers like myself. To be clear, this is absolutely a boon, particularly for readers, but also for any writer looking to have a broad impact. For your typical newspaper, though, the competitive environment is diametrically opposed to what they are used to: instead of there being a scarce amount of published material, there is an overwhelming abundance. More importantly, this shift in the competitive environment has fundamentally changed just who has economic power. The power I was referring to was Google; this Article was an articulation of Aggregation Theory a year before I coined the term. The relevance to AI-written code, however, is not necessarily about Aggregators, but rather about inputs. Specifically, what changed for publishers is that the cost of distribution went to zero: of course that was beneficial for any one publisher, but it was disastrous for publishers as a collective. In the case of software companies, the input that is changing is the cost of code: it’s not going completely to zero, at least not yet — you still need a managing engineer, for one, and tokens, particularly for leading edge models actually capable of writing usable code, have significant marginal costs — but the relative cost is much lower, and the trend is indeed towards zero. If you want to carry this comparison forward, this is an argument against there even being a market for software in the long run. After all, the most consumed form of content on the Internet today, three decades on, is in fact user-generated content, which you could analogize to companies having AI write their own software. That seems a reasonable bet for 2056 — if we even have companies then ( I think we will ). In the shorter-term, however, the real risk I see for software companies is the fact that while they can write infinite software thanks to AI, so can every other software company. I suspect this will completely upend the relatively neat and infinitely siloed SaaS ecosystem that has been Silicon Valley’s bread-and-butter for the last decade: identify a business function, leverage open source to write a SaaS app that addresses that function, hire a sales team, do some cohort analysis, IPO, and tell yourself that you were changing the world. The problem now, however, is that while businesses may not give up on software, they don’t necessarily want to buy more — if anything, they need to cut their spending so they have more money for their own tokens. That means the growth story for all of these companies is in serious question — the industry-wide re-rating seems completely justified to me — which means the most optimal application of that new AI coding capability will be to start attacking adjacencies, justifying both your existence and also presenting the opportunity to raise prices. In other words, for the last decade the SaaS story has been about growing the pie: the next decade is going to be about fighting for it, and the model makers will be the arms dealers. While this battle is happening, there will be another fundamental shift taking place: yes, humans will be using software, at least for a while, but increasingly so will agents. What isn’t clear is who will be creating the agents: I expect every SaaS app to have their own agent, but that agent will definitionally be bound by the borders of the application (which will be another reason to expand the app into adjacent areas). Different horizontal players, meanwhile, will be making a play to cover broader expanses of the business, with the promise of working across multiple apps. Microsoft is one of those horizontal layers, and the company’s starting point for agents is what it is calling Work IQ; here is how CEO Satya Nadella explained Work IQ on the company’s earnings call : Work IQ takes the data underneath Microsoft 365 and creates the most valuable stateful agent for every organization. It delivers powerful reasoning capabilities over people, their roles, their artifacts, their communications and their history and memory all within an organization security boundary. Microsoft 365 Copilot’s accuracy and latency powered by Work IQ is unmatched, delivering faster and more accurate work grounded results than competition, and we have seen our biggest quarter-over-quarter improvement in response quality to date. This has driven record usage intensity with average number of conversations per user doubling year-over-year. This feels like the right layer for Microsoft, given the company’s ownership of identity. Active Directory is one of the most valuable free products of all time: it was the linchpin via which Microsoft tied together all of its enterprise products and services, first driving upgrades up and down the stack, and later underpinning its per-seat licensing business model. That the company sees its understanding of the individual worker and all of his or her artifacts, permissions, etc. as the obvious place to build agents makes sense. There’s one big problem with this starting point, however: it’s shrinking. Owning and organizing a company by identity is progressively less valuable if the number of human identities starts to dwindle — and, with a per-seat licensing model, you make less money. That, by extension, means that Microsoft should feel a significant amount of urgency when it comes to fighting the adjacency battles I predicted above. First, directly incorporating more business functions into Microsoft’s own software suite will make Microsoft’s agents more capable. Secondly, absorbing more business functions into Microsoft’s software offering will let the company charge more. Third, the larger Microsoft’s surface area, the more power it will have to compel other software makers to interface with its agents, increasing their capability. This pressure explains the choices Microsoft made that led to its Azure miss in particular. Microsoft was clear that, once again, demand exceeded supply. CFO Amy Hood said in her prepared remarks: Our customer demand continues to exceed our supply. Therefore, we must balance the need to have our incoming supply better meet growing Azure demand with expanding first-party AI usage across services like M365 Copilot and GitHub Copilot, increasing allocations to R&D teams to accelerate product innovation and continued replacement of end-of-life server and networking equipment. She further explained in the Q&A section that Azure revenue was directly downstream from Microsoft’s own capacity allocation: I think it’s probably better to think about the Azure guidance that we give as an allocated capacity guide about what we can deliver in Azure revenue. Because as we spend the capital and put GPUs specifically, it applies to CPUs, the GPUs more specifically, we’re really making long-term decisions. And the first thing we’re doing is solving for the increased usage in sales and the accelerating pace of M365 Copilot as well as GitHub Copilot, our first-party apps. Then we make sure we’re investing in the long-term nature of R&D and product innovation. And much of the acceleration that I think you’ve seen from us and products over the past a bit is coming because we are allocating GPUs and capacity to many of the talented AI people we’ve been hiring over the past years. Then, when you end up, is that, you end up with the remainder going towards serving the Azure capacity that continues to grow in terms of demand. And a way to think about it, because I think, I get asked this question sometimes, is if I had taken the GPUs that just came online in Q1 and Q2 in terms of GPUs and allocated them all to Azure, the KPI would have been over 40. And I think the most important thing to realize is that this is about investing in all the layers of the stack that benefit customers. And I think that’s hopefully helpful in terms of thinking about capital growth, it shows in every piece, it shows in revenue growth across the business and shows as OpEx growth as we invest in our people. Nadella called this a portfolio approach: Basically, as an investor, I think when you think about our capital and you think about the gross margin profile of our portfolio, you should obviously think about Azure. But you should think about M365 Copilot and you should think about GitHub pilot, you should think about Dragon Copilot, Security Copilot. All of those have a gross margin profile and lifetime value. I mean if you think about it, acquiring an Azure customer is super important to us, but so is acquiring an M365 or a GitHub or a Dragon Copilot, which are all by the way incremental businesses and TAMs for us. And so we don’t want to maximize just 1 business of ours, we want to be able to allocate capacity, while we’re sort of supply constrained in a way that allow us to essentially build the best LTV portfolio. That’s on one side. And the other one that Amy mentioned is also R&D. I mean you got to think about compute is also R&D, and that’s sort of the second element of it. And so we are using all of that, obviously, to optimize for the long term. The first part of Nadella’s answer is straightforward: Microsoft makes better margins and has more lifetime value from its productivity applications than it does from renting out Azure capacity, so investors should be happy that it is allocating scarce resources to that side of the business. And, per the competition point above, this is defensive as well: if Microsoft doesn’t get AI right for its own software then competitors will soon be moving in. The R&D point, however, is also critical: Microsoft also needs to be working to expand its offering, and increasingly the way to do that is going to be by using AI to write that new software. That takes a lot of GPUs — so many that Microsoft simply didn’t have enough to meet the 40% Azure growth rate that Wall Street expected. I think it was the right decision. There are some broader issues raised by Microsoft’s capacity allocation. First, we have the most powerful example yet of the downside of having insufficient chips. Hood was explicit that Microsoft could have beat Wall Street’s number if they had enough GPUs; the fact they didn’t was a precipitating factor in losing $357 billion in value. How much greater will the misses be a few years down the road when AI demand expands even further, particularly if TSMC both remains the only option and continues to be conservative in its CapEx ? Secondly, however, it’s fair for Azure customers to feel a bit put out by Microsoft’s decision to favor itself. It reminds me of the pre-TSMC world, when fabs were a part of Integrated Device Manufacturers like Intel or Texas Instruments. If you wanted to manufacture a chip you could contract for space on their lines, but you were liable to lose that capacity if the fab needed it for their own products; TSMC was unique in that they were a pure play foundry: their capacity was solely for their customers, who they weren’t going to compete with. This isn’t the case with Azure: Microsoft has first dibs, and then OpenAI, and then everyone else, and that priority order was made clear this quarter. Moreover, it’s fair to assume that Amazon and Google will make similar prioritization decisions. I didn’t, before writing this article, fully grok the potential for neoclouds, or Oracle for that matter, but the value proposition of offering a pure play token foundry might be larger than I appreciated. All that noted, the safest assumption is that Microsoft, like the rest of Big Tech, will figure this out. Some software may be dead, but not all of it, at least not yet, and the biggest software maker of them all is — thanks in part to that size — positioned to be one of the survivors. It’s just going to need a lot of compute, not only for its customers, but especially for itself. OpenAI is still Azure’s biggest customer, but the fact that the maker of ChatGPT represents 45% of Azure’s Remaining Performance Obligations (RPO) is now seen as a detriment by the market. Bing was briefly interesting when it contained Sydney ; Microsoft quickly squashed what remains the single most compelling AI experience I’ve had and, one could make the case, Bing’s growth prospects. All of Microsoft’s products have a CoPilot of some sort; it’s not clear how well any of them work, and both Claude and OpenAI are attacking the professional productivity space. Microsoft 365 CoPilot has 15 million paying customers, but (1) that’s a tiny fraction of Microsoft 365’s overall customer base and (2) the rise of agents raises serious questions about the long-term viability of the per-seat licensing model on which Microsoft’s productivity business is built.

AI Business

▲ 0

0 views

Stratechery 1 weeks ago

Personal Day — Sick

Personal Day — Sick

Career

▲ 0

0 views

Stratechery 1 weeks ago

2026.05: The Chip Fly in the AI Ointment

Welcome back to This Week in Stratechery! As a reminder, each week, every Friday, we’re sending out this overview of content in the Stratechery bundle; highlighted links are free for everyone . Additionally, you have complete control over what we send to you. If you don’t want to receive This Week in Stratechery emails (there is no podcast), please uncheck the box in your delivery settings . On that note, here were a few of our favorites this week. This week’s Sharp Tech video is on the Apple Vision Pro not getting live events right. Tech’s Looming Chip Problem . My first obsession in tech was semiconductors, back when every new generation provided such astronomical gains that it fundamentally shaped how the entire software industry approached development. Then, for a few decades, semiconductors faded to the background: yes, they matter, but they were commodities, both in terms of availability and ongoing costs. Today, however, chips matter more than ever thanks to AI. Yet even though it’s no longer an esoteric topic, I actually think that the industry is not thinking about them enough : today we have chip shortages, but if demand doesn’t take the initiative to restructure supply than the shortages looming by the end of the decade could dramatically curtail AI’s impact and, by extension, the entire industry’s revenue potential. — Ben Thompson What Is Meta Doing, and Why?  Thursday’s Daily Update was a return to Stratechery’s bread and butter: unpacking Meta’s latest earnings announcement and explaining why Wall Street might be overreacting to the results. Come to read how Meta beat expectations again (continuing a trend that began in Q2 2025 ), and stay to learn why, despite Wall Street’s apparent permission to spend up to $135 billion on AI infrastructure , there are fair questions to ask of Zuckerberg’s enormous bets. We doubled back to discuss all this on the podcast this week , including thoughts on whether Zuckerberg is ultimately stuck in the same box he’s been in for years — the world’s most successful app developer, yearning to be something more — and the stark contrast between Apple and Meta in the AI era.  — Andrew Sharp Disappearing PLA Generals and “De-Risking.”  The past few weeks have seen both Canada and the U.K. make outreach to the Chinese government to gin up new business and de-risk from the U.S.; on Sharp Text, I wrote about why that’s a bad idea (and why the fawning over Mark Carney’s Davos speech annoyed me). Elsewhere, a warning: we don’t have many answers as to what the hell is happening at the top of the PLA, but Xi has now purged 5 of the 6 appointees of the Central Military Commission, and 50 senior officers since 2023, including his number two in command, Gen. Zhang Youxia, last weekend. This week’s episode of Sharp China probes the questions raised by all this (for Taiwan, for Xi, was Zhang leaking nuclear secrets to the US?) and it was a terrific coda to a week of rampant speculation in the wake of genuinely seismic China news.  — AS TSMC Risk — If hyperscalers and chip companies don’t build up a TSMC competitor they are set to forego billions of dollars in revenue and stunt the AI revolution. Intel Earnings, The Agentic Opportunity, Intel’s Mistaken Pessimism — Intel’s earnings were disappointing because the company is missing a huge opportunity by virtue of selling off its capacity. An Interview with Kalshi CEO Tarek Mansour About Prediction Markets — An interview with Kalshi co-founder and CEO Tarek Mansour about the value of prediction markets. Meta Earnings, Turning Dials, Zuckerberg’s Motivation — Meta is up, despite massive CapEx plans. The company is turning every dial to drive revenue, because Mark Zuckerberg thinks winning in AI is existential. The Scorpion and the Frogs — As world leaders look to China to de-risk from the U.S., it’s worth considering how we got here. Meta’s Tim Cook Doctrine Toshiba’s Breakthrough Laptop PC The Japanese AI Boom Needs A Little More Ambition 10 Questions on the Detention of Zhang Youxia, Corruption in the PLA, Rumors, Messaging, and Taiwan The Luka Trade One Year Later, The Hottest Team in the NBA and an OKC Reality Check, Another Giannis Injury Meta’s Plans to Spend $135 Billion, The ‘AI Bubble’ Bubble?, Why Hyperscalers Should NOT Invest in TSMC

Hardware AI Business

▲ 1

9 views

Stratechery 1 weeks ago

Meta Earnings, Turning Dials, Zuckerberg’s Motivation

Meta is up, despite massive CapEx plans. The company is turning every dial to drive revenue, because Mark Zuckerberg thinks winning in AI is existential.

Business Finance AI

▲ 0

0 views

Stratechery 1 weeks ago

An Interview with Kalshi CEO Tarek Monsour About Prediction Markets

An interview with Kalshi co-founder and CEO Tarek Monsour about the value of prediction markets.

Business Finance Data

▲ 0

0 views

Stratechery 2 weeks ago

Intel Earnings, The Agentic Opportunity, Intel’s Mistaken Pessimism

Intel's earnings were disappointing because the company is missing a huge opportunity by virtue of selling off its capacity.

Hardware Business

▲ 0

0 views

Stratechery 2 weeks ago

TSMC Risk

Listen to this post : You probably think, given this title, you know what this Article is about. The most advanced semiconductors are made by TSMC in Taiwan, 1 and Taiwan is claimed by China, which has not and will not take reunification-by-force off of the table. Relatedly, AI obviously has significant national security implications; at Davos, Anthropic CEO Dario Amodei reiterated his objection to the U.S. allowing the sale of Nvidia chips to China. From Bloomberg : Anthropic Chief Executive Officer Dario Amodei said selling advanced artificial intelligence chips to China is a blunder with “incredible national security implications” as the US moves to allow Nvidia Corp. to sell its H200 processors to Beijing. “It would be a big mistake to ship these chips,” Amodei said in an interview with Bloomberg Editor-in-Chief John Micklethwait at the World Economic Forum in Davos, Switzerland. “I think this is crazy. It’s a bit like selling nuclear weapons to North Korea.” The nuclear weapon analogy is an interesting one: a lot of game theory was developed to manage the risk of nuclear weapons, particularly once the U.S.S.R. gained/stole nuclear capability, ending the U.S.’s brief monopoly on the technology. Before that happened, however, the U.S. had a dominant military position, given we had nuclear weapons and no one else did. Perhaps Amodei believes the U.S. should have advanced AI and China should not, giving us a dominant military position? The problem with that reality, however, is Taiwan, as I explained in AI Promise and Chip Precariousness . AI, in contrast to nuclear weapons, has a physical dependency in Taiwan that can be easily destroyed by Chinese missiles, even without an invasion; if we got to a situation where only the U.S. had the sort of AI that would give us an unassailable advantage militarily, then the optimal strategy for China would change to taking TSMC off of the board. Given this dependency, my recommendations in the Article run counter to Amodei: I want China dependent on not just U.S. chips but also on TSMC directly, which is why I argued in favor of selling Nvidia chips to China, and further believe that Huawei and other Chinese companies ought to be able to source from TSMC (on the flip side, I would ban the sale of semiconductor manufacturing equipment to Chinese fabs). I think it’s a good thing the Trump administration moved on the first point, at least. However, this risk is not what this Article is about: there is another TSMC risk facing the entire AI industry in particular; moreover, it’s a risk the downside of which is already being realized. There was one refrain that was common across Big Tech earnings last quarter: demand for AI exceeds supply. Here was Amazon CEO Andy Jassy on the company’s earnings call : You’re going to see us continue to be very aggressive investing in capacity because we see the demand. As fast as we’re adding capacity right now, we’re monetizing it. Here was Microsoft CFO Amy Hood on the company’s earnings call : Azure AI services revenue was generally in line with expectations, and this quarter, demand again exceeded supply across workloads, even as we brought more capacity online. Here was Google CFO Anat Ashkenazi on the company’s earnings call : In GCP, we see strong demand for enterprise AI infrastructure, including TPUs and GPUs, enterprise AI solutions driven by demand for Gemini 2.5 and our other AI models, and core GCP infrastructure and other services such as cybersecurity and data analytics. As I’ve mentioned on previous earnings calls, while we have been working hard to increase capacity and have improved the pace of server deployments and data center construction, we still expect to remain in a tight demand-supply environment in Q4 and 2026. Here was Meta CEO Mark Zuckerberg on the company’s earnings call : To date, we keep on seeing this pattern where we build some amount of infrastructure to what we think is an aggressive assumption. And then we keep on having more demand to be able to use more compute, especially in the core business in ways that we think would be quite profitable than we end up having compute for. Earlier this month, TSMC CEO C.C. Wei admitted that the shortage was a lack of chips, not power; from the company’s earnings call : Talking about to build a lot of AI data center all over the world, I use one of my customers’ customers’ answer. I asked the same question. They told me that they planned this one, 5-6 years ago already. So, as I said, those cloud service providers are smart, very smart. So, they say that they work on the power supply 5-6 years ago. So, today, their message to me is: silicon from TSMC is a bottleneck, and asked me not to pay attention to all others, because they have to solve the silicon bottleneck first. But indeed, we do get the power supply, all over the world, especially in the US. Not only that, but we also look at, who support those kind of a power supply, like a turbine, like, what, nuclear power plant, the plan or those kinds of things. We also look at the supply of the rack. We also look at the supply of the cooling system. Everything, so far, so good. So we have to work hard to narrow the gap between the demand and supply from TSMC. The cause of that gap is obvious if you look at TSMC’s financials, specifically the company’s annual capital expenditures: After a big increase in CapEx in 2021, driven by the COVID shortages and a belief in 5G , TSMC’s annual CapEx in the following years was basically flat — it actually declined on a year-over-year basis in both 2023 and 2024. Note those dates! ChatGPT was released in November 2022; that kicked off a massive increase in CapEx amongst the hyperscalers in particular, but it sure seems like TSMC didn’t buy the hype. That lack of increased investment earlier this decade is why there is a shortage today, and is why TSMC has been a de facto brake on the AI buildout/bubble; I wrote last quarter : To put it another way, if Altman and OpenAI are the ones pushing to accelerate the AI infrastructure buildout, it’s Wei and TSMC that are the brakes. The extent to which all of Altman’s deals actually materialize is dependent on how much TSMC invests in capacity now, and while they haven’t shown their hand yet, the company is saying all of the right things about AI being a huge trend without having yet committed to a commensurate level of investment, at least relative to OpenAI’s goals. That Update was about the future, but it’s important to note that the TSMC brake has — if all of those CEO and CFO comments above are to be believed — already cost the biggest tech companies a lot of money. That’s the implication of not having enough supply to satisfy demand: there was revenue to be made that wasn’t, because TSMC didn’t buy the AI hype at the same time everyone else did. TSMC is, finally, starting to invest more. Last year’s CapEx increased 37% to $41 billion, and there’s another increase in store for this year to $52–$56 billion; if we take the midpoint, that represents an increase of 32%, a bit less than last year: Make no mistake, $54 billion is a big number, one that Wei admitted made him nervous: You essentially try to ask whether the AI demand is real or not. I’m also very nervous about it. Yeah, you bet, because we have to invest about USD52 billion to USD56 billion for the CapEx, right? If we did not do it carefully, that will be a big disaster to TSMC for sure. So, of course, I spent a lot of time in the last three-four months talking to my customers and then customers’ customers. I want to make sure that my customers’ demands are real. Wei made clear that he was worried about the market several years down the line: If you build a new fab, it takes two and three year, two to three years to build a new fab. So even we start to spend $52 billion to $56 billion, the contribution to this year is almost none, and 2027, a little bit. So we actually, we are looking for 2028-2029 supply, and we hope it’s a time that the gap will be narrow…So 2026-2027 for the short-term, we are looking to improve our productivity. 2028 to 2029, yes, we start to increase our capacity significantly. And it will continue this way if the AI demand megatrend as we expected. First off, this delayed impact explains why TSMC’s lack of CapEx increase a few years ago is resulting in supply-demand imbalance today. Secondly, notice how this year’s planned increase — which again, won’t really have an impact until 2028 — pales in comparison to the CapEx growth of the hyperscalers ( 2025 numbers are estimates; note that Amazon’s CapEx includes Amazon.com ): Remember, a significant portion of this CapEx growth is for chips that are supported by TSMC’s stagnant CapEx growth from a few years ago. It’s notable, then, that TSMC’s current and projected CapEx growth is still less than the hyperscalers: how much less is it going to be than the hyperscalers’ growth in 2028, when the fabs being built today start actually producing chips? In short, the TSMC brake isn’t going anywhere — if anything, it’s being pressed harder than ever. TSMC is, to be clear, being extremely rational. CapEx is inherently risky: you are spending money now in anticipation of demand that may or may not materialize. Moreover, the risk for a foundry is higher than basically any other business model: nearly all of a foundry’s costs are CapEx, which means that if demand fails to materialize, costs — in the form of depreciation — don’t go down as they might with a business model with a higher percentage of marginal costs. This is exacerbated by the huge dollar figures entailed in building fabs: $52–$56 billion may drive revenues with big margins, but those big margins can easily flip to being huge losses and years of diminished pricing power thanks to excess capacity. Therefore, it’s understandable that TSMC is trying to manage its risks. Sure, the company may be foregoing some upside in 2028, but what is top of Wei’s mind is avoiding “a big disaster.” What is important to note, however, is that the risk TSMC is managing doesn’t simply go away: rather, it’s being offloaded to the hyperscalers in particular. Specifically, if we get to 2028, and TSMC still isn’t producing enough chips to satisfy demand, then that means the hyperscalers will be forgoing billions of dollars in revenue — even more than they are already forgoing today. Yes, that risk is harder to see than the risk TSMC is avoiding, because the hyperscalers aren’t going to be bankrupt for a lack of chips to satisfy demand. Still, the potential money not made — particularly when the number is potentially in the hundreds of billions of dollars — is very much a risk that the hyperscalers are incurring because of TSMC’s conservatism. What the hyperscalers need to understand is that simply begging TSMC to make more isn’t going to fix this problem, because begging TSMC to make more is to basically ask TSMC to take back the risk TSMC is offloading to the hyperscalers — they already declined! Rather, the only thing that will truly motivate TSMC to take on more risk is competition. If TSMC were worried about not just forgoing its own extra revenue, but actually losing business to a competitor, then the company would invest more. Moreover, that extra investment would be stacked on top of the investment made by said competitor, which means the world would suddenly have dramatically more fab capacity. In short, the only way to truly get an AI bubble, with all of the potential benefits that entails , or, in the optimistic case, to actually meet demand in 2028 and beyond, is to have competition in the foundry space. That, by extension, means Samsung or Intel — or both — actually being viable options. Remember, however, the number one challenge facing those foundries: a lack of demand from the exact companies whom TSMC has deputized to take on their risk. I wrote in U.S. Intel : Our mythical startup, however, doesn’t exist in a vacuum: it exists in the same world as TSMC, the company who has defined the modern pure play foundry. TSMC has put in the years, and they’ve put in the money; TSMC has the unparalleled customer service approach that created the entire fabless chip industry; and, critically, TSMC, just as they did in the mobile era, is aggressively investing to meet the AI moment. If you’re an Nvidia, or an Apple in smartphones, or an AMD or a Qualcomm, why would you take the chance of fabricating your chips anywhere else? Sure, TSMC is raising prices in the face of massive demand, but the overall cost of a chip in a system is still quite small; is it worth risking your entire business to save a few dollars for worse performance with a worse customer experience that costs you time to market and potentially catastrophic product failures? We know our mythical startup would face these challenges because they are the exact challenges Intel faces. Intel may need “a meaningful external customer to drive acceptable returns on [its] deployed capital”, but Intel’s needs do not drive the decision-making of those external customers, despite the fact that Intel, while not fully caught up to TSMC, is at least in the ballpark, something no startup could hope to achieve for decades. Becoming a meaningful customer of Samsung or Intel is very risky: it takes years to get a chip working on a new process, which hardly seems worth it if that process might not be as good, and if the company offering the process definitely isn’t as customer service-centric as TSMC. I understand why everyone sticks with TSMC. The reality that hyperscalers and fabless chip companies need to wake up to, however, is that avoiding the risk of working with someone other than TSMC incurs new risks that are both harder to see and also much more substantial. Except again, we can see the harms already: foregone revenue today as demand outstrips supply. Today’s shortages, however, may prove to be peanuts: if AI has the potential these companies claim it does, future foregone revenue at the end of the decade is going to cost exponentially more — surely a lot more than whatever expense is necessary to make Samsung and/or Intel into viable competitors for TSMC. This, incidentally, is how the geographic risk issue will be fixed, if it ever is. It’s hard to get companies to pay for insurance for geopolitical risks that may never materialize. What is much more likely is that TSMC’s customers realize that their biggest risk isn’t that TSMC gets blown up by China, but that TSMC’s monopoly and reasonable reluctance to risk a rate of investment that matches the rest of the industry means that the rest of the industry fails to fully capture the value of AI. Yes, there are chips made in Arizona, but only a portion, and they need to be sent back to Taiwan for packaging and testing.  ↩ Yes, there are chips made in Arizona, but only a portion, and they need to be sent back to Taiwan for packaging and testing.  ↩

Security AI

▲ 0

0 views

Stratechery 2 weeks ago

2026.04: Netflix, Luka, and Life in La La Land

Welcome back to This Week in Stratechery! As a reminder, each week, every Friday, we’re sending out this overview of content in the Stratechery bundle; highlighted links are free for everyone . Additionally, you have complete control over what we send to you. If you don’t want to receive This Week in Stratechery emails (there is no podcast), please uncheck the box in your delivery settings . On that note, here were a few of our favorites this week. This week’s Stratechery video is on Apple: You (Still) Don’t Understand the Vision Pro . Netflix Questions. Netflix reached all-time highs last summer, as the saturation narrative of a few years ago was shattered by continued subscriber growth. Six years on, and Wall Street is getting nervous: do all of those subscribers actually watch Netflix enough, such that Netflix can show them ever more ads and/or charge them ever more money? Those worries were only exacerbated by the intended acquisition of Warner Bros.: is the Netflix growth story nearing its conclusion? I asked co-CEO Greg Peters all of these questions , including how and why Netflix is always learning and willing to change, and why the scarcity of compelling storytellers means that professional storytelling will endure in the face of user-generated content and AI slop. — Ben Thompson What TSMC Means to AI. For all the ink spilled on OpenAI, Google and Nvidia as the main characters driving investments and fascination in the middle of the AI boom — this Stratechery Article , for example — it can be easy to forget that all of them are dependent on the same company to fab the chips that make any of this acceleration possible. Wednesday’s Daily Update explored that dynamic through the lens of TSMC’s capacity constraints, including questions about restrained capital expenditures over the past few years, and the tension implied by the news of the company’s planned investments from here. We went deeper on an excellent episode of Sharp Tech , with an extended discussion of TSMC’s past struggles to wield its pricing power, and why AI companies of the future should be answering TSMC’s restrained spending by working to build up competitors.  — Andrew Sharp And Now, Some Basketball. A useful heuristic for modern life is to keep in mind that anytime the Internet appears to be in unanimous agreement, the truth is more complicated than what’s being presented on Twitter. Such is the case for the Luka Doncic trade, which shocked the world almost exactly one year ago, and was greeted last year with near universal certainty that it would be remembered for generations as an inexplicable disaster. In this week’s Sharp Text article I wrote about why I disagree , and why a trade that was indisputably a short term mess for Dallas may yet look more explicable as the years pass. Meanwhile, on a nearly two hour episode of Greatest of All Talk on Thursday , Ben Golliver and I had a bad time discussing a brutal Jimmy Butler injury, and then a great time making our All-Star picks, marveling at a Lakers ownership mess, and fielding listener reports on what the NBA looked like in London last weekend.  — AS Ads in ChatGPT, Why OpenAI Needs Ads, The Long Road to Instagram — OpenAI finally announced that ads are coming to ChatGPT. It’s an important step, but one with far more risk given the delay — and the delay means the ads aren’t yet the right ones. TSMC Earnings, The TSMC Brake Revisited, Why AI Needs Foundry Competition — TSMC admitted that it has invested too little in the face of overwhelming demand for AI; that’s why the industry needs to facilitate competition for the foundry leader. An Interview with Netflix co-CEO Greg Peters About Engagement and Warner Bros. — An interview with Netflix co-CEO Greg Peters about engagement, competition, and the Warner Bros. acquisition. Was Nico Harrison Wrong? — Looking back at the volcanic backlash to the Luka trade, and looking ahead to the future for Doncic in L.A. More Vision Pro, ChatGPT Ads TVs and Siri Chatbot How Westinghouse Lost its Way Why Diamond Transistors Are So Hard Canada and the New World Order; Second Order Questions for the US and EU; More Big Names Purged from the PLA? The All-Star Format and Starters, KD Passes Dirk, The Most Depressing Trade Deadline Ever?  Jimmy Butler Out for the Year, All-Star Reserves in Both Conferences, Trouble in Paradise for the Buss Family? A Call to Action for TSMC’s AI Customers, Wall Street’s Netflix Anxiety, Q&A on Tech’s Cignetti, OpenAI, Starbucks

Business Media

▲ 0

0 views

Stratechery 2 weeks ago

An Interview with Netflix co-CEO Greg Peters About Engagement and Warner Bros.

Listen to this post: Good morning, This week’s Stratechery Interview is with Netflix co-CEO Greg Peters . Peters became co-CEO of Netflix in 2023, and was previously Chief Operating Officer and Chief Product Officer. I interviewed Peters in January 2024 . Netflix reached all-time highs last summer, but is facing increased skepticism from Wall Street, particularly after the company announced its intention to acquire Warner Bros., which represented a stark departure from the company’s build-not-buy philosophy. In this interview we discuss the Netflix roller-coaster, why engagement is the metric everyone suddenly cares about, and how live events play into that. Then, we get into the Warner Bros. acquisition: why is Peters convinced that this is a home run, and what are the company’s arguments as to why regulators should allow this acquisition? As part of this discussion we get into the overall structure of the industry, including questions of bundling, professional versus user-generated content, and why Hollywood should be embracing Netflix instead of fearing them. As a reminder, all Stratechery content, including interviews, is available as a podcast; click the link at the top of this email to add Stratechery to your podcast player. On to the Interview: This interview is lightly edited for clarity. Greg Peters, welcome back to Stratechery. Greg Peters: Thanks for having me, it’s great to be back. So it’s been two years since we talked, which was actually pretty surprising to me, I felt like it was last year. A lot has happened since then — just from a stock market perspective, you are, congratulations, up 45% since the last time we talked. GP: (laughing) Thank you for noting that frame of reference, I appreciate it. Unfortunately, you’re down 37% from last summer. I thought that ending the regular reporting of membership numbers, which would immediately set off like a 10% jump or decline, or whatever it was, would flatten out the rollercoaster — it feels like the rollercoaster has gotten steeper and crazier. Did you expect that? GP: No, not expected. You might argue that that was a self-induced wound as we continued to try and do creative things to advance the business. But our job is to move things forward and push the edges and we’ve done a bunch of that stuff and I think you’re seeing some of the chaos and the reaction in that. Yeah. Well, one of the advantages of the member number is it was a very direct line from those numbers to the success of your business and part of that was a function of you had a pretty simple business. You sold subscriptions, and how many subscribers you had was a direct line to your bottom line. Now it’s a bit more complicated . You’ve urged Wall Street to focus on your financial metrics, but of course people want to know what the drivers of those metrics are. To follow on to that point, did you see this sense of a lack of clarity and uncertainty coming? Was that something you anticipated when you thought about changing how you report your earnings? GP: Well, the specific subscriber piece was because we were evolving from a model that was really simple. Literally there was like X times Y and you got to the business and as we were doing things like introducing extra members, so you’ve got different membership components. Obviously ARM [Average Revenue per Membership] looks very, very different across the world and there’s increasing span in terms of what that looks like in different places. And then the advent of advertising revenue, because now you have this whole other revenue source, which is not directly captured by that math, we knew that that subscriber was going to be less and less complete and trying to drive that out. And so to some degree, the internal way we say this is like, “We’d love to people start to treating us like a business”, and businesses ultimately are about revenue and profit, and that’s the numbers that we should look at, and certainly the numbers that we manage too. Right, I think the problem is people want to get to and know about what the drivers of the business are. GP: And those are the drivers so I think to some degree, the business is not so complex in its evolution that those drivers are not clear. Then I think we’re trying to provide enough structure behind that that says, “This is how you can assess how we’re doing”, and so in my mind, I don’t think that there’s a lot that we’re missing in terms of that. Now, obviously, to some degree, we want to insulate, or maybe I’d say protect, from a competitive perspective, some of the things that we are doing because we don’t want our own competitors to know what’s going on and copy those. There is a lot of talk by you and by Wall Street about engagement. As someone who, my bread and butter, maybe more than any other company, has been covering Facebook is one that I’m very familiar with. On one hand, this is kind of a bullish discussion because Netflix’s revenue per hour viewed is the lowest amongst your streaming peers, which suggests you have latitude — that’s another way to increase your business is just increase prices. On the other hand, there is concern that engagement growth is barely budging, which means your capacity to raise prices is maybe limited, maybe you kind of got locked in too low. Which is it? GP: Well, I would say we really do think about that revenue per engagement as an important sign of where our upside potential is and I don’t have any reason to believe that, over a period of time, we shouldn’t be able to meet or exceed essentially other revenue per engagement kind of metrics. There’s no first principles argument why that’s different, but I do also think that there’s complexity behind — we can talk about total engagement and I’m happy to tease apart the different components of that. That’s one of my questions because you actually spent a lot of time on the last earnings call saying there’s different types of engagement and trying to land this point. Like live, for example, drives buzz, drive signups, is that what you’re driving at? GP: I think there’s a bunch of different components of this. So there’s one, all hours are not created equal and this is a statement that we’ve made in the business for almost 20 years essentially. And then now knowing that that’s true, because we all know as viewers that that’s true, you’ve had hours that you think are transformational to your life that you’ve seen on TV and hours that probably didn’t matter that much. But getting to the point in the business where we actually understand that and can encode that value in a way that allows us to manage the business, that’s a whole different proposition. But we’ve made incremental progress around that. So that’s one dimension of difference is that like, what is the value delivered in the various different ways that we actually derive that value? Isn’t it fair to say though, Netflix’s value has kind of been in the — I don’t want to call it low value engagement, which by the way, I think is actually a tremendously valuable place to be, there is value in bulk and knowing that there’s just a large amount of stuff that will fill a large amount of hours. Even if on the flip side, and we’ll obviously get to this property later, I believe the streaming service with the highest, the opposite metric, the highest dollar they get per viewed is in HBO Max. Which you think about HBO, you think about certain shows, you don’t need to watch it all the time, but you want to get whatever the hot new show is and that’s how their model works. GP: Yeah. You and I have disagreed a little bit on this. I think it was off the record though, so I had to bring it into the discussion! GP: I would say I think we’re operating quite across a wide spectrum and I definitely believe that we are delivering hours that people love. You think about the Stranger Things finale, those are very valuable hours that I think people were desperate to enjoy. Also, I think we’re doing things in live events. You mentioned live events that are very punctuated moments that I think are very, very high value hours. So look, there’s no doubt that we want to do more and we want to do more across quite a wide spectrum. I think it’s also worth noting that each individual values their entertainment differently. So what might be low value to you might be high value to somebody else and so part of what we’re trying to do is capture those components. But HBO is, it’s got great content, so there’s no contesting that, I think that’s one of the reasons that we were excited about doing the deal , for example, as well as doing things internally like producing more Bridgertons and more Stranger Things, and all those things that we also think are high value that come from our own development. Yeah. We’re going to get a lot to the deal later, as you can imagine. I do have a bone to pick with you with the live thing. GP: Okay, great. I think you said in this earnings, you’re very explicit, “Oh yeah, these live events drive buzz, drive signups”, which was one of my arguments in favor of you doing these live events a few years ago. But I felt like you got an earnings call like two or three years ago and you’re like, “Oh, it’s so small, it doesn’t really make a difference”, and now you’re like, “Actually, no, this is very important in this regard”, I need more consistent clarity from you guys in your earnings to back up my thesis. GP: Okay, great. Well, I think maybe we can try and provide some of that clarity right now. So in terms of the total number of hours and the total number of investment, live is a small fraction of our total content portfolio, so that statement is just fundamentally factually true. I think of this as the cake and the icing maybe, which is, while small, it’s a great complement to these other kind of things that we have. And as a portfolio, it makes up an increasingly important part of that, so it’s great to have these punctuated moments, they do drive acquisition. But to be clear, they’re not driving a majority of acquisition. So if you were to say, in terms of the materiality of the business- So it’s the amount of acquisition relative to the time is very high. GP: That’s exactly right, that’s a very good way to describe it. But has that been a shift of you recognizing that, or did you always recognize it and now you’re just trying to talk more explicitly about that? GP: No, I think we had a thesis that that was going to be the way that live would work the business, but now we’re seeing it in practice. And oftentimes we go from, “Okay, this is the hypothesis, we’re starting to see confirmatory data points”, and then the next step you get to is you’re actually building the model for, “What does that value look like?”, and that model looks different for a live event than it would look like for a returning season of Bridgerton, let’s say. Has it looked different than you expected when you formed those theses? GP: I would say it’s roughly consistent, but I’d also say, look, just be frank, we’re dealing with a reasonably small number of data points, so I don’t know that I would want to conclude super broadly. One interesting dimension for us is to be doing now live outside the United States, it’s going to be interesting to see how that works in different markets, what matters. World Baseball Classic in Japan , yeah. GP: We’re trying different kinds of events. We see some live, WWE is a perfect example, which really, it’s a retention driver. You’ve got a core fan of folks and they are going to watch that again and again, again, and that supports improvement and retention. Then you have these other live events that are really more about this cultural conversation, moment in time, acquisition, and buzz really. And so we’re building this out and learning more as we go. I think you’ve been doing the NFL one , I think this is the second or third year, one of my pet theories was it wasn’t just the NFL, but the NFL on Christmas is a time when older people have their children in the house who can show them how to use Netflix. Was that a thought at all? GP: Not particularly. I would say it basically gets more to the events angle that we have on sports, really, which is that, again, we’re trying to build a strategic understanding of how sports fit into this live strategy and we really have come at it from the angle, which is that we want to think about it as an event. And how do we bring more specialness, if you will, around it. Whether it’s adding talent to a football game, let’s say, or creating events that have never existed before, and coming up is an amazing free climber who’s going to free climb Taipei 101 . You know Taipei 101 very well, obviously. I do know Taipei 101 very well, yes. GP: So, this is the kind of thing that, we’re building these events from scratch, so to speak. So, I would say it’s more about that, that we thought about, “Why do NFL on Christmas?” — because that’s a special moment to do NFL. Do you think you would have this depth of thought and analysis of engagement in the different types of drivers absent the introduction of advertising? Or is the tail wagging the dog here? If you’re in ads, you have to understand engagement? GP: No, this was a discussion that we were having and trying to wrap our arms around way before we ever launched advertising. We’ve always carried this sense that we want to understand the complexity and nuance and engagement, that’s important to our business. We all felt like, “Okay, these are different things and it’d be great for the business if we could understand how they’re different and really attribute value to them in the right way”, and it’s very hard. It’s incredibly hard to go do this, even though it’s so intuitive, and then, we’ve been iteratively building our way, over a decade, into a more evolved and sophisticated understanding of that and really I’d say decade-and-a-half. Did that drive you into advertising, that sort of understanding? GP: Not per se. I would say, look, bluntly, we had debated advertising a bunch of different times and decided not to do it because simplicity, focus, run hard against the business opportunity we had. Until we realized, “Oh, actually, we need to start opening up the aperture and bring more into it” — and quite frankly, advertising has been incredible because it’s just such a natural way to bring lower price points to a broader set of- It’s a win-win-win. GP: It’s a win-win-win, exactly. And then, what we’re also finding is that technology in ways that you would be very familiar with are making actually the ability to provide a better advertising model, not only for consumers, but for brands, that there’s a huge runway of innovation that we’re going to do there. Would you say, and I’m not going to say this question is about another company, but maybe it’s about another company , that this anti-ad sentiment that gripped some segment of Silicon Valley, such that maybe there’s been a reluctance to go into ads that was unwarranted and actually should have happened sooner? GP: I think there is an understandable reason for that reluctance because I think that you and I can both cite multiple examples where businesses drove an ad model in a negative way. I think actually linear TV is a pretty good example of this, where it really got about revenue extraction at the cost of the experience and the consumer and it’s a bad business decision because it’s a short-term, essentially, optimization, rather than a long-term optimization. But we, as consumers, all have this experience and so I think that that sentiment comes from those experiences. I think I understand why that is the case, but of course, you don’t have to do that. You can actually build, as you said, win-win-win advertising models that really work for consumers as well as for advertisers and the business. Well, you just led me to — this is the exact concern, I asked someone else about who I was talking to today, about questions he had. He’s like, “Look, the concern I have…”, I think he brought up Viacom specifically, this idea where flattening engagement in a business that is measured on business metrics, on financial outcomes, leads to a temptation, to your point, to push too hard on those financial drivers, such that you achieve them, but then the experience deteriorates, engagement gets hurt more and it’s a vicious cycle. How do you make sure you avoid that? GP: That vulnerability always exists in business and outside of ads, we can cite multiple examples. Let’s take another one, like easy cancel, we’ve always believed in easy cancel, I can guarantee you, we can make- Right. You don’t even have an annual plan still, right? GP: No, we don’t and basically one-click cancel. If you don’t want to subscribe, we should make it easier for you to not subscribe and we believe that that’s a long-term business positive decision because basically saying, “We’re going to make a great consumer experience, we’re going to build trust with consumers, we’re going to let them try the service, if we’re not delivering the value, they should be able to get out and we want to hopefully get them back, win them back when we are delivering that value”. That’s another example of we could do a bunch of definitely short-term optimizations that would work for quarters, but it would be bad for the business. And so, it’s just being disciplined that you got to have a long-term view and not do those shortcuts and really win for the long-term. A quick question to go back to the sports and live events, is the goal to have at least one must-see live event every month, say, for the sports fan? Where if you’re a hardcore sports fan, no, you’re not going to have the full season of NFL for reasons that are apparent, you don’t want to be a renter or whatever it might be, but if you’re a sports fan, you’re going to keep Netflix year round because there’s going to be something every month. Would that be the ideal outcome? GP: Yeah, I’d say that’s roughly the working hypothesis. I’d say it’s slightly different than what you just said it, but like for these virtual audiences that you have around the world in different segments with different entrants, essentially, we’re trying to have an live event at enough frequency a month is probably a pretty good metric. Well, that’s when the subscription period is. GP: Yeah, it connects with that obviously and we could query whether or not six weeks, eight weeks does a lot of the same work, but essentially, the whole idea is you should always have something you’re looking forward to, right? And something that is something that you can’t stack and bank, you’re not going to let it sit and then six weeks later… Right, sign up for one month, “This is my Netflix month, I’m going to plow through everything”. GP: Exactly. So, that’s definitely how we’re thinking about some of the value that live can add. What is driving the increase in advertising revenue? I think you increased— GP: Two-and-a-half basically over ’25, 2.5x over 2025, and then we’re targeting double- And then 2x next year. GP: Exactly, you got it. So, okay, what are the drivers of that? One driver is just inventory growth. We’re getting more subscribers on the ads plan, that means there’s more inventory to sell. Is the vast majority of your subscriber growth on ads plan? GP: No, I wouldn’t call it the vast majority. The majority of our growth in the ads markets is on the ads plan, but vast majority I think overstates it. And then, the other dimension of growth is just better monetization of that inventory. We’re rolling out more ads products, better measure, all the things that you go do, better targeting. We’re actually getting to the point where it’s going to be really interesting where we’re going to see some amount of personalization of ads or ad-specific creatives and things like that that’ll be partly based on the data that we have. Obviously, thinking about privacy, data safe, secure ways to do that, but when you improve those things, you get better monetization of that and those are the two drivers of the ads revenue growth. How has the ad revenue growth done relative to your expectations when you started advertising like three or four years ago, versus a year ago when you reset everything, your own ad server and things along those lines? GP: I think of that trajectory as being a very common trajectory that I’ve seen in launching new things at Netflix, which is that you start with a good first principles argument, like, “This all makes sense”, you launch it and then you realize that there’s 20 things that you got wrong or that were harder than you expected and such, and then you start working those problems. Maybe it’s a little slower at the beginning, and I think about like launching Latin America or the first time we were really going ex-US, there’s a lot of problems we ran into that we didn’t anticipate. And then, you start working those problems, then all of a sudden you start to get to a faster trajectory than you actually anticipated because then you start to build optimizations. You get all the catch-up. GP: Well, not only catch-up, but there’s optimizations that you hadn’t really built into the model from the beginning that you’re now applying to it. And so, as is usually the case, the map is not the territory, the model is not reality and you initially underpredict, or you fail to catch all the things you have to go do, but then you also fail to catch the things that you can do that are better. Yep. This is my favorite anecdote of the interview because last time we talked I was like, “Oh, yeah, I felt that in my subscription business”, I felt that curve you’re talking about in my subscription business too, so I think I know what you’re talking about. You did spend a lot of time on the call, one thing I picked up on, talking about total viewers as opposed to members. You have almost a billion total viewers, is that an advertising-driven metric that’s in your head now or is that just the nature of how you’re changing your subscription models or what? GP: I think it’s a component of two things, because actually, as we report the per household membership number, one component of this is extra members, which, again, is not a material mover or a significant mover of that number, but it’s a component of it that actually gets to more people because we don’t treat that as a separate membership. We treat that than the total household numbers. And then, to your point, advertising has made us think more about like, “Okay, actually, we have to think more about the individual viewers and what’s that number?”, obviously reporting that total number becomes important for advertisers because they’re looking at that. That’s a place where advertising is much more influenced how we think about what’s the total viewer count rather than the household subscriber number. Is there any room for advertising to the non-advertising tiers, whether it be on the home screen or something like that, or is that absolutely a no-go? GP: Well, no, I would say it’s back to the long-term thinking, right? We want to do the right thing for members, which yields really good long-term business results, we think. Now, to be clear, we’ve given advertising to non-advertising plan members on certain types of shows back to your NFL games on Christmas. NFL advertising just on the format’s almost been built together and so, those are natural and we expect that — pulling them apart almost creates a worse experience. Totally, absolutely. Because I lived internationally for a long time, so I would watch lots of League Pass games or whatever the equivalent of the NFL was, and I found it distracting and annoying to have dead time in between. I preferred — I’m not going to say how I got streams — but I preferred streams from the US. GP: It’s bizarre right? Anyway, so that’s an example of a place that we think that the type of content and the format is right to put ads on, even for the non-ads tier folks. Why did you buy Warner Bros. ? GP: Well, I would say one, worth noting that we came in from this perspective of we’re builders, not buyers, so we have a default orientation toward that, but we also thought it’s our responsibility to look at everything and we got in there and looked at what Warner’s had and we were like, “Wow”. We started building the value case for this and we got to impressive numbers and there’s really three big centers of value that come from this. One is there is a component of theatrical that really works and is complementary with the streaming model. We see that because we do this with Pay-1 deals all the time, right? We did a big deal with Sony, for example, that we just renewed , so we know that those work synergistically. But it’s nice when they’re paying for the marketing, not you. Then you sort of get a free ride. GP: Well, I wouldn’t call it a free ride because I think you’re encoding the value that’s created in that title in the licensing fee. If the market’s working efficiently, when they spend money on marketing to build value in that title, we’re paying for that in terms of the license fee, right? Right, good point. It’s worth more, the IP’s worth more. GP: Yeah. I would just say that this is a question of that that was externalized before and this is internalized. In building internal business, it’s all the things that we just talked about. You have to go spend energy figuring out what you don’t know and getting proficient at it and building that and that’s just something that we’ve looked at it multiple times, but we always just said like, “Hey, rack and stack the priorities, this is the cycles we’ve gotten”. But if they’ve already built the capability, then it’s a different question. GP: 100% and we want to leverage that capability and so we don’t want to blow that up or destroy that in any way, shape or form. The production side of things, we use them as production partners right now, they’re producing shows for us as we speak. They’ve got great production capabilities, they expand our capability to our infrastructure and the capacity we’ve got production. They’re very good developers of IP as well, so it’s a bunch of great capabilities that are complementary there. And then you also get to the HBO service brand programming identity. You mentioned it before, HBO is a real acme of quality and it signals to consumers that shows that they’re going to get. We think that that is an incredibly important opportunity, essentially, to elaborate our offering, to think about how we assemble plans and build the capability to actually give different folks what they want. We also think that those assets, the titles that are in Warner Bros. writ large, film, series, etc., they’re being underexploited right now. We have a global footprint that’s bigger and a better streaming service capability and knowledge to leverage those. This all makes sense, but at the end of the day, it’s still $83 billion, so it has to be accretive. I mean, you can earn incremental revenue through — this could help you get more subscribers, it could help you get higher revenue per subscriber, it could drive more engagement, you can show more ads, you can save on content acquisitions and licensing costs. Where does this pencil out? You don’t get the escape of, “All the above”, or does that mean you’re not going to be able to answer? GP: The reality is that it is all of the above, and so we build a model. And the model, it doesn’t have one driver, it’s got multiple drivers of value, you’ve listed a pretty good taxonomy of those drivers. In reality, but if you were to say, “What is the biggest one of those?”, I do think that if you put your finger on one thing, there’s quite a library, a significant library of content of existing shows that are out there. This is not speculative, we’re going to go do something with the IP or all these different things, but you have existing library of content. Right now, it’s not getting as much viewing as it could, we know we can drive more viewing on it, and this is, again, just to — we do this all day long, is basically take a look at what would a title do on our service? Because we have to do this when you think about licensing and what’s our valuation model there, willingness to pay, etc, so we basically use those same mechanisms, those same models to assess, “Okay, how would all this library of content do on Netflix?”, and when you drive that viewing, and that’s drivers of our business in all those different dimensions that you’re talking about. That’s the whole point, this is all full circle, right? So even if that’s all true, you can drive all this viewing. How does that actually manifest in your financial results? GP: It’ll be many things. So it’ll be basically improved retention, we will see more subscribers because of it, it’s again back to the things that we listed at the beginning, they’re all there. It will drive more advertising for the folks on our advertising plans and have ads associated with that, so that’s all certainly the case, more value delivered. We actually think about when we deliver more value, we go to what’s the right pricing model associated with that too, so that’s a component as well. But it’s also, there’s different aspects of this that maybe we actually don’t talk about that much but we see on how we outbound our service into other service offering bundles. If you get a package with a pay TV provider that includes Netflix, there are all sorts of benefits that you get in terms of retention improvements there, we think of this as the same thing. We’ve got a bunch of subscribers to HBO, let’s say, the majority of those are subscribers to Netflix too, so we actually see that by putting these two things together, we can actually improve the offering for consumers. Will those still be two separate services or would they be together in one? GP: This is the kind of thing we would want to sort out and we’ve got some thinking originally. I knew you weren’t going to answer, but I had to ask. GP: Yeah, we want to do some more work on that one. But there’s clearly, I would say we know from our work in doing this with other providers that there are benefits. We can make a win-win, we can make a better product for consumers, lower price ultimately, and it works better for the business, we know that that’s capable. What does Wall Street miss about this? They don’t seem very happy about it. What do you think is the missing link there, that you feel such high conviction that you change from, “We’re builders, not buyers”, to, “No, we’ve got to get this”? GP: Well, to be clear, when you say, “We’ve got to get this”, actually, I would not agree with that. Got to get this at a specific price. GP: Yeah, exactly. And it’s like all the deals that we do, we understand what the value model is. We compete aggressively up to that approach and we take the same disciplined approach and almost everything we do. So again, this is me speculating for investors and you can talk to them directly, but my general sense is that investors don’t like uncertainty. You go into this limbo period essentially while you’re waiting for the deal to come to fruition and that’s an awkward place to be. I also think it’s fair to say it’s probably right for folks to bring some skepticism around our ability to execute a big integration. We’ve never done it before, so folks can look at that and say like, “Okay, these folks have a demonstrated track record of being builders, but they don’t have a demonstrated track record of being buyers”. Warner Bros. doesn’t have a great record of being acquired. GP: And there is that. Obviously, we asked ourselves these same questions and we said, “Okay, when we look at these models”, we’re like, “That’s great, but can we execute this?”, and at the end of the day, we looked at it and said, “When we look at where are their sensitivity to those value drivers, this is the business that we’re in, can we bring content onto our platform and monetize it? Yeah, we do that every single day, we know how to go do that”. There’s discreet components which are new to us, we’ve never run a theatrical business before. That’s fair, that’s a totally fair comment, but we’ve got a team that’s done that, that knows how to do that, and we don’t think that the complexity of bringing that group in is that high. Given their record, you’d probably rather Warner Bros. run the theatrical business than you run the theatrical business. GP: They’re good at it, I’ll just say that. They’ve got a great track record, they’ve been doing an amazing job. I do believe you about the theatrical window, I think there’s some skepticism about that. I think one of the things that [Netflix co-CEO] Ted [Sarandos] said in the earnings call that I actually had already planned to ask you about was the extent to which nothing at Netflix seems to be sacred. You’ve changed your mind about a lot of things from DVDs to original content to sports to acquisitions to ads, so why not the theater? I think I actually asked you this last time, but it seems more pertinent than ever. Why is Netflix able to change its mind? GP: Well, I’d say we’ve often said we believe in two things, capitalism and that users first. Well, That was part two is what are the things that you won’t change your mind about? GP: Well, I think those are pretty important so I’d put those on the list, but there’s another way to look at this, which is that we’ve often had what I characterize as a scientist mindset, which is that you have a case based on the data or based on the conditions that you’re operating in, and then when things change, you change that. And for us, we often change very quickly, and I think that that speed of changing throws people off because they look at this as like, “Oh my gosh, they just changed everything” — well, yeah, if the world changes, we change to respond to that, and maybe the difference is that we just do it very, very quickly. You’re CEO number two or number three, I don’t know, 2A, I don’t know how you would count it. GP: I’m not sure either. How do you maintain that? Does this go back to the culture doc back in the day ? Are you able to keep this in the long run? GP: Oh, it’s definitely in the DNA. When I joined, it’s one of the things that I loved about the company, which is that attitude. I know that we talked about some of these big pivots which are maybe more recent, but it was going on way back when, so in smaller ways that maybe were hidden, but you’d have a position around how does — we did free trial, I don’t know if you remember free trial, we had that for a longest period of time, and then we did a bunch of pretty hardcore science on, “How does this work and is it really working?”, and we’re like, “Oh, actually it doesn’t work as well as we expected”, so we just changed. It’s a good example. On the regulatory point, I have two questions. One, to go back to something that Ted said in the call. He said, and I’m quoting, “This is really a vertical deal for us” — explain that. How is it a vertical deal? GP: If you think about the components of this, so there’s the theatrical, the production. These things essentially verticalize the business that we’ve got and we’re the distributor for those components, and then you look at the HBO piece and that’s where I mentioned, sometimes you hear the people do the math, which is like, okay, you take our 325 million subscribers, you take 100 million from HBO and it’s 425 million subscribers. That’s not the reality of it, the reality of it is that the vast majority of those HBO subscribers are already Netflix subscribers. Right, but so that part’s a horizontal part though. GP: And that part, to your point, so that’s distinct from this other one, and that’s where we would argue though it’s a complementary approach where the people that are going to subscribe to one are the people are going to subscribe to the other, and actually, by bundling and packaging them together, we can provide a better offering for those consumers as well as expand that offering to more places around the world. There’s a bunch of countries around the world that there is no HBO Max, there is no service there. There’s countries around the world where they have no plan today to get to and that we’re already operating in, and we can obviously bring an audience for that service and for that content. The second point is Ted made the point that TV, “Isn’t what it used to be”. Who are your competitors? GP: Well, you’ve said this and I agree with it, that time and attention are the scarce resources, and we really believe that. And you’ve heard us say this for probably a decade plus. Actually, in your defense, literally the next question — one, I agree, the only scarce resource is time , and two, this is a point you guys have been making on earnings calls for a long time. Is that your saving grace, that you’ve actually been saying that for many, many years on your letters to shareholders and whatnot? GP: We’ve always thought about it that way. We’ve always thought this is how we have to compete, and it’s fair to say, I think when you get into then a regulatory question, the market definition becomes more precise. And when we look at this, if you see what people, what consumers are watching on TV, you know the numbers on YouTube, they’re watching a lot of hours on YouTube. The programming that they’re watching on YouTube, YouTube is now doing NFL games too, it’s going to do the Oscars , you can watch long form movies and series on it, the BBC just did a deal with YouTube in the UK as well as Channel 4 and ITV, so there’s just all this examples where I think it’s becoming a much more blurred space. You’ve got tech competitors, Amazon with MGM and they continue to invest in Prime Video, you got Apple doing big movies, and so there’s competition at all levels, multiple places for creators to bring their opportunities and we look at that writ large and that’s where we think about that’s a very robust and broad competitive space. But YouTube is the threat , right? That’s really the big one. GP: We think of them as a very important threat, I would say we’re competing with all these different folks for sure, but YouTube is the formidable competitor. They drive a lot of hours, they have a great model, and so it’s no doubt that they’re going to push hard on this. You always report them as YouTube on TV. How large is the lead if you were to include mobile and computer and all those other things? GP: It’s significantly bigger, and again, a little bit, it comes down to we do think about time and attention as a very broad competitive landscape. But you also have to get a little bit tighter when you think about what are we going to go do to defend against that space and what are the spaces that we want to prioritize defending and how we think about that. What are the advantages of professional content versus user-generated content? And are any of those advantages structural and sustainable in the long run as the capability of producing very high quality content gets better and better? GP: No, I think they are sustainable, and here’s the theory behind it. Professional becomes a little bit of a subjective term, but maybe let’s just say that where storytelling and the capability, or let’s say the skill at storytelling at the highest level in the human population, I think of that as a fairly rare commodity. Not everyone can do that storytelling, and the ability to do that and do that consistently essentially is where we’re thinking about that demarcation. Our model allows us to compete more effectively for those world-class storytellers. Interesting. So it’s basically, there’s another scarce resource here, which is on the consumer side, it’s time and attention, but there is one on the supply side, and if you can pay upfront and have more of that venture type of model as opposed to you have to make it first and capture it on the backend, you get a sustainable advantage on the supply side scarcity as well. GP: That’s right, and that’s one version of the supply side. The other one is quite frankly that our model monetizes better on a per hour basis, and so to some degree, that translates into a better competition for those scarce storytellers. And those things, I think we just should just know. You get it, but just to make it explicit, those things are relatively independent of the production cost or production cost efficiencies, so that doesn’t really actually change that. You can have very highly produced terrible content. GP: Yeah, and we’ve seen it in a bunch of different places and those don’t typically do very well. Is AI slop going to save you? If it overwhelms the UGC platforms and basically it’s like you’re a refuge, so this is all actual, real. GP: I think it’s a credible — I don’t know if that’s the reality so I can’t say with certainty that’s where we’re going to land, but it’s a credible possibility, I think. You mentioned the huge value of Warner Bros. is the library. Can valuable library content even be made today? I think an analogy that you could make is that library IP, particularly from the ’80s and ’90s, that era, the Friends, the Seinfelds, all those sorts of things, it’s like the fossil fuel of streaming. It’s unbelievably valuable, it’s unbelievably useful, and we’re not getting more of it. GP: Yeah, I hear that a lot. I guess I try and put a broader historical framework on this, and I think if you go into the music space, I think you’ve heard that I don’t know how many times across multiple different generations and decades. I think there’s to some degree that the commentators, maybe you and I at our ages, grew up in those eras, and to some degree, we have an affinity for that content. Obviously the best era. GP: Exactly, that’s right, so we speak about that stuff. But I firmly believe that if you go 15, 20 years from now, people are going to be binge-watching Stranger Things for the seventh time, I think that that’s absolutely going to happen. The other thing too that happens is you get this compression effect, right? You stack up 10, 20, 30 years worth of- Right. I just put the 80s and 90s together. GP: Yeah, exactly, and you say all of the output of those two decades versus how we’re looking at the last five years, and so that’s just a little bit unbiased in that regard. Do we, or I guess I should say I, maybe have actually in the end underestimated how easy it is to switch and what I mean is I’ve had the overall thesis about the bundle fragmenting, which obviously happened, but then the assumption it would reform and the bull case for Netflix is that would reform around you. Is it unrealistic? You go back to the original bundle , that was downstream of towns who couldn’t get a broadcast signal banding together to put up towers. You had a physical constraint that forced everyone into an optimal business model that you probably wouldn’t have gotten otherwise. If you don’t have that constraint, because everything’s on the Internet, it’s just bits, is it unrealistic that that happens organically? And there’s a bit about acquiring WB where it happens financially because you have financial power relatively speaking thanks to your position, but it’s never going to come to you naturally without pushing it. GP: I think that there’s one other important dimension of this which I feel like is your actually theory, which is the other constraint is the attention and discovery constraint. So there’s the physical constraints, and also channel carrying capacity, a bunch of different things. This is just where you can monetize the content better because you can get people to watch it. GP: That’s right, and so I think there’s an argument that when you build that gravity, you build a better product experience, all those different things, we should be a more efficient monetizer of any given unit of content, and I think that’s our theory and that’s why we think there’s a lot of upside value in the Warner Bros. case. Now, you could argue that the market efficiently allocates that by licensing, there’s lots of different ways that you would unlock that. In theory, but how long is it going to take? GP: Exactly. I think also, when you get into the position where companies feel like this is an existential change and there’s just a lot of dynamics which can sort of gum up those works. And so then you start to look at, “What’s the most effective way for us to unlock that and accelerate that process?”. Is there other companies that should follow in this, assuming the WB thing goes through? GP: Meaning by other companies? Is that what you mean? Well, you say you’re a builder, not a buyer. Are you going to have a new identity going forward? GP: Let’s see how this goes. We should probably prove that we can go do this and deliver value there. But look, at the end of the day, we’ve mostly been on a process of expanding the number of mechanisms for adding value. We started off as licensers, that’s the only thing we had, it’s the only thing we did. We said, “Okay, let’s get into production”, we’ve gotten increasingly good at production, I would say, and being able to produce at a scale and diversity from a geography perspective that no other content creator really has done before. So that’s been a long, in some regards, many year process to build that capacity. We’re experimenting with other things like back to being what you might call a more classic aggregator in terms of taking broadcasters’ content writ large and just putting it through in a channelized model, so we’ll see how that goes. This is yet another mechanism. I don’t know how repeatable this mechanism is, realistically, I don’t know that we can go do this many more times, but mostly we’re just in the business of trying to find more ways to add more value into the entertainment offering. How satisfied are you with the homepage changes that you’ve implemented? GP: The most satisfying component for me is not even what you see, but it is what I and the team knows we’ve built in terms of extensibility and the ability to improve the offering over 10 years, and that’s the real value. You know how this works, you have a hyper-refined user experience, you do a huge shift and you’re just hoping basically it gets back to parity pretty quick. So you’re taking all that evolution over 10 years and you’re sort of starting from fresh. But what we see is the trajectory of changes that we’re driving through with the improvements we’re seeing. We’re like, “Okay, this is a great velocity beyond and I know that it’s going to return in multiple different ways over the years to come”. This sort of ties back to a question before. Does Netflix have a quality of content problem or a surfacing of relevance problem? Is this one of those things you can make the case, “Oh no, just a relevance problem”, and now you feel like you’re going to have the opportunity to prove that? GP: I’m not quite sure I understand the question. Well, now that you have this with this homepage, the way I understand the initiative generally is you’re going to be able to get a lot better at personalization, a lot better at understanding the customer, getting them stuff that is really interesting to them. There’s a heavy bias with shows in general towards recency, in part because that’s the easiest variable to understand and people generally like new things. But if I, a Netflix viewer, can get something from 2013 that is perfectly attuned to what I want to see, that’s more valuable than it being in 2025 or whatever it might be. GP: Yeah, okay, so a couple thoughts there. I think it’s really two dimensions of improvement. I think of the UI and the UI capability essentially is a force multiplier on the content offering we have. If we can make that 5% better, then all of that spend, all of that investment that we make has that return and that compounds with the ability to invest more in that content space, so we can add more that goes into a more efficient engine and so we need to do both. Just one thing on the recency component — there’s sort of two centers of value, and this is probably an oversimplification, but we’ve often thought about this as two centers of value. There’s a center of value you get from watching something that you find incredibly fun and compelling in a room by yourself, and that’s a great center of value and that gets to maybe your 2013 perfect show for you. But there’s a center of value that also comes from the fact that you can then have a conversation, or we can talk about that show you watched and I think why recency has this value and why live, to some degree, has this value is that we have that conversational value. We’re sharing something beyond the moment of just actually experiencing it. Now I have to ask, I wasn’t going to put it on here because I ask it all the time, one of the things Netflix has never changed is release everything all at once, you kind of broke it up a little bit with Stranger Things, it wasn’t the whole final season all at the same time. But if that is important to you to have that conversation, will that ever be — is that on the, “Might be revisited someday” list, or no, “This is going to always be this way” list? GP: I would say we wanted to think about where are the right moments to really have what we’ve known as that water cooler effect, and we’ve seen it in different places. But you’re playing with different trade-offs and there’s always trade-offs, and so we’ve also seen that the all-at-once release model, we think we have evidence, builds bigger shows as a result. So I think there are going to be these moments, and we want to do it in a way, again back, that’s customer first. Is there a bit where it’s better upfront for a show because that builds a show better, but once a show is big, then that’s the time to capitalize? GP: Could be a component of it, but I would say we really want to do this in a viewer and fan kind of way and we thought about the Stranger Things and we’re like, “This is, one, it was a lot”, the Duffer Brothers delivered a lot of content into that final season and we’re like, “We can do this and we can do it in relatively short order, so it’s not too far apart”, each of these units is a contained narrative moment, so it’s not like we’re leaving people hanging for the next week or things like that. And we think it’s going to be super fun to do it across three holiday periods. So again, it’s a lot of very specific situational, I think, thinking around that. Well, speaking of building big shows, I have to ask you about KPop Demon Hunters . Is this going to just be a fun memory from 2025 or will this have a lasting impact on Netflix? GP: I think it’ll have a lasting impact. We start to see the durability of this in terms of whether you look at re-watch or other signals in terms of folks out there, whether fan events and consumer products and stuff like that, it seems to be quite durable. And if you look at it compared to other sort of similarly situated franchises you might know from other competitors, we see the same volume of activity that it really indicates we think this will be something we’re going to talk about 10 years from now. At what point did you realize this is just going to blow everything else out of the water, particularly in terms of movies? GP: Yeah, it was relatively early on that we’re like, “Okay, this is performing quite impressively”, but then for me, the thing that it shows up is that when it shows up in your other, “normal life” in all sorts of interesting ways, you got Djokovic doing the soda pop thing at a tennis match and it shows up in SNL , and then you’re just like, “Oh my God, this has just permeated the culture in a way that you can’t stop”. There’s a narrative in Hollywood that Sony screwed this up by licensing it to you, my view is you screwed it up by not owning it entirely. Who had the bigger screw up? GP: Well, you could flip it around and say that we’re both quite happy that we had this culturally defining moment because I think that that’s true. Certainly we benefited from it in that regard and Sony, I think, is going to do a great job with it as going forward. There’s always things you can go do better, you always try and learn from it, but I would say we’re both thrilled to have had this moment. Just to circle back, to sort of wrap this up, I have a couple questions about you and Hollywood generally. To go back to the acquisition, what I hear from Hollywood is they’re scared, they don’t like it. Why shouldn’t they be scared? GP: Well, I think it’s worth noting that we’re talking about a storied asset. This is a studio that has produced some of the most amazing content over a century run, so it’s fair for folks to look at this in that way. I also think change is scary and every time you shift something up, there’s associated anxiety to it and I think oftentimes it’s like, “Let’s just keep the status quo”, right? But we also have to embrace the reality of things are changing, the competitive landscape is different and we as an industry, we’ve got to compete aggressively with the other forms of peoples’ time and attention and energy out there, and we want to go do that. We want to make this kind of storytelling even more compelling. That means presenting it through a great user experience that we just talked about, it means make sure that we’re producing, we’re investing in more production, we’re improving the quality of that production and so I would say, why is there a strong case for optimism? Because I think we’re going to be good stewards of that history and that heritage, and we’re going to be able to give those shows and the shows that are going to be produced by these people and by these institutions an even better opportunity to find amazing audiences and to create the stories that shape generations. The things that the people that are our kids will talk about 20 years hence as being the things that they thought were the most amazing things they’ve ever seen. Well, to that point, who’s harder to convince that your actual competition is, “Okay, if we must nearly define it, YouTube, but actually everything on your phones”: regulators or everyone in Hollywood, and to convince them that you’re all in the same boat together? GP: Quite frankly, I think most people know this, once you get into a conversation. They sort of know it in their bones. It’s just, it takes a while to come out. GP: That’s right. And I think it shifts from being an individual and understanding that, and then sometimes you have to get into this mode where you’re advocating for a group of people and understandably so. They should be using these opportunities to push their agendas as they should, so I think that’s really the difference. How do you keep the talent that makes, say, an HBO what it is? Because I’m sure, I would imagine people are circling already. GP: Actually, I feel very confident about that because we’re going to keep that HBO team and that HBO team is good at working with that talent and giving them the environment that they need to tell those amazing stories and they get to do it under a great brand that speaks to the kind of program they’re trying to make, and we’re going to give them a bigger audience. Well, and an actual functional business model. GP: Well, yeah. I would say that I think they’ve got a growing business model and we can just make it even more robust. What does a creator want? I think that’s what they want. My final question, and it ties into this, in the popular conception, you guys are the smart guys, you come in and you kick Hollywood’s rear end — it’s run by a bunch of stupid dinosaurs, they don’t understand how the world’s changing. As you’ve become the biggest and most important player in Hollywood, sort of to this point, and you’re no longer the upstart, what have you come to learn, if anything, that those dinosaurs understood that you didn’t? GP: I think it’s worth just noting that, especially when a business has been around a long time, it’s learned a lot, the industry has learned a lot, and so how do you give creators great environments to tell their stories? How do you make sure that their creativity- Was that a shift? You mentioned that before that there’s a scarcity in supply, there’s only so many people that can tell great stories, which is one of those truths everyone kind of knows, but you’re kind of not supposed to say to a certain extent. Did Netflix always think that way or was that something you had to learn? GP: I think we’ve thought that way for quite some time, and maybe in contrast to some of our other competitors that come at it from a different point of view. So we’ve always believed that, but I would say that that’s something that the industry understands very, very well and has done a great job at cultivating that. In my mind, for me, it’s just remaining trying to create a sense that there’s many things that we still have yet to learn, many things that we want to challenge our assumptions around, and so that we aren’t locked into a set of positions that are basically not fit for the purpose that we are facing in terms of the competition we have and the growth that we want. So I just want to create that constant sense of self-challenging around the positions that we hold. Greg Peters, great to talk to you again. GP: Good to talk with you, thank you. This Daily Update Interview is also available as a podcast. To receive it in your podcast player, visit Stratechery . The Daily Update is intended for a single recipient, but occasional forwarding is totally fine! If you would like to order multiple subscriptions for your team with a group discount (minimum 5), please contact me directly. Thanks for being a supporter, and have a great day!

Entertainment Business Media

▲ 0

0 views

Stratechery 2 weeks ago

TSMC Earnings, The TSMC Brake Revisited, Why AI Needs Foundry Competition

TSMC admitted that it has invested too little in the face of overwhelming demand for AI; that's why the industry needs to facilitate competition for the foundry leader.

Hardware AI Business

▲ 0

0 views

Stratechery 3 weeks ago

Ads in ChatGPT, Why OpenAI Needs Ads, The Long Road to Instagram

OpenAI finally announced that ads are coming to ChatGPT. It's an important step, but one with far more risk given the delay — and the delay means the ads aren't yet the right ones.

Business Marketing AI

▲ 0

2 views

Stratechery 3 weeks ago

2026.03: Technology Doings

Welcome back to This Week in Stratechery! As a reminder, each week, every Friday, we’re sending out this overview of content in the Stratechery bundle; highlighted links are free for everyone . Additionally, you have complete control over what we send to you. If you don’t want to receive This Week in Stratechery emails (there is no podcast), please uncheck the box in your delivery settings . On that note, here were a few of our favorites this week. This week’s Stratechery video is on AI and the Human Condition . What Technology Did for United. I am hardly the only technologist fascinated by airlines, so the opportunity to interview United CEO Scott Kirby was obviously one I was excited about. What makes United particularly interesting to me, however, is that my personal experience with the airline long ago convinced me that this is one of the most interesting tech stories around: United has become one of the best airlines in the world by investing in technology, which not only benefits all of its customers but has also let it invest in all the other parts of being a premium airline. It’s both a reminder of how technology is the only way to progress generally, and also how the real challenge is making up-front investments that may take years to pay-off. —  Ben Thompson What Technology Is Doing to Legacy Media. The biggest media story of the last 12 months is Netflix and its attempt to buy Warner Brothers; the biggest, cattiest media story among media , however, is Bari Weiss and her ongoing efforts to remake CBS News. On Sharp Text this week I wrote about the chaos surrounding Weiss and CBS , my theory on why she’s so polarizing, and why, with my tech podcaster hat on, I think her project to revive CBS News is doomed to fail. — AS What Technology Should Do For Apple. The Vision Pro finally has my dream content: a live broadcast of an NBA game in immersive video, and even better, the first broadcast featured my Milwaukee Bucks! Unfortunately, I found the experience incredibly disappointing : while the Vision Pro was immersive, as soon as you gained the sensation you were there Apple’s production immediately reminded me I wasn’t, changing cameras, cramming in studio shows, and generally treating an entirely new format as if it were simply TV. I am more convinced than ever that Apple has this all wrong: the Vision Pro doesn’t need bells and whistles, it simply needs to let me feel like I am there. In this case, less is more, both in terms of a single game, and also in the amount of content that could be made available. — BT Apple: You (Still) Don’t Understand the Vision Pro — The first live sporting event was broadcast in the Vision Pro, and it’s a big disappointment. The experience could be amazing, but Apple actively ruins it. Apple and Gemini, Foundation vs. Aggregation, Universal Commerce Protocol — The deal to put Gemini at the heart of Siri is official, and it makes sense for both sides; then Google runs its classic playbook with Universal Commerce Protocol. Meta Compute, The Meta-OpenAI Battle, The Reality Labs Sacrifice — Mark Zuckerberg announced Meta Compute, a bet that winning in AI means winning with infrastructure; this, however, means retreating from Reality Labs. An Interview with United CEO Scott Kirby About Tech Transformation — An interview with United CEO Scott Kirby about how rebuilding United’s technical infrastructure laid the foundation for transforming the airline. News in the Era of Irreconcilable Differences — The Bari Weiss era at CBS is probably not the crisis her critics allege it to be, but two fundamental realities make a successful revival look close to impossible. Vision Pro Frustration The Mac Icon Regression The Remarkable Computers Built Not to Fail What’s at Stake for China in Iran; Canada’s PM in Beijing; Notes on Chips, Apps, and Drones It’s Trae Day in Washington D.C., The End of An Era in Atlanta, Will AD Move Next? Rockets and Nuggets Stock Going Opposite Directions, The Wolves are Alive and Well, Buzz on Ja, AD, and MPJ Apple And Its Lack of Vision, The Transformation of United Airlines, Q&A on Grok, Meta, and Streaming Economics

Media AI Business

▲ 0

1 views

Stratechery 3 weeks ago

An Interview with United CEO Scott Kirby About Tech Transformation

Good morning, This week’s Stratechery Interview is with United CEO Scott Kirby . Kirby has had a remarkable career in the aviation industry. He was one of the chief architects of the industry’s consolidation in the earlier part of this century, and, over the last decade, has transformed United Airlines into a consistently profitable and aggressively growing behemoth. This might seem, at first glance, to be an odd interview subject for me. However, there are two interconnected angles undergirding my coverage. First, I have been a witness and personal beneficiary of Kirby’s changes; over the last decade the transformation in United’s product has been very tangible to me, and very much appreciated. Secondly, however, and more pertinent to Stratechery, is that it has been clear to me that United’s transformation has been driven first and foremost by an investment in technology. United first earned my loyalty by having by far the best website and app experience; that sort of loyalty is exactly what Kirby has been focused on, and it has in turn made it possible for United to improve many more aspects of the customer experience. In this interview we cover all of that, including Kirby’s background, belief in differentiation, and how United’s improvements were made possible by rebuilding the company’s tech stack from the ground up. We also touch on the challenges of competing internationally, AI, Starlink, and my own personal wishlist for my most used airline. As a reminder, all Stratechery content, including interviews, is available as a podcast; click the link at the top of this email to add Stratechery to your podcast player. On to the Interview: This interview is lightly edited for clarity. Scott Kirby, welcome to Stratechery. Scott Kirby: Thanks for having me. This is perhaps an odd venue for you given this is a site about technology, and you are an odd subject for me given you’re an airline executive, but this is actually an interview I’ve been looking forward to for a very long time for reasons that I think will become clear in our conversation. I am, however, like many of my readers, a bit of an aviation geek, and for what it’s worth, I’ve been a Premier 1K member for many years. SK: Thank you! You’re welcome. So, I guess that’s my conflict of interest declaration. SK: It’s a cool, sexy industry, even if it’s not technology. Well, it’s fun, because it’s so sexy from a product perspective and can be very not sexy from a financial perspective. SK: Yeah, we’re trying to change that. I also have some very personalized feedback for you, but I’ll save that for the end. Before I get into United and the tech angle, I do want to explore, I always like to open these interviews by learning more about my subject, you, specifically. Where did you grow up? When did you get interested in airplanes? I saw that you went to the Air Force Academy, so it was clearly quite early. SK: Yeah. I grew up in what at the time was a small town, Rowlett, Texas, not even big enough to have a high school, so I had to go to high school in a different city. But I early on decided I wanted to go to the Air Force Academy. I read a lot, and I had read history, World War II, and World War II was about airpower. So, I wanted to be a pilot, and then an astronaut. Went to the Air Force Academy, realized while there I was not going to be a fighter pilot, so I chose not to go to pilot training. Got my graduate degree in operations research, which is applied math, wanted to do something with math so just by accident wound up back in the airline industry, because I wanted to do something with math. But it’s been great ever since. Well, a somewhat related question, when did you learn to count cards ? SK: I used to play poker a lot, and when I got out of the Air Force Academy, I went down to Biloxi, Mississippi for four-and-a-half months, and I was going out on this international boat where you could play poker and it was just like shooting fish in a barrel. People were terrible, and I would make a bunch of money and a buddy of mine told me about, “Oh, you should count cards”. So I got the green book — I forget the title of the author, but it’s called the Bible for Card Counters [ Blackjack for Blood: The Card Counters’ Bible, and Complete Winning Guide ]. I learned to count cards, I went to DC working at the Pentagon, I started going off to — it’s self-taught, but started going up to Atlantic City, and then, of course, went to Vegas a lot. If you’ve ever read or saw the movie about the blackjack team that was in Boston, two different times when I was in Vegas, somebody came and asked me if I wanted to join their team, because it’s obvious to tell if there’s another card counter. SK: But, anyways, that was a long time ago. They’ve made it impossible now I think for card counters to be successful, but you could do pretty well back in the day. I’ll circle back to that, but you actually moved back to the airline industry with America West. SK: Yeah, America West. So you are a part of America West, a series of acquisitions that, ultimately, ends up at American Airlines. Tell me about that, that path, and how you rose through the ranks. SK: Sure. I was at America West, the number two executive, I got appointed to be the Chief Commercial Officer on September 10th, 2001. And what level did you join at? SK: I joined as a Managing Director of Route Planning and Scheduling. It was a small airline — like 80 airplanes at the time I got there, based in Phoenix, Arizona. A guy named Bill Franke , who is still around, owns a bunch of ultra low-cost carriers around the globe, I saw him last week when I was out in Phoenix, but was the CEO. Not an airline guy, he hired a bunch of us young kids. What I remember about the interview with Bill, at the time I was at American Airlines, at a subsidiary of American, it was doing well and I went to interview with Bill, and he said, “I’ve been told that you’re an up-and-coming star, and you’re going to go far at American. American is on top of the world, why would you possibly want to come to little America West?”, and I said, “Well, at American, they are doing well, they’re the best in the world, but if you want to change or try anything new, you have to prove beyond a shadow of a doubt that you won’t break anything and it won’t screw up, which is impossible, so you can’t change or try anything new”. And he said, “Well, everything at America West is effed up” — he didn’t say effed — “Is effed up, and you can do anything you want, and if it works, you get promoted, and if it doesn’t, you get fired”. I was like, “Perfect, that’s what I’m looking for”, so off I went. You obviously succeed, you rise up to number two, then you set off this series of consolidation. What was the industry pressure that was happening around that time? SK: You actually said it right. We, little America West, drove industry consolidation in the industry. The first merger was America West, the first I realized I was running the network and the commercial side of the business, we tried this experiment with the transcontinental markets, it failed miserably. Our most overlap was with the best airline in the world at the time, Southwest Airlines — we had to get bigger and change to survive. And so, the first deal we did was US Airways , who was in bankruptcy at the time. People described it, one it’s impossible to raise money for. I had 52 meetings with investors trying to raise money for that, some of your tech clients will probably recognize me going around trying to talk to people about raising money, until one finally said yes. We raised a billion dollars, and got the merger done. We did it in September of 2005. By the way, so, two small airlines merging, the same month that both Northwest and Delta filed for bankruptcy, so a really tough time in the industry to do it. SK: It was really successful, right out of the gates it was successful. A year later, we made a hostile bid for Delta Airlines, who was in bankruptcy and planning to come out on a standalone basis. That drove changes in their executive team and board — we failed in that, but— Right, you had no idea what you were creating. SK: Yeah. We knew we were creating a risk, or we knew that we were starting a ball rolling, we figured it was good no matter what happened. That caused Delta to then merge with Northwest . Next we went after United, and actually had a deal with United and I negotiated that deal along with their Chief Financial Officer who led the negotiations for United and I can remember going to dinner with her in Washington DC, which is where we’re all negotiating, we were even drafting the press release and she told me, “Just so you know, we’ve decided to leak this to test the market reaction”, and I knew we were in trouble when that happened, because it meant Continental would get in. Continental came in, and Jeff Smisek had one of the more infamous, he’s the CEO of Continental, quotes at the time that they decided on the pretty girl instead of us , and they merged with United , but we drove that to happen. And then, of course, American files bankruptcy, and we made a hostile bid against American . Everyone told us it was impossible, and we got it done. So, that little America West team, a bunch of young kids who really didn’t know enough to know that everything we were doing was just, in theory, impossible really changed the whole industry. It may have seemed impossible, but, like you said, it completely changed everything. I presume you’re on board that consolidation has been good for the industry. Perhaps it’s a different question, has it been good for the consumer? SK: It absolutely has been good for the consumer. You look at air travel today compared to where it was 20 years ago and it is a far better experience for you. You’re a frequent flier, it’s far better today. Back then it was just a heavily commoditized industry where the product, the service, the quality, the communication, none of that mattered. It was a pure, “Just get the lowest cost possible” industry, and it is much better for consumers today. And, by the way, it’s still a great value. In fact, I’ll just give you the most recent stats. In the six years pre-COVID to today, so, 2019 to 2025, airfares actually went down by about 2%, inflation was up 23%. Airport inflation, which drive our air costs, are up over 50%, airfares went down pretty significantly. That, by the way, is going to cause some ultra low-cost carriers to go out of business, already in bankruptcy, they’re going to go away, so a little bit of that is going to reverse. But airfares consistently for the whole period of deregulation have grown at much less than the rate of inflation, but the product and the service and the quality is infinitely better today than it was 20 years ago. To go back to the poker bit, you have a reputation for being very data-driven and you mentioned, “Well, back then it was shooting fish in a barrel”, was that the case when you started at America West? Was just by virtue of looking at the numbers, being buried in spreadsheets, was that why you were able to take over the industry? SK: No. We started with the weakest hand of cards at the table, we just played it the best, to keep the poker analogy, we had to bluff and bluster, we had the weakest hand, but we played it more aggressively than anyone. You had nothing to lose. SK: Well, we did, we had the whole company to lose. SK: But we just played it more aggressively than anyone else. We also recognized that doing nothing was a huge risk in and of itself. In fact, one of the things I say, “The biggest mistake most people make in their careers is never making big mistakes”, they’re so afraid of taking risk that they miss all of the opportunities and we were balanced about recognizing the opportunities, but also the risks of doing nothing — deciding on the status quo is a decision in itself. Actually, that reminds me of one of my favorite quotes is from Teddy Roosevelt that goes something like, “If you’re in a crisis, the best thing to do is quickly make the right decision, the worst thing to do is make no decision, and the second-best thing to do is to make the wrong decision. Do something!”. That makes a lot of sense. You do talk much more, though, about things like the customer experience, and things that are not necessarily data-oriented. Is that just part of softening the edges, as you know you’re one of the big three CEOs, or has it been a real shift in mindset? SK: Well, I think it is probably a shift in mindset, but it’s also something I’ve always thought. I am really good at data, I’m good at math, understand data, use data to inform my opinions, but data doesn’t tell you the answer. In the big questions, they’re uncertain and data helps inform your view of the world, but you can’t just rely on data. Doing something new and different is — you can’t ever use data to prove it, you have to trust your gut, trust your instinct. Two of my favorite quotes that are similar, but both Henry Ford and Steve Jobs. Henry Ford once said something like, “If you ask people what they want they’d say a faster horse”, and Steve Jobs famously never used surveys. I hate surveys, never used surveys, because you’re trying to invent and create something that people don’t know they want yet, you can’t get that from data. But data can inform your opinion, and help you build the foundation that lets you have the big leaps of insight. Was that the issue in the early 2000s when you were talking about everything was purely commoditized? Because if you did any survey you asked people what they want, they wanted direct flights, I would assume, and low prices. And that was very tricky. SK: That was the industry convention, and all the modeling, and all the math supported that and really no one had tried very hard to challenge that orthodoxy. And, for me, the first big challenge to that orthodoxy was JetBlue when they started, and they put live TV on the airplanes. I thought that was a gimmick at the time, but then I went and flew it and like, “Holy cow, this is a meaningfully different experience”. JetBlue was 12 or 13 airplanes at the time, and we had gotten up to 100 and so we cut a deal with a company that was called LiveTV to put live TV on all of the America West airplanes. They would have been growing eight or nine times with that contract, a huge transformative deal, live TV, nobody else is willing to do it, everyone else talked about it like a gimmick and it would have been the biggest CapEx investment outside of aircraft we’d ever made by far at America West, got the board to approve it. I called the CEO of LiveTV to tell him we got that done and he didn’t call me back, I’m like, “This is weird”, it took, like, 24 hours. I was starting to text people, or call people and say, “Is he okay?”. SK: “Did something happen?”, he finally called me back, and he said, “I’m so sorry”, he said, “We told JetBlue that you were going to do the deal, and they bought the entire company to keep you from putting it on the airplanes”. So, that would have been like 2000 or something, maybe 2001, ’99 or 2000 probably, and that was my first big point of, “You can get people to pay for a better experience”. Well, tech is on the extreme in terms of thinking about massive fixed costs upfront, and then zero marginal costs in terms of actually serving customers. Airlines are challenging. You mention that would have been one of your largest CapEx expenses ever. You have large fixed costs, you have the airplanes, but you also have ongoing marginal costs in terms of employees. Also fuel, which can be very much out of your control. When you’re thinking about managing these companies, or making that decision about this LiveTV, how do you think about balancing those different types of costs that go into your COGS? SK: Yeah, we’re much more operating expenses as opposed to the CapEx. Our CapEx can be big but if you just look at depreciation, for example- If you’re on a 30 year depreciation cycle, it doesn’t add up to much. SK: Yeah, it’s a relatively low. Most of our expense is employees and labor and maintenance and landing fees and all things that are variable with how much you are actually flying. So, it’s very different than tech in that sense. But I think it’s probably similar in that you’ve got to make big bets upfront, particularly about the aircraft and what you’re going to do with the aircraft. Right, you’re stuck with them for a long time. SK: And that is the one decision, almost everything else you can fix that you screw up but if you make a bad fleet decision, that can be the unpardonable sin, that’s what’s happened to the ultra low cost carriers, they bought too many airplanes that are way too big. And when you do that, you can’t fix it because they are 30 year assets, when you make big fleet decisions, you bet the company. Those are decisions that you have to feel really good about the analytics, but at the same time, the analytics, who knows what’s going to happen in 30 years? Good grief, we don’t know what’s going to happen this year, much less the next 30 years. This is the way I think of — people get lost in the spreadsheet, I often say, the spreadsheet doesn’t have the answer. The spreadsheet is good for, “Under this set of assumptions, here’s what the outcome is”, but in your brain, you’ve got to have a whole variety of differing assumptions to say, “I can make it through, I can live with this decision if things get a lot worse than I hope, if things get a lot better”. You’ve got to be able to think through a wide variety of outcomes, and the spreadsheet is just a starting point for the real analytical process that needs to go on in your brain. So when I think about flying in general, just as just a consumer, I think of there being two markets, domestic and international. Given that I lived in Taiwan for most of the last two decades, I was obviously much more focused on international. When I get my 1K, I’m always on the miles or spend side, not on the segment side. Is that a division that you think about in terms of your business and your competitive set? It seems kind of obvious from the outside, do you think that it’s more of an integrated whole? SK: We do, but they are an integrated whole. So, our flight from Taipei to San Francisco. Two a day now, makes me happy. SK: Yeah, it has a fair number of people that are connecting from other places. San Francisco is a great gateway, it’s the best gateway to the Pacific, we can connect the whole country. Obviously, given the tech and also just the heritage relationships between the Bay Area and Asia, there’s a lot of the traffic is coming locally, but a lot of it is connecting and from others. And there’s other divisions, business versus leisure travel, premium versus price sensitive so there are a lot of divisions, but we’re the largest international airline from the US. In fact, we fly to more cities internationally than all the other US airlines combined. So when you talk about those, you talked about all those divisions, like premium versus low cost, or whatever it might be, what matters domestically? Is it different there? I think about domestic, I think about all the segmentation that you’ve been doing, for example. SK: So, the primary way I think about the segmenting in the market is brand loyal customers versus price sensitive commoditized customers. That makes sense. SK: And commoditized customers typically are, this isn’t a perfect distinction, but typically, people that don’t fly very much, they care about the schedule, so they’ll pay a premium for a nonstop. But all else equal, they don’t really care what airline they fly. They don’t fly enough to have be in the frequent flyer program, they’re just looking for the cheapest price, and it’s a commodity for them. The brand loyal customers are the ones that we’ve focused on really trying to over-index on and when, and those customers care about the service, they care about the product, they care about the technology. I think one of the biggest and under-reported advantages- Oh, we’re getting to that, don’t worry. We’re just building up, we’re building up to this. SK: But the app on United, it’s like, I don’t know what airline in the world is second best, but there’s no question that they’re just a second. It’s true. It is 100%, there’s no comparison. Well, actually on that point, I had a little soliloquy here in the middle about my personal experience, because I’m very focused on international. So several years ago, I booked a business class flight on Singapore Airlines, long regarded as one of the best airlines in the world. I had to cancel that trip and it blew my mind that in order to get a refund, I had to submit a refund request, and it was literally a WordPress form on their website. I love WordPress, I’m very familiar with their ecosystem, doesn’t feel like the best tool for claiming a refund. I’ve told this story before in other contexts, but there’s an addition to it, which is in preparation for this interview, I went to find the email trail from this. It turns out they did email me eight weeks later asking why I wanted a refund, and I don’t think I ever got the refund. SK: Oh, geez. You should go back! I know, we’re going back, it was an expensive ticket! SK: By the way, I will tell you, there are some airlines, I know at least one in the United States, who do that on purpose because it helps what they call breakage. The harder they make it for you, like I’m talking about the things we were doing to make refunds easier, and putting up flight credits on our website and things so that you would know you had them. Somebody’s like, “Are you crazy?”. “Make people search for it!” SK: People will use them now instead of you just get to keep the money. I’m like, “Oh my God, you don’t understand the customer, thank God I’m competing with you”. Well, so anyhow, as I was filling out that stupid form, but I can remember thinking this in my head, I decided then and there, and I’m in Asia Pacific, so obviously it was United was the option, I was only flying United if at all possible, and the reason was solely because the website and the app were so superior to everyone else, by a massive margin. I could change flights, I could cancel, I could do whatever, I didn’t have to talk to anyone, and I followed through. I used to fly a lot more EVA Air, which is one of your Star Alliance partners, not the best partnership, I have complaints about that as well, but now I almost exclusively fly United. The one exception is EVA does have Taipei-Chicago direct flight, that one’s very nice, but to me, but this is the part that I always said I want to get to, I realized that for me, someone who flies a fair bit, changes my schedule a lot, nothing matters more than technology, and United had the best tech in the world. That’s my soliloquy. I think I’m putting the ball on a tee for you, but is this the sort of outcome that you foresaw when you come to United in 2016 and you start really making these investments ? SK: Yes, but it’s been even more impactful than I thought, and I’ve always thought that using technology to anticipate a customer’s needs, solve them before, make it easy, treat them fairly, was more important even than things like the product. And you’re seeing with your Singapore Airlines experience, you can’t get over something like that when you treat people that way and force them that. Yeah, walk me through this journey because again, I fly on a lot of airlines, I do fly United the most. How did your technology become so good? What was the change management process here? How did you go through it? What was it like when you showed up there? SK: So one, we have a great team, we have great people. It’s the one part of the budget that’s been sacrosanct that I every year increase the budget, tell them to spend more money. Even during COVID, we didn’t cut it, so leading by example, but also pushing ideas really hard that are important, and pushing things that everyone thinks we shouldn’t do or that are impossible. I’ll give you one example. One of the biggest culture changes, this isn’t really customer facing directly, but it is indirectly, that we made was, as soon as I became CEO, I eliminated what are called delay codes. Every airline in the world has a delay code, if there’s a flight that’s even a one-minute delayed, “Is it the flight attendant’s fault?”, “Is it the pilot’s fault?”, “Is the FAA’s fault?” — somebody gets blamed, and that’s true at every airline in the world. And you’d have these meetings that would go on for hours where groups of people were trying to pin the blame on a different group and management teams at United and around the world say, “Oh no, we never hold anyone accountable for that, this is all about we have to have this data in order to run the airline”, and I just always thought it didn’t matter if people were held accountable or not. People didn’t have four or five hour meetings about a single flight if they didn’t think they were going to be held accountable and so that’s all that mattered — I just made them stop and they said it’ll be impossible to run the airline. Within a couple of months, we now have built technology, we have better data to analyze what’s happening with delays and all the things that are happening, put cameras up, use AI, I use AI to figure out what’s going on, far better data than any airline in the world. And because of that, it also made our culture infinitely better. We took that off of the plate and told employees, “We want you to just do the right thing. You’re a team, work as a team. It doesn’t matter, there’s this delay, it doesn’t matter whose fault it is, you’re not going to get blamed”. It’s transformational for the culture and we built a better data architecture that lets us start explaining delays to customers and doing things that no airline has the ability to do it, because they’re still doing paper and pencil, sitting in the conference rooms, blaming each other for delays. And that’s a deep in the weeds kind of thing, except what sounds like a simple decision, but it has done as much as anything to change not just the technology, but the culture at United. What has been the balance between the consumer facing tech changes, whether that be in the app — I know a manifestation of that data is you now have LLM generated explanations for why flight is delayed, which is very cool; I remember the first time I was sitting on the right side of the airplane, I got a notification that my bag was being loaded and I saw my bag going up the track, that was very cool — versus the stuff you’ve done on the backend, like managing your irregular operations, all those sorts of things. Is that a cohesive effort or you pick one thing to focus on and go back and forth? How do you think about that? SK: We’re trying to do it all, because it all ties together. One of the big enablers was, that we started after I got here, we spent several hundred million dollars just getting off of our legacy mainframe systems, several hundred million rewriting code. Is that the key? That is the the key enabler? SK: That was a key unlock. You can’t do what we do unless you do that first, it was a key unlock. So, just for the nerds, what was the system that you were running on before? SK: Oh, we were on, it was called SHARES, but it’s a Fortran-based system that was written in the ’60s. And so, what are you on now? Are you just on a regular cloud environment now? SK: We’re on our own stuff. Yeah, it’s all cloud and it’s our own stuff, but we spent hundreds of millions. It’s not done. In fact, the last cutover’s coming next year for the last piece. Oh, now I’m feeling nervous. SK: Most of it’s done. Just out of curiosity, when is the last cutover? SK: Well, we’ve done it a piece at a time. So none of them have ever been really impactful because you’ve done a piece at a time, and there’s several pieces next year, so they come, but we’ve never had any issues with it because we’ve done them a piece at a time. And when you do it a piece of time, there have been problems before, but it’s easy to roll back when you do it a piece at a time, relatively easier to roll back. Has anyone else done this? SK: Not that I’m aware. No, certainly not to the extent that we’ve done it. They’re still on old legacies, because it’s hard. If you’re sitting at airlines, which haven’t been great margin businesses, you’re sitting and somebody says, “I want to spend $150 million next year to just take it off of the legacy mainframe system and put it into modern coding system”. No obvious, immediate benefit. SK: And there’s no ROI, there’s no nothing. “I don’t know what I’m going to be able to do with it, but it’ll just be easier to do stuff in the future”, I’m probably the only CEO that says yes to that, none of the others do. What was the process here? Who convinced you to do this? Why did you make the decision? SK: I think they did bring it to me, I’m not sure. Actually, I think I may have been the one that was pushing, “Let’s get off these systems”. It’s okay, you can take credit. SK: Partly I was a programmer. I mean, I double majored, my other degree is comp sci, I understood it. When I was at the Pentagon, I got responsibility for managing a whole set of tech stack, I didn’t know Unix. I remember at the time, you’re going to be the system admin for a Unix system, and they gave me a manual that was three inches thick and at some point I erased the kernel, which took me about six months to recover from, but I learned a lot by doing that. But anyway, I understood when I was working, I wrote Fortran and did those kinds of programs. It’s impossible to get anything done and you put it in a modern system, man, you can change it overnight. Try something new. I think that explains a lot. I mean, especially to people in our audience, if understanding you did this level of rewrite of the underlying system, no wonder you can roll out features so rapidly. SK: In fact, I rewrote a couple of the programs that I was responsible for. Nobody asked me to, I just did it when I worked at the Pentagon and put them in Unix, put them in C, but on a Unix platform. That’s super interesting. Well, to go back to the question about costs, because this ties into it — it strikes me about this, you just talked about it, it’s a real example of shifting spending to upfront costs. Was this something you had to convince the board of? Be like, “Look, this is as good as buying a bunch of new airplanes, it’s a similar cost”, or were you able to just make the call? SK: The board has been really supportive of me and everything I’ve done. It helps when it’s all working — when you have a track record, they’re willing to give you a lot of rope. But not everything has worked, and when it doesn’t work, we pull it back, but they’ve been willing to give us a lot of rope. Another thing that we do that’s totally different than every other airline is at the end of the year, when we go through our budget every year, I at the end of it say, “Here’s another 150, some number, millions of dollars that I want you to spend on the customer”, I don’t pretend to know what the best thing is, I hardly ever eat the food on airplanes, we spent a bunch more on food, we spent a hundred million more last year on upgrading the quality of wines. Not my thing, but other people do know. That was my next question! Actually on this point though — wine is literally my next question — but when you did this commitment to rewrite, when you were thinking through as far as cost structure goes, and you mentioned that this has exceeded your expectations in terms of the benefits, but where did you expect it to manifest? Did you think you would get lower marginal costs, you’d be better, or did you see this being able to sustainably charge higher prices? What was your expectations, and how has it actually played out? SK: It wasn’t about cost at all, it was about doing more for customers. My goal for 20 years has been, from a technology perspective, get to a world where if your flight is delayed or canceled, pretend I’m on the airplane and I’ve called our network operating center and asked them what’s going on, what would you tell me? I want to be able to automatically tell every customer that. And it was not going to be possible to do it when we were on those legacy systems. We’re better by far than any other airline in the world at doing that, but I think we’re still in the pregame on getting it. I think we have the data architecture that allows us to do it, and AI is really the right tool to effectuate that and I think we’re getting much better at it, but we’re talking about including things like weather forecasts for your flight 24 hours out, so you can look at the maps of, “Here’s what’s happening”, if there’s weather coming into your city, here’s the other routes that you could take to get there. Just all kinds of cool stuff that will really give you very inside baseball of what’s going on, take some of that uncertainty away. Yeah. Well, I know I talked to Flighty , you may be familiar, the flight tracking app, and he both agrees that you are by far the best app, and also feels like you keep stealing all his good ideas, so I think he means that as a compliment. SK: I think we talk to them a lot. In fact, I told our team, I don’t know what they’ve done, I shouldn’t say this in public, but I told our team, “We should just go buy Flighty, we’ll keep running it as Flighty, but I just want smart people that aren’t sitting at headquarters that believe anything is possible. Don’t tell me why we can’t do it, but go buy Flighty, and just let those guys tell us what to do”. They’re really good. I’ll let [Flighty Founder and CEO] Ryan [Jones] know. Who had the idea to greet frequent flyers by name, and why wasn’t this always the case? SK: It wasn’t me, but it’s part of enabling it with the technology. We also have your birthday in there, if you fly on your birthday, the flight attendants or someone will usually tell you happy birthday. We have all kinds of stuff like that about you. Well actually, my last birthday I took off from San Francisco on United on the 11:00 PM flight, and the day before my birthday, and I landed in Taipei the day after my birthday, so I missed my birthday completely. No one noticed, alas. SK: (laughing) You’re supposed to go the other way, then you can do a double. Do a double New Year’s is the thing that people do, they start in Guam and fly to Honolulu on United, you can do a double New Year’s. Well, anyhow, I have the menu from my last flight here, because I did have a question about wine. I called you the best airline in the world a few years ago, I think on Twitter, solely because of the technology edge. And I sort of granted in the tweet, “Yeah, the food and wine aren’t as good and the flight attendant’s going to be grumpy and da, da, da, da”. Well, first off, being greeted by name helps a lot, I love it. Secondly, when and how did United suddenly have the best wine in the world? I told you this is literally my question on here. On this flight, I had a Mascot Cab, it’s a 4.5 on Vivino, a $200 bottle of wine. It was very, very good. The biggest problem is that every other wine on the list was also really good and I wanted to have all of them. Am I talking to the United CEO, or the Emirates CEO? SK: (laughing) That’s a high class problem. So really what happened on this was this was one of those at the end of the year I say — most airlines go back and say, “You have to cut the budget by X million dollars to meet our plan”, and I do the opposite and say, “Here’s how much extra we’re going to add, purely for customers”, and we have a great team. It’s led by Andrew Nocella and Toby [Enqvist] , who’s our Chief Operations Officer, a whole bunch of people are involved in it. But they wanted to spend an extra $100 million on wine, and I trust him. Although I looked at, because I’m not a oenophile, I can’t tell the difference. I’m like, “Really? We’re going to spend $100 million on wine?”, because it’s like that is not what — I haven’t had a drink on an airplane in at least 20 years, so it’s just not what makes a difference to me. But I trusted them and they were right, and so we’re doing it next year. One of the wines they sent me, my wife and I had it. Oh, there’s been a step up! So I noticed this happen about 16 months ago. I’m like, “Holy cow, there’s more good ones”, and then this year it’s actually taken another step up. SK: Yeah. There’s some wine from South Africa, which I should know the name, because I drank it over the holidays, it was fine, it was good, I just can’t tell the difference. And they sent me one that’s coming on, that won some big award, global award. But anyway, that’s an example of hire great people. So what makes this possible, why can you spend extra $100 million dollars? SK: Because I’ve had a belief that air travel is not a commodity and that customers care. And no matter how good we are, we have to get better every year, or we will lose them, and we will deserve to lose them if we don’t get better every single year. And I believe that if we do that, we’re going to get market share, particularly for premium customers, or particularly for brand loyal customers who fly a lot, and they’re going to move to us, and that has worked like a charm. In every one of our hubs, we’ve won massive market share since we started this. We do win every year, we get more market share, and it’s about having a long-term vision instead of the short-term. I’m just trying to maximize profitability for this year. In the short-term, it would probably be better to not spend that couple hundred million extra, but in the long-term, it’s a great investment to win customers for the future and so, we’re doing it, we’re going to keep doing it every year. Have you actually been preparing for better wine for basically your entire career? First, by driving consolidation such that there’s a possibility to pursue these things, then with the tech enabling these new features? And the payoff is I get to have a great glass of wine on this flight. SK: Yeah, I wouldn’t have said it was going to be about the wine, but what I was trying to do was build an airline that could invest for customers and win brand loyal customers, could win not just on schedule and price, that was my goal. And wine wasn’t what was going to make the difference to me, but it is one of the things that makes a difference to a lot of other people. You made some very interesting comments last fall about the US having what you call a trade deficit with international airlines. Explain that to me. SK: So about two-thirds of the long-haul international seats are on international airlines, even though only about 40% of the passengers are international and that is because those airlines have different kinds of state subsidies. Quasi-monopoly positions, if you’re in Singapore, most airlines can’t fly and compete with them, most of the route, Middle East carriers — and I’m not complaining about that, by the way. The city of Dubai, the massively wealthy, impressive city state of Dubai doesn’t exist without Emirates, they go part and parcel together, it helped build it and Singapore doesn’t exist without Singapore Airlines. It’s important, it makes sense for those places to do that. But we’re then trying to compete with airlines that are essentially loss leaders for their country, their city state and so we’re kicking around some ideas of we don’t want subsidies, not trying to hamstring or tie up the other competitors around the globe, but how do we rectify that trade deficit and get an equal amount of long-haul international traffic? Well, what are some examples of what you want? SK: Well, I can’t tell you yet. But is this something where you need outside help, or you think you can do it yourself? SK: I don’t know what you mean by outside help. Well, I mean you’re starting to fix it to a certain extent. SK: Size would help. Size in what sense? SK: Being bigger would help. And this is just more leverage on your costs? SK: Well, it’s that, and it’s more ability to actually have all of the US customers. You say you love United and you mostly fly it, but there’s a nonstop flight on EVA Air for Chicago, it’s the same thing, you look in a place like New York, it’s more fragmented. And so we have customers that fly United almost all the time or they fly Delta, but when they go to the Middle East, it’s fragmented enough that they fly on Emirates, and if we’re bigger and have more offerings for those customers, possibly, it makes it more rational for them to fly us when they go to the Middle East. And if we do that, then we can invest in — we would have the profitability cashflow to invest in better product, and really the place that the international airlines beat us is on product, premium product in particular. Ours is really good, all the US airlines are really good, all the European airlines are good, but they over-invest there. They lose money, but they over-invest there and that’s the gap that I am now thinking about how to close. I want us to be the best airline in the history of aviation and I think we’re already the best in the world, you said it too, but that’s collectively. We’re not the best at everything yet, and I’d like us to get closer to that. So what’s the limitation in getting larger? Because I see two angles. One is can you even get airplanes? And then number two is, “We could cut an interesting deal with JetBlue, for example, if the government will let us, we could consolidate even more, that’s our bread and butter”. SK: You’ve talked about the two interesting possibilities. Which one is a harder nut to crack? SK: (laughing) They both have challenges. You’ve had some spicy thoughts about Boeing in the past. How’s the relationship these days? SK: Well, we have a good relationship, but when they were stumbling I called them out on it, especially when they were stumbling and not willing to acknowledge that they were stumbling. I don’t mind people making mistakes, myself, people that work for me, I can’t stand excuses. The best thing I learned at the Air Force Academy is, “No excuses, sir!”, and I cannot stand when people make excuses. And when Boeing was screwing up and not acknowledging that they had problems, that meant they weren’t going to fix it. I think under [Boeing CEO] Kelly [Ortberg], they’ve done a really good job of getting on the right path, they’ve turned it around for the narrow bodies. Actually, the wide bodies, 77s have been a little behind, but we think we’re going to get 20 787s this coming year, we’re supposed to be getting 24 a year, but we only got 8 last year, and we’re going to be up to 20 this year, so I think they’re there. I think Boeing, the ship has been righted and I feel good about Boeing. Just to wrap up this part, the reason I was enthusiastic about this interview beyond being a United 1K flyer, I think it’s easy in the weeds of technology to forget why technology is critical for human welfare and improvement generally. From my point of view, to your point about facing some structural disadvantages relative to other airlines, you overcame that from my perspective with technology, and the great thing about that is it benefits everyone. Even the basic economy flyer gets to use the United App and website, maybe not the refund part of it, but then by extension, that gives you the cost structure to be able to offer me better wine. That’s pretty inspiring. But do you feel you’ve reached the limit of how much you can get from that? SK: No, not at all. In fact, I think in some ways we’re just getting started. The gap in profitability, Delta and United this year, the two of us, they have a similar strategy, although I think they were ahead of us on starting it, but we’ve caught up and are passing them. But the two of us are going to be 100% of the industry profitability this year and the rest of the industry’s under strain because they’re very much in the commodity business, they focused on the commoditized travel. And so, I think as we continue to get better, the pressure on them gets higher. I think it’s going to lead to a better outcome for us, better outcome for our customers as well, but I think those leads are only set to grow and we’re ahead of all of our competitors. You take technology, we’re ahead. I can almost guarantee you that we’re investing more next year than anyone else — even though we’re ahead, we’re investing more. I tell our team, I said earlier on this, our job is to come up with cool stuff to wow you every year and we don’t get to take credit for the things we did last year. We got to do new stuff every year, or we’ll lose customers and that’s our attitude and we’re going to keep it going, and no one else is doing it, so it’s just remarkable to me. We’re doing all this and no one’s copying the things that matter, which is great. Does it go back to just because you have to rewrite the entire foundations of the technology stack? SK: It does. And it’s that, and it’s also a finance-driven, it’s short-term finance-driven cultures that if you can’t prove this has an ROI, you can’t do it. And as data-driven as I am, all the big decisions that I’ve made in my life, there was no way to calculate the ROI on those decisions. It’s just like you’re playing poker. There’s no way to know for sure what the other guy has, and you make bets when they’re good bets, and they’re just not willing to make bets. How do you actually tell this story? Because I think that was one of my fun things to drop on Twitter, “Oh, United is actually the best airline in the world”, because it outrages a lot of people who have, whether it be stories of people dragging them off planes or their guitar or whatever, or their own personal experience with delays, people always blame the airline for everything. How do you actually convince people that it actually is different now? Or is it just a matter of people who fly a lot, they will pick up on it and you can be in Chicago and take over the market, for example. SK: People that fly a lot do pick up on it, but we’re certainly not perfect, and there are things where we screw up and I hear stories and I’m like, “Ugh, I can’t believe we did that”, but when I hear those kinds of stories, I think, “How do we make that better? How do we change that?”. It’s not an excuse, how do we change it and make it better? And technology has been a big part of it. One of the things, we didn’t talk about it really, but we had this, we called it Hell Week in Newark, where the airport was closed. It was a combination of storms and the FAA, we came close to needing to shut down Newark to recover, we didn’t quite get there, but we came close. And you dig through it all, at the root of it was technology problems with crew scheduling, just archaic processes. One of them, we went out to the NOC, network operating center, the week after it to thank people for all the hard work and to talk to them, but also just to talk to them, and I’d never really focused on it. I was at the crew desk and talking to them, being like, “Why does it take us so long to get our crew?”, they were talking about how we had so many to repair and we just were behind and it just kept snowballing on us. And it turned out that we were making phone calls to every crew member that got reassigned. So you have a storm come through, you need to cancel 200 flights. Well, there’s an average of four flight attendants per flight, two pilots, so you’ve got to make 1,200 phone calls, and most of the time, people don’t answer the first time. And our crew desk was on the phone, 95% of their time was on the phone. I was like, “Why are we doing that? Why don’t we just text them?”, “Oh, our union contract said they have to have a call”, “Well, why don’t we robocall them?”, and it was like, “What are you talking about? We can’t do that”. Three months later, we robocall them and it’s perfect. It’s better for employees, you just tell them what it is, it’s better for everyone, but we completely re-architected how we manage crews. And now, when we have these big events, we watch other airlines like, “Oh, they’re about to get in trouble, you can tell they’re in trouble”, and we’re able to repair and go on. It was great during the shutdown and other places but just the investments that we’ve made operationally and for the customers really just make the experience better. What about AI? AI, in some respects, you’ve been using machine learning for a lot of things, it’s a big part of your investment. You do have the LLM-generated text about why your flight is delayed. Everyone wants to talk about AI, I feel compelled to give you an opportunity to say something about it. SK: Yeah, sure. But we’re also talking about airplanes in the sky. Does it really matter? SK: I’ll tell you what we’re doing, and then if you want me to, I’ll give you my broader view on AI, which is different than everyone else and won’t be as popular. So I think AI is a wonderful tool for many classes of problems. Our ability to communicate with customers and tell them exactly what’s going on in a delay is one example of that. We’ve got some work that we’ve done with AI on airspace when storms happen, because we have this massive dataset, we started in Denver — decades of data of when storms hit, which routes are open, what’s the best way to sequence aircraft in, and we can make the Denver airport dramatically more efficient, we can do it at every airport, dramatically more efficient with AI. At the moment, it’s a challenge, because the FAA can’t ingest that data, they have not modernized their systems and can’t ingest that data, but it creates benefits for us anyway. But also, we’re experimenting with the art of the possible for the future to run the airline better. Essentially, our network operating center, the brains of the airline, I think almost everything we do in there, AI can do it better than humans can do it, because we can look at — we have massive data sets, it’s pretty easy to define the objective function of what it can do, and it’ll be really good. But I will put myself in the distinct minority of people on the broader level about AI and the gradient descent models that are driving AI, that I do not think we’re going to get to artificial general intelligence with that, and I think there’s a whole class of problems that they are terrible at that they’re being used for. On the consumer AI use, I’ll give you two things that I dislike about it. They’re good at selling you stuff that they want you to buy, and the emperor has no clothes, telling you what you want to hear. My simple example, go to AI and ask it, “Where can I watch Landman? Which streaming service can I watch Landman, or any other show, for free?”, you cannot find it. It’s maybe a hundred pages down, I haven’t gone far enough, it’s all advertising. And then, the second one, I’m telling you what you want to hear. It’s designed, that’s the feedback it gets, to tell you what you want to hear. My mom recently broke a bone in her back and she’s laid up, and on the 12th day, things weren’t getting better and it was getting worse, so she was using ChatGPT, and she asked it, “I was really doing good and this is my 12th day and it’s really bad”, and it gave her this detailed, “You used to be acute, now you’re subacute”, all these terms, and said, “But this is normal on the 12th day”. That’s BS, that’s not right. So I got on a different device and said, “On my 12th day, I felt remarkably better”, and the same system told me, “That’s normal, on the 12th day, you get remarkably better”, it is designed to tell you what you want to hear, not what you need to hear. So my caution to people with AI is I think brain power is still really important and the ability to think independently. AI can be used for some things, but you need to still use your brain. Why did you make the jump to Starlink ? Obviously, Hawaiian had been there and I used it on JSX years ago. You’re the first big airline. Now, Air France is on, Lufthansa just announced today, what made you take the leap? SK: I knew for a long time that we were going to go, for a while, that we were going to go. It’s a step function improvement in technology, no one else can get there and others are going to try to launch Low Earth orbit satellites, but good grief, they’re light years behind, they’re not going to get there. So Starlink was going to be the answer. It took time to negotiate, because we wanted to own the consumer data, and at the beginning, Starlink did, so that was hard, and then, the other thing was I wanted to let my big competitors in the United States finish their deals with other providers and get locked in so that we would — eventually, everyone’s going to have Starlink. That’s my question. SK: But it’s going to take them time. They don’t know it yet, but it’s going to take time, they don’t know it yet. And every day that they don’t put it on their airplanes, it is one more day of advantage for us. That’s my view , Starlink is going to be a monopoly in the long run. Are you worried about that or are you just like, “Yeah, you’ll have to pay for it eventually”, but for now, you get differentiation for the next however long? SK: We’re going to be way ahead, and then we’re going to come up with the next great thing. They’re going to always be behind us. How important is advertising to this deal, to having all the seat-back devices? The main thing to use it, you have to be a MileagePlus member, that gives an identifier, you know who they are. SK: Yeah. I think the advertising piece is the icing on the cake. It’s not the cake, but it’s icing on the cake. So it’s not going to be a material driver, or maybe it will be someday? SK: No, I think it’ll be a big financial issue. But I really am focused on winning customer loyalty, we’ll get some bigger payoff for that. Is the real payoff credit cards? Are you actually a credit card company? SK: No, we really aren’t, that’s a bit of a misnomer. We’re a loyalty company and credit cards are part of it, but the credit card doesn’t exist unless you can take the cool, exotic vacation to Tahiti or Cape Town or wherever you want to go. I just want to board early, that’s all I care about, and I want to be greeted by name. That’s the thing is the higher you get, the less the perks matter. I have three quick things, I know we have like three minutes left, my own personal beefs. Number one, one underrated thing about United is you have the best bedding in the industry. Two blankets, two pillows, good for me as a warm sleeper. Please, don’t change this, don’t cheap out on this. SK: Okay. By the way, I’ve never used the bedding either, because I go to sleep as soon as the plane takes off the runway. You never put your seat down, yeah. Well, that little gel pillow is amazing. You have a shop online that lets you buy United onboard items, it does not have the gel pillow, I need a way to buy this pillow, so that’s number one. Number two, SFO. As an American, I appreciate the egalitarian nature of baggage carousels in US airports, which means that your status has zero impact on when your bags come out, unlike in Asia, the status bags always come out first. However, after waiting for one hour for bags last week and almost missing my connecting flight, I need to pull my elitist card and ask you to fix that. SK: Okay. Here’s two things. One, that obviously implies, that you were waiting on a connection, that you were coming in international. My number one ask to change something with this administration, and I think we’ve got a chance at getting it done, is it is ridiculous that you have to collect and recheck your bag. It’s absurd. SK: It’s a hangover from the Lockerbie disaster when all the systems and technology and security protocols were different. It’s a perfect example of government inertia. There are tens of thousands of government employees who are working on that, millions of hours of lost time for you and other customers, and it is totally unnecessary. It’s pointless. SK: Totally unnecessary. I’m working to get that changed. No one checks it anyway, I just walk in and drop it right back in. All right, thank you. Good, that’s even better. SK: You shouldn’t need to touch it. If you fly domestically, you don’t have to touch it. We’ve already scanned your bag for security somewhere else in the world, we know it’s safe. Why, because it’s international, do we have to touch it? It makes no sense at all. Well, that’d help your business too. You could do more connections international-international. Number three, I know the build up of Denver means this isn’t going to happen, but you had a Madison-San Francisco flight that died during COVID. Please, I want you to bring it back. SK: No, no, no, that flight did well. The issue with San Francisco right now is the runway construction this year, and so we have fewer flights than normal because the runway construction’s been going on this year. So you’re saying there’s a chance? SK: I’m telling you there’s a chance, one of the best lines in the history of movies , you’re telling me there’s a chance. I love that line. Scott Kirby, thank you very much. SK: Thanks, Ben. This Daily Update Interview is also available as a podcast. To receive it in your podcast player, visit Stratechery . The Daily Update is intended for a single recipient, but occasional forwarding is totally fine! If you would like to order multiple subscriptions for your team with a group discount (minimum 5), please contact me directly. Thanks for being a supporter, and have a great day!

Business

▲ 0

0 views

Stratechery 3 weeks ago

Meta Compute, The Meta-OpenAI Battle, The Reality Labs Sacrifice

Mark Zuckerberg announced Meta Compute, a bet that winning in AI means winning with infrastructure; this, however, means retreating from Reality Labs.

Hardware AI Business

▲ 0

0 views

Stratechery 4 weeks ago

Apple and Gemini, Foundation vs. Aggregation, Universal Commerce Protocol

The deal to put Gemini at the heart of Siri is official, and it makes sense for both sides; then Google runs its classic playbook with Universal Commerce Protocol.

AI Business

▲ 0

0 views