# Don't Look Up > Source: > Published: 2026-08-11 19:26:22+00:00 If you liked this piece, you should subscribe to my premium newsletter. It’s $70 a year, or $7 a month, and in return you get a weekly newsletter that’s usually anywhere from 5,000 to 18,000 words, including vast, detailed analyses of [ NVIDIA](https://www.wheresyoured.at/the-haters-guide-to-nvidia/), [, and](https://www.wheresyoured.at/howmuchmoney/) __Anthropic and OpenAI’s finances__[. My Hater's Guides To the](https://www.wheresyoured.at/premium-the-haters-guide-to-the-ai-bubble-vol-2/) __the AI bubble writ large__[,](https://www.wheresyoured.at/hatersguide-saas/) __SaaSpocalypse__[and](https://www.wheresyoured.at/hatersguide-privatecredit/) __Private Credit__[are essential to understanding our current financial system, and my guide to how](https://www.wheresyoured.at/hatersguide-pe/) __Private Equity__[pairs nicely with my](https://www.wheresyoured.at/how-openai-kills-oracle/) __OpenAI Kills Oracle__ __Hater's Guide To Oracle, as well as the Hater’s Guide To Oracle (Part 2).__Subscribing to premium is both great value and makes it possible to write these large, deeply-researched free pieces every week. On Friday, I'm going to pull together exactly how much money is needed to keep the AI bubble inflated in the next three years. It's gonna be a laugh-riot. Or very scary, one of the two. If you want to get in touch — and especially if you have any juicy information about Anthropic, OpenAI, or any other companies in the AI bubble — hit me up on Signal at ezitron.76. I’m also on IB on The Terminal. [ Last week I put out one of the most consequential newsletters I’ve written yet](https://www.wheresyoured.at/the-ai-demand-bubble/), pulling together multiple distinct financial analyst notes from Wells Fargo, Barclays, and UBS that directly estimated that 70% or more of the AI revenues of Microsoft, Google, and Amazon were from either OpenAI or Anthropic. To be clear, UBS estimated that next year, Anthropic and OpenAI’s compute spend would be 48% of *all*Google Cloud revenues — which means that they likely account for even more than 70% of its AI revenues, but I wanted to be fair. This was both a colossal pain in the arse and a story that I knew would piss off a lot of people, because of its huge ramifications. Some outright dismissed it as “doomerism,” while others insisted it was a good thing, because OpenAI and Anthropic are growing so fast. 24 hours later, [ Bloomberg ran a story](https://www.wheresyoured.at/news-microsoft-disclosures-suggest-openai-sales-account-for-around-70-of-fy26-ai-revenue-more-than-7-of-fy26-revenue/) estimating, based on OpenAI’s $24.1 billion dollar contribution to Microsoft’s Fiscal Year 2026 revenues and previous statements, that OpenAI alone contributed to 70% or more of Microsoft’s AI revenues for the year. For some context, Microsoft has spent $261.3 billion dollars in capital expenditures since the beginning of 2022. Meanwhile, [ Apollo chief economist Torsten Slok said Friday](https://www.apollo.com/wealth/insights-news/insights/daily-spark/in-ai-the-41-percent-depends-on-the-59-percent?ref=wheresyoured.at) that profit margins in AI are “...higher the further you get from the end user,” and then said something I think I’ve said maybe four times in the last three months: The bottom line is that the most profitable part of the AI value chain depends on the least profitable part continuing to grow revenue or raise capital. Capital can bridge the gap for a while, but not indefinitely. And therein lies the risk: will the ROI show up for AI's end customers fast enough to sustain the spending that is generating those upstream margins? Good bloody question Torsten! The answer is “probably not.” Let’s get real *simple* about this because everybody wants to make AI *so complex.* # The Future Growth of Google, Microsoft, and Amazon Is Contingent On Anthropic and OpenAI Spending $200bn+ in 2027, Which Requires $250bn to $300bn in Funding Sidenote:before we go any further, I need to be clear thatAI is not the reason that these companies are growing, outside of the compute spend from Anthropic and OpenAI. AI is not “boosting other product categories” or “helping other categories grow,” because if it was, they’d tell you specifically. I get so many emails from people sending me the overall revenues of these companies, mostly from people that don’t appear able to read, but nevertheless, I want to add this note on the off chance they learn. If we assume, on the low end, [ that Jensen Huang is right and he’s going to sell $1 trillion or so of GPUs](https://www.wheresyoured.at/the-ai-demand-bubble/#analysts-estimate-anthropic-and-openai-represent-more-than-70-of-all-ai-revenues-and-we-are-building-hundreds-of-billions-of-dollars-of-data-centers-for-nobody:~:text=pay%20their%20bills.-,This%20Is%20A%20Huge%20Problem%20Even%20If%20You%20Don%E2%80%99t%20Want%20To%20Think%20About%20It,-Jensen%20Huang%20has) (roughly 30GW of billable IT capacity), that’s somewhere between $360 billion and $435 billion of annual compute revenue demand. Right now, there are (outside of hyperscalers buying compute for them, and whatever it is Meta is up to) two companies that spend more than $500 million a year on AI compute, namely Anthropic and OpenAI. Both are unprofitable, and both lose tens of billions of dollars a year. If we take [ OpenAI’s testimony from the Musk-Altman trial](https://archive.ph/submit/?url=https%3A%2F%2Fwww.bloomberg.com%2Fnews%2Farticles%2F2026-05-05%2Fopenai-to-spend-50-billion-on-computing-in-2026-brockman-says%3Fsrnd%3Dhomepage-americas&ref=wheresyoured.at) as gospel, it’ll spend around $50 billion on compute this year, and if we (kindly) assume Anthropic will spend $50 billion itself, that brings us to $100 billion. To get to that level of spend, __Anthropic__ __and__[have raised a combined $217 billion in the first half of 2026. Every neocloud is effectively an outgrowth of this spend, either through](https://openai.com/index/accelerating-the-next-phase-ai/?ref=wheresyoured.at) __OpenAI__ __direct__[or by proxy](https://investors.coreweave.com/news/news-details/2026/CoreWeave-Announces-Multi-Year-Agreement-With-Anthropic/default.aspx?ref=wheresyoured.at) __contracts__ __via__ __Microsoft__ __or__[. Outside of hedge fund and investor](https://www.reuters.com/business/coreweave-offer-compute-capacity-googles-new-cloud-deal-with-openai-sources-say-2025-06-11/?ref=wheresyoured.at)__Google__ [and](https://www.coreweave.com/news/jane-street-signs-6-billion-ai-cloud-agreement-with-coreweave?ref=wheresyoured.at) __Jane Street__[, neoclouds do not have significant customers at the level that would warrant all this capex.](https://www.cnbc.com/2025/09/15/coreweave-stock-jumps-on-disclosure-of-6point3-billion-order-from-nvidia.html?ref=wheresyoured.at) __NVIDIA__So, the world is building AI compute capacity with the expectation of *at least* $360 billion in annual revenue, all while we struggle to find single-digit billions in AI compute spend. The only way all that compute gets used is if either A) Anthropic and OpenAI rent all of it or B) massive (and I’m talking multiple $10 billion-a-year customers) appear virtually overnight. In both those cases, the money to pay for that compute will have to come from somewhere. Remember: for Anthropic and OpenAI to be able to afford their current (and comparatively meager) spend, both have had to raise nearly a *quarter of a trillion dollars *in this year alone. The vast, vast majority of the world’s compute revenue — I’d wager anywhere from 70% to 90% — is contingent on venture capital propping the AI labs up, and to make matters worse, *it is no longer sufficient for them to just “grow fast,” but to grow so fast that they can spend (per estimates from Wells Fargo, Barclays, and UBS) $197 billion on compute in 2027 just on Google Cloud, Amazon Web Services and Microsoft Azure. *This does not include the billions that both will spend on CoreWeave, Cerebras, or Oracle. Sidenote:[. I will also add that these estimates only include Microsoft Azure for the first two quarters of 2027 (as estimates are only up to the]The numbers are all in this newsletterendof Microsoft’s FY27, which runs from July 1, 2026 to June 30, 2027). It’s likely that the number is more like $220 billion. It’s likely more. Also, before you ask: revenue concentration appears to be getting worse over time, because these 70% estimates mostly rely on the continued growth of overall compute and AI model rental platforms like Vertex, Foundry and Bedrock. Sorry, I got too complex again. **Anthropic and OpenAI are only set to spend $100 billion on compute this year, and had to raise over $200 billion to do it, which makes it likely they’ll have to raise $150 billion each leading up to or in 2027. ** And let’s be clear about something: the future growth trajectories of Amazon, Google and Microsoft (not to mention Oracle, CoreWeave, and every other neocloud) are contingent on the continued ability for Anthropic and OpenAI to *raise and have the demand necessary to spend that money.* Let’s get specific! Stephen Ju of UBS estimates that Amazon’s AI revenue — 73%+ of which is OpenAI and Anthropic’s compute spend and revenue share (per Barclays) — accounts for 26% of AWS’ 2026 revenue and 30% of AWS’ 2027 revenue. Michael Turrin of Wells Fargo estimates that 25% of Microsoft Azure’s (calendar year) 2026 revenues come from AI (of which OpenAI is an estimated 70%). Brad Zelnick of Deutsche Bank estimates that AI will contribute 33% or more of Azure’s revenues in FY2027. [ As I mentioned last week](https://www.wheresyoured.at/the-ai-demand-bubble/#:~:text=OpenAI%E2%80%99s%20ongoing%20work.-,Google,-is%20in%20a), UBS estimates that 48% of Google Cloud’s 2027 revenues will come from OpenAI and Anthropic. Zelnick of Deutsche Bank also projects in its most-likely scenario that AI revenues will make up 37% of all Microsoft’s cloud revenues in FY2029. I feel like I need to spell this out more. These are analysts from major banks and financial institutions. Their estimates, which are based on detailed financial models, inform Wall Street and investors’ expectations, as well as informing Bloomberg Intelligence’s consensus estimates for revenues. These are serious numbers and Wall Street* will be mad if they are not met! * The other problem is that cloud is becoming an increasingly-larger part of the revenues of these companies. See the below chart that bakes in consensus analyst estimates up to 2029: I, again, will simplify: if more and more of the revenues of these three companies are coming from cloud segments that are increasingly-dominated by AI revenues mostly driven by two unprofitable, unsustainable companies, **then the literal future of Microsoft, Google, and Amazon is whether Anthropic and OpenAI can pay them. **This is not complex, it’s not contrived, it’s not *doomerism* or *hating*, these are the estimates from analysts and what they require to stop them from putting executives in The Wicker Man. You can cut this situation in any way you want, but there’s no getting away from the fact that we’re four years in and the vast majority of demand comes from two companies that can’t afford to sustain it, and won’t be able to even under the most mold-poisoned of booster projections. **Hyperscale growth is contingent on the success of their AI plays, and at 70% of AI revenues, “AI plays” refers to “two unsustainable AI labs.” ** Perhaps another visualization would help! Below is a chart of the expected percentage of year-over-year growth that cloud revenues are estimated to contribute on a quarterly basis to revenues. Cloud revenues are the lynchpin of growth for Microsoft, Amazon and Google, though for whatever reason analysts estimate that YouTube and Google Search will re-accelerate. This is a huge issue when 33% of Azure revenue, 48% of Google Cloud, and (per Ken Gawrelski of Wells Fargo) 60% of AWS revenue growth is coming from companies that have been, assuming all the money crosses, sent a combined $115 billion from Amazon and Google in 2026 alone. I realize I’m repeating myself, and I’m sorry, but it’s all so insane! The future of some of the largest companies on the stock market is contingent on both *spending hundreds of billions a year in capex* and *the continued existence of the only real customers for AI compute. * Sidenote:None of this even discusses Oracle, which is completely dead if it doesn’t either drastically pull back on or entirely cancel its capex, as I wrote back in[,]January[, and]April[. Congratulations to the New York Times on getting there a few weeks later, I’m sure]July[to]this remarkably-similar sentence[was a coincidence.]mine The counter arguments are, from what I can tell, as follows: - Anthropic and OpenAI will simply continue to grow faster and faster, spending more and more on compute, reach *profitability*, and then*do so to such a level that they will need hundreds of billions of dollars of compute.* - Other companies — which have yet to emerge in any way, shape or form — will also need billions of dollars of compute. - Hyperscalers will make untold trillions of dollars’ worth of revenue. - If this doesn’t work out, “there will be uses for it after, like the dot com bubble,” .__even though it’s nothing like the dot com bubble__ - If this doesn’t work out, “there will be uses for it after, like the dot com bubble,” - Nuh uh! - The massive revenue backlogs are proof that there’s tons of pent-up AI compute. To be clear, “growing super fast” is no longer sufficient for OpenAI and Anthropic. Assuming that OpenAI actually intends to pay for [ its reported $750 billion in compute commitments through 2030](https://www.wsj.com/tech/openais-planned-cloud-spending-hits-750-billion-as-computing-efforts-ramp-up-6ac3f58a?ref=wheresyoured.at), it will have to raise hundreds of billions of dollars a *year*while also having the *actual demand necessary to use that compute.*No matter how big, handsome, and amazing you think either of these companies are, *their expected compute spend will require them to make as much revenue as Microsoft, Google and Amazon in the next four years,* **and if they don’t, hyperscalers will not meet analyst and investor expectations.** To make matters worse, for them to *even be able to pay hyperscalers, capex investment must continue, as it’s become blatantly obvious that the capacity necessary to make all this money doesn’t currently exist.* Hyperscalers have not yet *spent the money necessary* to reap the “rewards” of their massive contracts with OpenAI and Anthropic, and analysts estimate that these three companies will spend another $1.5 *trillion* through the end of 2027. So, again, let’s review: - OpenAI and Anthropic are on the hook for over $1.1 trillion in spending commitments, with hundreds of billions of dollars’ worth across Amazon, Google and Microsoft. - OpenAI and Anthropic represent 70% or more of AI revenues across these companies, largely from ever-increasing amounts of cloud compute spend, and analysts have set expectations based on their ability to continue doing so. - Microsoft, Google and Amazon are dependent on their cloud segments for overall revenue growth, and Anthropic and OpenAI make up large swaths of that growth. - To be specific, their estimated compute spend across these platforms is over $200 billion in 2027. - To pay for their estimated $100 billion in 2026 compute spend, they had to raise over a combined $217 billion. - The only way that Anthropic and OpenAI can pay for that compute is if they both *raise the money to do so*and*have the demand necessary to justify it.* - The only way that hyperscalers can get paid if they do so is if they can build the capacity necessary to fulfil these demands. - To do all this, hyperscalers will have to take on increasingly-large amounts of debt, issued in the bond markets alone.__with an expected $250 billion this year and $400 billion next year__ - To do all this, hyperscalers will have to take on increasingly-large amounts of debt, The demand for AI compute does not exist at scale outside of Anthropic and OpenAI, and it is not emerging anywhere that I can see. We are no longer in a situation where single or even double-digit demand for AI compute is sufficient. Based on the amount under construction, we need — even *with* OpenAI and Anthropic* — *hundreds of billions of dollars’ worth of demand in the next few years just to monetize the data center capacity under construction. AI boosters will insist that this is happening in the shadows, and that “all available compute will be used,” making the mistake of conflating scarcity of GPUs with overwhelming demand. If Microsoft 70% of Microsoft’s estimated $34.43 billion in AI revenue is from OpenAI, that leaves over a depressingly-low $10.33 billion across *every single possible AI service that Azure has, including renting GPUs, AI models and Microsoft 365 Copilot…*which means that, *in the literal best-case scenario*, there’s low-single-digit billions of revenue in AI compute to non-AI labs. Sidenote:[, with 365 Copilot revenue at an atrocious $3.858 billion. It has “other AI revenue” at $5.2 billion, which includes all AI GPU and API access revenue. Stinky!]Turrin of Wells Fargo has the estimated revenues at $34.5 billion If there was meaningful demand for AI compute or AI software, Microsoft would be representative of it as one of the largest vendors of both cloud software and cloud compute. Microsoft would, by virtue of its massive infrastructure and brand recognition, be receiving a large share of blue chip GPU rentals, as would it be representative of the ability for anyone to sell AI software at scale. $10.33 billion in *annual revenue* is a catastrophic failure. It is around a quarter of [ Microsoft’s $41 billion in Q4FY2026 capex](https://www.cnbc.com/2026/07/29/microsoft-msft-q4-earnings-report-2026.html?ref=wheresyoured.at). It suggests that there is a calamitous lack of demand across both those renting GPUs *and*market demand for software built on top of AI models, and that Microsoft spent $261 billion in capex since 2022 to create annual revenues that amount to less than a third of the quarterly revenue of the Intelligent Cloud segment ($39.31 billion). There is no spinning this positively other than to ignore it outright. If Microsoft doesn’t have the demand, nobody has the demand. No, $10 billion is not “a lot,” especially for a company with tens of thousands of salespeople, a huge customer base, and [ a headstart of several years](https://www.wheresyoured.at/premium-the-haters-guide-to-nvidia-part-2/#how-nvidia-used-hyperscalers%E2%80%99-post-2021-desperation-to-create-fomo-for-gpus-and-imaginary-demand-at-scale). Amazon and Google are doing equally poorly, which is why *nobody*wants to talk about their actual AI revenues. In fact, folks, it’s time for a thought exercise!Wells Fargo estimates that Microsoft’s FY2027 AI revenue will be $54.5 billion, or roughly 58.8% year-over-yearincludingOpenAI’s compute spend. If we increase it again by that much, FY2028 will be at $86.55 billion. Consensus estimates have Microsoft spending $186.4 billion in FY2027 and $214.3 billion in capex in FY2028. It’s hard to square how exactly this ever pays off. # Why Was Everybody Wrong About Hyperscaler AI Demand? I’m gonna say it with my full chest: anyone who said that “AI was paying off” for Microsoft, Google, Amazon, or Meta was wrong. Everybody who said the capex was well-spent was wrong. They are yet to admit they’re wrong because revenue growth has yet to slow and stock prices remain elevated. And that last part is why everybody* got it wrong.* [ As I discussed in last week’s premium](https://www.wheresyoured.at/premium-the-haters-guide-to-nvidia-part-2/#how-nvidia-used-hyperscalers%E2%80%99-post-2021-desperation-to-create-fomo-for-gpus-and-imaginary-demand-at-scale), hyperscalers started buying GPUs because their overall revenue growth had begun to slow over the course of a little over a decade, with *everyone*— NVIDIA included — [:](https://www.wheresyoured.at/premium-the-haters-guide-to-nvidia-part-2/#how-nvidia-used-hyperscalers%E2%80%99-post-2021-desperation-to-create-fomo-for-gpus-and-imaginary-demand-at-scale:~:text=Then%20everyone%20hit%20a%20wall.) __hitting a wall in 2022__Microsoft’s FY2023 revenues only grew 6.9% year-over-year. NVIDIA’s FY2023 (which ran February 2022 to January 2023) was effectively flat, sitting at 0.2% as the post-pandemic surge of demand for gaming GPUs and networking gear puttered out,[revenues dropping by 21% year-over-year.]with Q4 2023 Meta, Google, and Amazon, all of which have fiscal years that align with the calendar year, saw growth deteriorate: Google went from 41.6% year-over-year growth in 2021 to 10.3% and 9.7% in 2022 and 2023. Meta went from 37.2% year-over-year growth in 2021 to negative 1.1% in 2022 and 15.7% in 2023. Amazon went from 37.6% year-over-year growth in 2020, to 21.7% in 2021, to 9.4% in 2022, to 1.8% in 2023 (and has really never recovered.) While buying GPUs didn’t really help *revenues* until OpenAI and Anthropic became big enough to start feeding hyperscaler and venture capital cash into Microsoft, Google, and Amazon’s mouths, buying GPUs became a dick-measuring contest that pumped stock values, all as Wall Street assumed every dollar of revenues came from AI. In 2023, [ Microsoft, Google, Apple, Meta, Amazon, and NVIDIA added trillions in market capitalization](https://www.cnbc.com/2023/10/17/amid-ai-buzz-big-us-tech-giants-add-2point5-trillion-in-market-cap.html?ref=wheresyoured.at), with [. It didn’t matter that](https://www.cnbc.com/2023/04/28/tech-earnings-calls-show-mega-cap-companies-going-big-on-ai-.html?ref=wheresyoured.at) __the media actively encouraging them to spend more money on capex__[, much like how Amazon’s Q3 2024 earnings — which specifically missed expectations for cloud, the only place that Amazon was making any money from AI —](https://www.cnbc.com/2024/07/30/microsoft-msft-q4-earnings-report-2024.html?ref=wheresyoured.at) __Microsoft missed on cloud revenues in Q4 FY2025__[.](https://www.cnbc.com/2024/10/31/amazon-amzn-q3-earnings-report-2024.html?ref=wheresyoured.at) __caused its stock to pop because overall revenues were higher than expected__In fact, I think that’s mostly what kept this going. [ I ran the numbers](https://docs.google.com/spreadsheets/d/1J2J-24rrlxgbqH5KzJcDG4E6Hp3w5j9FG_Oz-ALu4KE/edit?usp=sharing&ref=wheresyoured.at) on the premium over two five-year-long periods — 2015 to 2020 and 2021 to 2026 — and found that while stock returns were dramatic, actual revenue growth has slowed dramatically. As you can see, revenue growth, outside of Microsoft, slowed dramatically, all as PP&E (properties, plants and equipment, the part of the balance sheet where they keep GPUs and data centers — and other stuff, obviously) grew by *$754.5 billion. * I’ll get back to that in a little bit. Yet because *the stock price went up*, everybody assumed that every dollar of revenue came from investments in AI GPUs. In 2023, a year when AI likely contributed less than $4 billion including OpenAI’s compute spend, Business Insider said that its AI bet was “[ already paying off](https://www.businessinsider.com/microsoft-earnings-show-bet-on-ai-bing-already-paying-off-2023-4?ref=wheresyoured.at),” all because Azure kept growing: CFO Amy Hood said on a call with analysts that Microsoft expects revenue growth at its cloud division, Azure, to be 26% to 27% next quarter from the same quarter the year prior — with "roughly 1 point from AI services." To be clear, *one whole revenue point is pathetic. * Anyway, in both [ April](https://www.theguardian.com/technology/2024/apr/25/microsoft-earnings?ref=wheresyoured.at) and October 2024, [that Microsoft was “sailing” as the “AI boom fueled double-digit growth in its cloud business” in a year where (based on working back from Wells Fargo’s estimates for FY25, which started in Q3 2024) it’s estimated to have made less than $6 billion in total revenue from](https://www.theguardian.com/technology/2024/oct/30/microsoft-earnings-increase-ai?ref=wheresyoured.at) __The Guardian reported__*anything*AI-related outside of OpenAI. Microsoft’s total revenue for that fiscal year was $281.7 billion. In October 2025 — the end of Fiscal Year 2025 — [ Business Insider would again say that its AI bets had paid off](https://www.businessinsider.com/satya-nadella-96-million-pay-salary-microsoft-ai-filing-2025-10?ref=wheresyoured.at), specifically adding that “Microsoft's AI push also increased revenue by 15% to $281.7 billion.” Per Wells Fargo’s estimates, Microsoft’s total AI revenue for FY2025 — including what it received from OpenAI — were $14.86 billion, or around 5.28% of revenue, or roughly 6.1% of annual growth for in a year it spent $64.6 billion in capex. When you remove OpenAI’s estimated $9 billion in compute spend, that leaves around $5.7 billion in AI revenue, or around 2% of overall revenues for the year. The rationale is pretty simple: - The stock price kept going up. - The revenues kept going up. - The executives kept ( *vaguely*) giving AI credit for growth. - Hyperscalers kept spending tens or hundreds of billions in capex. - Everyone assumed that these were “smart people” that “wouldn’t spend all that money without there being a massive return.” Well, they did, and there wasn’t. You can fart around all you want about the theoretical or imaginary promises of AI or AGI or whatever, but this didn’t work. # The Media and The Markets Helped Launder AI’s Reputation, Inflate The Bubble, and Explain Away Its Obvious Failures Yet all of this kept going because the tech industry’s collective reality is based on stock prices, Twitter, and a tech and business media that appears to fall for just about anything as long as a wealthy person says it. And this chart is the entire reason: I maintain that 2021 broke the world for many reasons, but one of them is that it set unrealistic revenue goals as money flooded back into the economy post-pandemic off the back of the most-pornographic years of Zero Interest Free Money Policy. To explain, I’m going to crib a little [ from my latest premium](https://www.wheresyoured.at/premium-the-haters-guide-to-nvidia-part-2/#how-nvidia-used-hyperscalers%E2%80%99-post-2021-desperation-to-create-fomo-for-gpus-and-imaginary-demand-at-scale), The Hater’s Guide to NVIDIA (Part 2). The post-2021 hangover was brutal. Meta, Google and Amazon, all of which have fiscal years that align with the calendar year, saw growth deteriorate: - Google went from 41.6% year-over-year growth in 2021 to 10.3% and 9.7% in 2022 and 2023. - Meta went from 37.2% year-over-year growth in 2021 to negative 1.1% in 2022 and 15.7% in 2023. - Amazon went from 37.6% year-over-year growth in 2020, to 21.7% in 2021, to 9.4% in 2022, to 1.8% in 2023 (and has really never recovered.) Microsoft’s FY2023 (July 1, 2022 through June 30, 2023) revenues only grew 6.9% year-over-year. NVIDIA’s FY2023 (which ran February 2022 to January 2023) was effectively flat, sitting at 0.2% as the post-pandemic surge of demand for gaming GPUs and networking gear puttered out, [ with Q4 2023](https://s201.q4cdn.com/141608511/files/doc_financials/2023/Q423/Q4FY23-CFO-Commentary.pdf?ref=wheresyoured.at) revenues dropping by 21% year-over-year. Nobody really knew what to do, [ with just about everybody getting their asses handed to them by the markets](https://www.cnbc.com/2022/10/28/big-tech-falters-on-q3-2022-results-as-meta-has-worst-week-ever.html?ref=wheresyoured.at). Yet the savior was already incubating. In March 2022, [ NVIDIA announced the Hopper GPU architecture](https://nvidianews.nvidia.com/news/nvidia-announces-hopper-architecture-the-next-generation-of-accelerated-computing?ref=wheresyoured.at), and while initial sales and shipments in September were good, they weren’t enough to restart growth until the November launch of ChatGPT convinced everybody that they had to do AI, and that the only way to “do AI” was buy GPUs. The very same month, [.](https://techcrunch.com/2022/11/16/microsoft-and-nvidia-team-up-to-build-new-azure-hosted-ai-supercomputer/?ref=wheresyoured.at) __Microsoft and NVIDIA announced they were building another OpenAI supercomputer using Hopper__Something about ChatGPT would fundamentally break the brains of Microsoft’s competitors. At 1 p.m. on a Friday shortly before Christmas [2022], Kent Walker, Google’s top lawyer, summoned four of his employees and ruined their weekend… …[because] the entire agenda of the company had changed — all in the course of nine days. Sundar Pichai, Google’s chief executive, had decided to ready a slate of products based on artificial intelligence — immediately. He turned to Mr. Walker, the same lawyer he was trusting to defend the company in a profit-threatening antitrust case in Washington, D.C. Mr. Walker knew he would need to persuade the Advanced Technology Review Council, as Google called the group of executives, to throw off their customary caution and do as they were told. It was an edict, and edicts didn’t happen very often at Google. But Google was staring at a real crisis. Its business model was potentially at risk. What had set off Mr. Pichai and the rest of Silicon Valley was ChatGPT, the artificial intelligence program that had been released on Nov. 30, 2022, by an upstart called OpenAI. It had captured the imagination of millions of people who had thought A.I. was science fiction until they started playing with the thing. It was a sensation. It was also a problem. ChatGPT immediately gave executives AI psychosis. [ In January 2023](https://www.cnbc.com/2023/01/23/microsoft-announces-multibillion-dollar-investment-in-chatgpt-maker-openai.html?ref=wheresyoured.at), Microsoft would invest another $10 billion, and a few weeks later, [as its “preferred cloud provider,”](https://www.googlecloudpresscorner.com/2023-02-03-Anthropic-Forges-Partnership-With-Google-Cloud-to-Help-Deliver-Reliable-and-Responsible-AI?ref=wheresyoured.at) __Google would sign a partnership with early-stage AI firm Anthropic (founded by former OpenAI executives) to use its TPUs and GPUs__[, making AWS “Anthropic’s primary cloud provider,” which forced](https://techcrunch.com/2023/09/25/amazon-to-invest-up-to-4-billion-in-ai-startup-anthropic/?ref=wheresyoured.at) __only for Amazon to barge in and invest $4 billion a few months later in September 2023__[. By the third quarter of 2023,](https://www.cnbc.com/2023/10/27/google-commits-to-invest-2-billion-in-openai-competitor-anthropic.html?ref=wheresyoured.at) __Google to invest up to $2 billion in October 2023__[, primarily to Microsoft, Amazon, Google and Meta,](https://www.tomshardware.com/tech-industry/nvidia-ai-and-hpc-gpu-sales-reportedly-approached-half-a-million-units-in-q3-thanks-to-meta-facebook?ref=wheresyoured.at) __NVIDIA would be selling half a million H100 GPUs__[.](https://apnews.com/article/meta-ai-zuckerberg-llama-chatgpt-9431120efcc8e598c3d34af9b5201d1c?ref=wheresyoured.at) __which had just farted out its own open source ChatGPT “competitor,” Llama__NVIDIA was saved. [ Q1 FY24 (May 2023) revenues blew estimates out of the water](https://www.cnbc.com/2023/05/24/nvidia-nvda-earnings-report-q1-2024.html?ref=wheresyoured.at), and [, data center demand caused revenues to jump 170% year-over-year. Microsoft, ever helpful to its good friend and collaborator,](https://www.cnbc.com/2023/08/23/nvidia-nvda-earnings-report-q2-2024.html?ref=wheresyoured.at) __by Q2 FY24 (August 2023)__[, allowing it to](https://www.cnbc.com/2023/06/01/microsoft-inks-deal-with-coreweave-to-meet-openai-cloud-demand.html?ref=wheresyoured.at) __would sign a multi-billion dollar deal with CoreWeave to rent capacity in June 2023__[, taking advantage of the “ChatGPT moment” that was when “things got real” to quote CTO Brian Venturo.](https://www.coreweave.com/news/coreweave-secures-2-3-billion-debt-financing-facility-led-by-magnetar-capital-and-blackstone-to-meet-surging-demand-and-ongoing-expansion-of-specialized-cloud-infrastructure-to-power-ai?ref=wheresyoured.at) __raise $2.3 billion in debt a few months later__By the end of FY24, NVIDIA’s revenue had jumped 125.9% year-over-year. Everybody went AI crazy. The [ media would fall over itself claiming that AI could do basically anything](https://www.wheresyoured.at/how-to-argue-with-an-ai-booster/#:~:text=publish%20outright%20misinformation.-,Roose,-%2C%20along%20with%20his), and justify one of the largest expenditures in history. This was partially helped [, which was mostly caused by NVIDIA being the only vendor and selling the vast majority of them to hyperscalers who were yet to really show any return on their investment. Per the New York Times:](https://www.nytimes.com/2023/08/16/technology/ai-gpu-chips-shortage.html?ref=wheresyoured.at) __by a media-driven hype campaign around the availability of GPUs__Their desperation is palpable. On social media, blog posts and conference panels, start-up founders and investors have started sharing highly technical tips for navigating the shortage. Some are gaming out how long they think it will take Nvidia’s wait-list to clear. There’s even a groan-worthy YouTube song, set to the tune of Billy Joel’s “We Didn’t Start the Fire,” in which an artist known as Weird A.I. Yankochip sings “GPUs are fire, we can never find ‘em but we wanna buy ‘em.” Nevertheless, revenue was growing, seemingly in line with capital expenditures, and as hyperscalers realized that the media and the markets had toddler-like attachments to reality, they piled into NVIDIA GPUs en masse. And man, FOMO was in full force. The GPU shortage was timed perfectly with [ one of the worst years in the history of venture capital](https://www.cnbc.com/2022/05/28/start-up-investors-issue-warnings-as-boom-times-unambiguously-over.html?ref=wheresyoured.at), creating an air that the only way to get out of the depths of Hell was to invest in AI in any way, shape or form for both startups and hyperscalers alike. Yet [ when you look at the numbers](https://docs.google.com/spreadsheets/d/1J2J-24rrlxgbqH5KzJcDG4E6Hp3w5j9FG_Oz-ALu4KE/edit?usp=sharing&ref=wheresyoured.at), very little actually changed for the hyperscalers. Since the launch of ChatGPT, year-over-year growth has never returned to pre-2022 levels, other than for Microsoft, which hit its highest year-over-year growth (17.8%) since FY2022 (18%) after a prolonged period in the 14-percents. Everybody conflated the massive capex spend with the return of growth to the tech industry versus an industry-wide swindle. Hyperscalers were rewarded with stock pumps and __pay__[ bumps](https://www.businessinsider.com/satya-nadella-96-million-pay-salary-microsoft-ai-filing-2025-10?ref=wheresyoured.at) for an “AI revolution” that mostly amounted to spending hundreds of billions of dollars on GPUs to make tens of billions of dollars in revenue, all because revenue kept growing and both analysts and the media refused to talk loudly about the lack of any payoff. The media’s credulousness was used against it. The assumption, as I mentioned, was that all this money wouldn’t be spent without an obvious return, and because the numbers are so *flabbergasting*, it’s easy for you to say “$10.33 billion is a lot of money!” (because it is) and to dismiss the massive costs as “just part of building the infrastructure,” even if it isn’t clear how these numbers ever match up in the future. For whatever reason, the media continues to give hyperscalers and anybody in AI the benefit of the doubt when it comes to the efficacy and outcomes of large language models or the catastrophic economic mismatch in the returns. Every time the response is “these are smart people!” or “these are the early days!” or “it’s just like the dot com bubble!” because nobody is particularly interested in being *right* so much as they are about *being right about the particular consensus of a particular moment.* While I understand the professional harms of saying that AI was bullshit in 2023 or 2024, there was never any excuse to automatically give hyperscalers credit for “AI paying off” at any point in history, and further excuses of “it being the early days” are intellectual crutches” used by people that either want the powerful to win, have a vested interest in doing so, or have resigned themselves to watch it happen. The fact that the media has actively shrugged off the 70% story (outside of Bloomberg’s coverage, at least) is a sign that it doesn’t really want to reconcile with the truth, and honestly, *I kind of get it.* When you’ve spent three years saying that hyperscalers were growing because of their vast spend on AI, filling in the gaps of every narrative and assuming they don’t want to tell you revenues *because they’re oh-so-good*, it’s hard to move in reverse. The other problem is the monstrous and abusive marketing campaign from the AI industry itself, and those who use AI on a regular basis. If you are against the consensus that AI will grow ever-larger every single quarter forever, you will be harassed and dogpiled across multiple social media platforms by everyone from AI influencers to actual journalists. The fact that it’s more professionally dangerous to critique the powerful than it is to align with them is disgusting, but I should be clear that these tactics only reinforce that I’m on the right track. [ As Nik Suresh noted in his recent piece](https://ludic.mataroa.blog/blog/ai-mania-is-eviscerating-global-decision-making/?ref=wheresyoured.at#:~:text=II.%20Heretics%20Will%20Be%20Shot), refusing to say that AI is giving you massive productivity benefits will lead to actual professional consequences, because so much is riding on the overall grift about what AI can do (which is much, much less than the boosters will promise). This runs antithetical to productivity or good sense, and everybody involved in it should be both eternally shamed and shunned from any sensible business. And while the AI industry and its fandom will claim that people like me are “skeptics” and “haters,” the outright hatred and vitriol that they spew for not falling in line behind a nakedly false narrative built on outright disinformation is disgraceful. It only serves to prove my point that something is very, very wrong with the tech industry, our markets, and the overall information ecosystem. # The Consequences of The Rot-Com Bubble When [ I wrote the Rot-Com Bubble in 2024](https://www.wheresyoured.at/rotcombubble/), the AI bubble was a series of exploits used to mask the end of tech’s era of hypergrowth. The tech media’s immediate attachment to AI and ChatGPT was nothing to do with actual technology and everything to do with their attachment to being involved in whatever future the powerful decided had arrived. After years of depressing coverage of cryptocurrency and the metaverse and a deeply-depressing 2022, ChatGPT represented a product they could use, be built upon to create other products they could use, and lead to an entire era of new people to follow and report on. It gave retail investors a reason to dump money into stocks. It gave founders an API to build on top of, an infrastructural layer to smooth out, and a dream to sell venture capitalists who [ had near-unilaterally sucked at their jobs for years](https://www.wheresyoured.at/the-enshittifinancial-crisis/#the-devil%E2%80%99s-deal-of-investing-in-ai-startups:~:text=nothing%20else.-,In,-fact%2C%20investing%20in), all wrapped in a fuzzy sense of “progress” that mean that valuations could be high and the time horizon for returns could be effectively infinite. And that last part is what was so important for *everybody* involved. Startups, Microsoft, Google, Amazon, Meta, CoreWeave, and anyone else involved in the AI bubble were *immediately* given near-infinite runway and manufactured consent to burn as much money as they wanted to. Even today — in the third quarter of the year of our lord 2026 — I am still asked on podcasts “whether we’re in the early days.” This is the power of narratives, and how willing so many people are to explain away the failures of the powerful rather than having the courage to face them. And it’s all because the stock prices haven’t gone down yet. Sidenote:Need an example?[, boasting that its “annualized revenue” (a month times twelve, also known as $29.1 million a month in revenue) has hit $350 million.]Legal AI startup Harvey is currently raising a $500 million funding round at a valuation of $15.5 billion Harvey last raised in March —[, two months]$200 million at a $11 billion valuation[, better known as $16.6 million in monthly revenue. If it completes the round as discussed, Harvey will have raised $1.2 billion in the last year or so, and I have yet to read a single story that reconciles a company making less than $30 million in monthly revenue needing over a billion dollars, or shows any kind of concern about that at all.]after it hit $190 million in annualized revenue The argument here, by the way, is that “the goal for Harvey is to get every law firm as a client.” I have to give credit to Jensen Huang, though. He realized quickly that reality is dictated not by actual revenues but how you can manipulate the market with those revenues. NVIDIA’s continued position as the largest company on the NASDAQ relied upon a near-constant flow of new orders of GPUs, but brainwashed investors, misinformed by a media with few good information sources, had begun to conflate both GPU purchases and any revenue growth with purchasing them. The power of the narrative is such that everybody has rationalized what’s happening as “good investment.” Hyperscalers spending hundreds of billions on capex *makes sense* because *AI is driving growth,* and the *market is at all time highs. * The same extends to the endless flow of nebulous circular deals, like [ the supposed $500 billion infrastructure deal between NVIDIA and The Avengers of Private Credit, including Apollo, Blackstone, BlackRock, Brookfield, Goldman Sachs, and KKR](https://www.cnbc.com/2026/08/10/nvidia-wall-street-asset-managers-500-billion-ai-push.html?ref=wheresyoured.at). While NVIDIA dumped a little on the deal, the media still doesn’t seem to think this is a bad thing, even though it’s the loudest possible sign that there isn’t actually *real demand from anyone that can actually afford to buy these GPUs.* Even then, the “$500 billion” deal isn’t even a $500 billion deal, [ per The Wall Street Journal](https://www.wsj.com/business/deals/nvidia-wall-street-firms-strike-ai-financing-deal-targeting-500-billion-c50377db?ref=wheresyoured.at) (which has, for some reason, deleted this paragraph from the story): The massive financing would involve different vehicles, rather than one large collaboration, according to one of the people. The commitments could grow larger than $500 billion over time, this person said. Oh, okay. So it’s not actually $500 billion. It’s a bunch of smaller deals. Great. Sure. Anyway, surely you must think this is a little worrying? That this is what NVIDIA is reduced to doing to keep up with demand? What? You think it’s a *good thing?* The financing could help signal to stock investors that Nvidia and its partners in the artificial-intelligence boom have the firepower to build the infrastructure they need. Debt investors have lent companies hundreds of billions of dollars over the past year through direct bond sales and debt deals tied to individual data-center projects. Jesus fucking *christ.* Not a single word about the fact that the only two companies that would actually want this compute can’t afford it, or how much people are spending on compute (you know, the thing that data centers sell), or anything about AI at all beyond that it’s “part of the infrastructure buildout.” Yet when you actually open the press release (which I found on my Terminal but cannot for the life of me get a link to), there’s one glaring detail everybody left out, emphasis mine: Memorandums of understanding signed with six of the world’s premier financial institutions to create these partnerships aim to establish the first compute financing platforms of their kind at global scale to enable the AI infrastructure buildout across NVIDIA’s ecosystem, including leading frontier AI labs, enterprises and AI clouds. Under these strategic partnerships, NVIDIA will work with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to create dedicated pools of capital at significant scale at attractive rates for NVIDIA customers. That’s right folks. __There is no deal! It’s an MOU! It’s fucking theoretical!__ And I have not seen this fact reflected *in a single god damn story about this god damn “deal” in one god damn place!* **It’s from the press release from the fucking companies!** This “deal” is also very, very weird, with NVIDIA claiming it’s “establishing independent compute financing platforms” with the largest asset managers in the world. It isn’t clear what the money will do, where the money will flow, who it will flow to, when it will flow there, how it will be structured, from whom the money will be raised, or really anything other than “number so big, number so huge.” This announcement — and that is, at this point, all it is — exists entirely to have people say that NVIDIA has “*booked $500 billion in revenue*,” even though even in the *kindest possible read* not a *single dollar has actually been raised*, nor has a *single actual contract been signed. *If you need an example, take NVIDIA’s $100 billion investment in OpenAI that also involved it building 10GW of compute capacity — [ a memorandum of understanding that never materialized in a deal](https://www.wsj.com/tech/ai/the-100-billion-megadeal-between-openai-and-nvidia-is-on-ice-aa3025e3?ref=wheresyoured.at). [ Here is what Jensen Huang had to say about said memorandum of understanding](https://www.bloomberg.com/news/articles/2026-02-01/openai-investment-was-never-a-commitment-nvidia-s-huang-says?ref=wheresyoured.at): “It was never a commitment,” Huang told reporters in Taipei on Sunday. “They invited us to invest up to $100 billion and of course, we were, we were very happy and honored that they invited us, but we will invest one step at a time.” You’ll notice there are no actual details about any deals happening, mostly because nothing has actually happened beyond a few marketing calls and a lot of heavy breathing from the press. No money has been raised, what will likely happen — if anything — is that NVIDIA will end up backstopping a few $10 billion data center deals, or perhaps invest a few billion in equity into an SPV built to raise debt to buy GPUs [ as it already did with xAI](https://www.datacenterdynamics.com/en/news/nvidia-to-invest-in-xai-special-purpose-vehicle-cash-to-be-used-to-buy-nvidia-gpus/?ref=wheresyoured.at). Jensen Huang has said as much [ in his hilariously-oafish announcement of the MOU](http://x.com/jensenhuang/status/2086934705207959965?s=46&ref=wheresyoured.at): In some cases, NVIDIA may provide a residual-value support mechanism for up to 25% of an opportunity, assessed carefully on a project-by-project basis. That support is limited, residual-value based and designed to complement — not replace — independent underwriting. Hey, wait a second, is this circular financ- Is this circular financing? This initiative is designed to address that concern. We are bringing independent, long-term institutional capital into the AI infrastructure market. The demand is real: it comes from frontier AI labs, AI-native startups, enterprises, cloud providers and countries building AI services. The capital providers independently underwrite each project — including the customer, demand, utilization, cash flow and residual value. NVIDIA provides the platform; the investors make independent financing decisions. This is the beginning of an open capital market for AI infrastructure. Folks, this isn’t circular financing at all! It’s just that NVIDIA will pay some sort of 25% “residual-value support” so that private credit can use that as collateral to raise debt to *buy GPUs from NVIDIA*. If anything it’s spherical! Look, Jensen, *if the demand was real, you wouldn’t have to announce a rinky-dink-maybe-$500-billion-no-IT-loads-refused-MOU!* If there were actual diverse demand for NVIDIA’s GPUs commensurate with analyst expectations, you wouldn’t have to do these bizarre, painfully-circular deals that exist only to inflate its revenues and further prop up the existence of unprofitable AI labs! Anyway, if you’re wondering about what the point of this all is, Jensen Huang has your answer: Where is the return on investment? The return is in the usefulness of AI. Companies are using AI to write software, discover drugs, design products, serve customers, automate operations and build new services. AI factories make this possible. More compute creates better AI; better AI creates more usage; more usage creates more revenue; and more revenue drives more compute. This is the virtuous cycle of the AI industrial revolution. *We are four fucking years and over a trillion dollars into this garbage, Jensen! This is the best you can do? THIS? AHHHHH!* # The Bubble Is Built On Ridiculous Expectations, And Everyone Needs To Wake Up In FY2027 — which began on February 1, 2026 — analyst consensus has NVIDIA’s revenues at $393.7.6 billion for the year, growing to $565.7 billion in FY2028 and $694 billion in 2029. What this means is that despite hyperscalers spending over a trillion dollars on AI data center capex in 2026 and even more in 2027, *that’s just not enough to keep up with Wall Street’s expectations.* To get specific, NVIDIA’s FY2026 revenues were $215.9 billion, with 89% of that coming from the data center segment (read: GPUs and the associated gear), and this is with the near-entire focus of the world’s largest companies and credit markets on building AI data centers and [ banks that fear they’re “choking” on data center debt](https://www.ft.com/content/08aba5e4-5834-4e79-a48d-989a2c5bad0f?syn-25a6b1a6=1&ref=wheresyoured.at). Sidenote:[, UBS estimates that around 50% of NVIDIA’s data center revenue comes from Meta, Google, Microsoft, Amazon and Oracle, with Deutsche Bank estimating it’s as high as 60%.]per my last premium newsletter Analyst expectations are set to believe that it will triple its revenue in the space of two years. Honestly, $500 billion wouldn’t even be enough. NVIDIA needs every hyperscaler to keep spending more capex every single quarter, without fail, as well as hundreds of billions of dollars’ worth of *new* AI chip spend to arrive *from an industry where the only two companies with any real need for all this compute have only ever lost tens of billions of dollars, and literally can’t afford to pay for it.* I realize many people get number blindness past a certain scale, so I will put it very simply: - There is not enough money to keep up with analyst expectations for NVIDIA’s revenue. Even with every hyperscaler buying more and more GPUs every quarter, it will have to triple revenue in the next three years to keep up, at a time when (and negative in the case of Google and Amazon), making further capex contingent on debt, as AI revenues are not covering their costs.__hyperscaler cashflows are deteriorating__ - There are only two companies that actually spend more than a few hundred million a year on AI compute — OpenAI and Anthropic — and both of them will lose tens of billions of dollars this year and more next year. - Outside of these two companies, the only other companies spending more than a few hundred million a year are either the hyperscalers renting capacity to sell back to Anthropic and OpenAI, or Meta, a company that does not have an AI strategy or meaningful revenues. - No, AI is not “helping ads.” Meta has actually said this. They have mentioned incremental, single-digit engagement boosts in a few blogs, and everybody interpreted from there. I also understand why people want to bury their heads in the sand here. Right now, the numbers are all the highest they’ve ever been, and they keep going up, which means that anyone saying that things are going wrong has to expose themselves to torrents of abuse and aggression from both posters and peers. I also think doing so is an act of cowardice. While I don’t expect people to start saying that this is all bullshit and headed for the gutter, I see an astonishing flippancy about everything I’ve been writing about from much of the mainstream media. To not warn people that AI revenues are heavily-centralized around two companies that burn endless billions of dollars, and that the commensurate demand isn’t there as a result, is to both fail your readers and actively empower the powerful. I haven’t even gotten into [ the $1.65 trillion in off-balance sheet obligations](https://archive.ph/lOlv5?ref=wheresyoured.at). It’s unclear how hyperscalers afford them if OpenAI and Anthropic can’t afford to pay them. It’s unclear how *any* of this works. And then there’s the problem that [ for the 190GW of capacity in planning, we need somewhere between $1.62 trillion and $2.92 trillion in annual compute spend](https://www.wheresyoured.at/premium-ai-is-getting-way-too-expensive/#the-190gw-of-planned-ai-data-centers-need-roughly-162-trillion-to-292-trillion-in-annual-compute-demand) for an industry that, in the best-case scenario, has roughly $110 billion in annual demand, with 90%+ of that coming from two companies that can only spend that money if they’re fed it by venture capitalists. Everyone can — and will — keep ignoring what’s happening as long as it requires an ounce of courage to think about reality. Many will bury their heads in the sand, make the kindest reads possible of every AI story, pump up every AI narrative, and celebrate every mediocre “achievement” right up until NVIDIA or a hyperscaler misses on analyst expectations, or AI labs start circling the drain. Everything you’re seeing right now is an attempt to extract further capital and hype from the system to continue inflating the bubble. Every single asset manager in the NVIDIA “deal” has some sort of investment in AI data centers and/or neoclouds (especially Blackstone, who has been in CoreWeave since its earliest days), and absolutely nobody gives a shit about what LLMs can do outside of their ability to con investors and generate fees for their funds. Every journalist that continues to ignore the obvious instability, circularity and centralization of the AI industry fails their audience by not discussing it* before* any discussions of AI’s theoretical returns or abilities. It is no longer ethically sound to ignore the problems. Do with that statement what you will. In the end, it’s pretty simple: Microsoft, Google and Amazon’s hypergrowth era ends if Anthropic and OpenAI can’t spend hundreds of billions of dollars on compute, and companies like Oracle, CoreWeave, Nebius, and IREN face apocalyptic circumstances when they fail to do so. NVIDIA’s future is entirely dependent on these companies’ abilities to convince the credit markets that everything will go fine, and its circular financing operations are both sustained and continue to expand. And every VC invested in AI needs a miracle, because there are few signs that these companies can be sold to anyone or taken public. When it becomes safe to do so, many will attempt to either rationalize ignoring reality or pretend they saw it coming. If they did, they chose not to tell you. If they didn’t, they chose not to look. If you liked this piece, you should subscribe to my premium newsletter. It’s $70 a year, or $7 a month, and in return you get a weekly newsletter that’s usually anywhere from 10,000 to 18,000 words, including vast, detailed analyses of the biggest events and companies in the AI bubble. If you want to get in touch — and especially if you have any juicy information about Anthropic, OpenAI, or any other companies in the AI bubble — hit me up on Signal at ezitron.76. I’m also on IB on The Terminal.