Monologue: The Pale Horsemen Arrive

**Ed Zitron** (0:05)
And I looked, and behold, a pale horse, and the writer's name was the Smiling Man. This is your weekly Better Offline monologue, and I'm your host Ed Zitron.
This week I put out one of my most consequential newsletters of all time, called 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 will be 48% of all Google Cloud revenues. Which means that they will likely account for even more than 70%, but I wanted to be fair. This was both the colossal pain in the ass and the story that I knew would piss off a lot of people because of its huge ramifications. As a result, I've had a ton of pushback from people that either outright deny that this is the case or say, actually, it's good the two unsustainable companies are the majority of AI revenues for companies that spent over a trillion dollars. Actually, that's great. Anyway, 24 hours later, Bloomberg ran a story estimating based on OpenAI's $24.1 billion contribution to Microsoft's fiscal year 2026 revenues and previous statements, the OpenAI alone contributed to 70% or more of Microsoft's AI revenues for the year. For some context, Microsoft has spent $26.13 billion in capital expenditures since the beginning of 2022 And other than getting to have the swagger of someone who just laid out a 7,000-worth thesis founded in deep research and hard numbers, only to get their work validated in less than a day, this story also confirms my greatest fears about the state of the AI industry and the actual demand for AI compute. Microsoft has spent over $200 billion, and that's not including the $13 billion invested in OpenAI or the losses it's incurred running their services at cost, at least in 2024, but potentially beyond, to create a customer that is now worth 7% of its annual revenue that cannot sustain its existence without near constant flows of venture capital funding. Microsoft's year-over-year growth for fiscal year 2026 was about 17.7%, but when you remove OpenAI's $24.1 billion in revenue, which comes from OpenAI's compute spend and revenue share, Microsoft's year-over-year growth was more like 9.24%, the lowest it would be since fiscal year 2023 when it only grew by 7%. But there is one big difference, and that's, well, capital expenditures in the entire fiscal year 2023 were $31.8 billion, and Microsoft spent $35.8 billion in the fourth quarter of the fiscal year of 2026, a year when it spent $115 billion, and it wants to spend another $175 billion in capex in fiscal year 2027, which begins July 1st, 2026 It's annoying, but we're moving on. Oh, and OpenAI, arguably Microsoft's largest customer and responsible for nearly half of its growth in the last fiscal year, is going to run out of money. And by nature of it being 70% of Microsoft's AI revenues, it's very obvious that there are no other large customers that exist that could possibly take up all of this infrastructure, other than, of course, weird Wario Amadeus Anthropic, who is in exactly the same position as OpenAI, but with a slightly more annoying voice, because we're near the exponential.
This also means that any future revenue guidance and analyst expectations are now set under the explicit assumption that Anthropic and OpenAI will both have the need for and be able to pay for increasingly larger levels of compute at increasingly higher prices. Based on the analyst notes I reported on, which are in the newsletter, or at least quotes, UBS, Wells Fargo and Barclays all have the expectation that much of Google, Amazon and Microsoft's revenues are going to come from OpenAI and Anthropic for the next few years, and they don't seem to be diverging from that point in any way. Everyone just thinks that this is where it's going to come from. Any failure for these revenues to appear out of insolvency, lack of demand or unbuilt data center capacity, meaning they can't get paid, will piss off a Wall Street that's been brainwashed into believing that these three companies are growing because of diverse demand for AI services, rather than having pissed a trillion dollars up the wall to mostly to just feed themselves money. I'm not confident that the market will actually understand what Bloomberg has said. I'm looking now Microsoft is up 2%, Jesus fucking Christ. So let me spell it out for you. 70% of Microsoft's AI revenues are from either its revenue share of OpenAI's money or OpenAI spending money to rent out compute from Azure. This means that Microsoft is now dependent on OpenAI for future revenue growth, as the company now represents more than 7% of its annual revenues. And there does not appear to be significant contribution from anything outside of single-digit billions of annual revenue from Microsoft 365 Copilot, a product mostly sold by tricking people into signing contracts like Dr. Facilier in the beginning of Princess and the Frog. Microsoft believed it would create an independent ultra-growth engine from its investments in OpenAI. Instead, it has created a parasite. At 70% of AI revenues, OpenAI represents the failure of Satya Nadella's empire. Microsoft has spent $260 billion on infrastructure for a demand that does not exist outside of two different companies that burn tens of billions of dollars a year. And to make matters worse, OpenAI and Anthropic must grow their compute spend aggressively every single year to keep up with the expectations of analysts, which will mean it will need more and more money, which means anyone investing in Microsoft is effectively gambling on how long OpenAI and Anthropic can continue to raise money before they simply exhaust every available form of capital.