Let AI Burn

Author: Ed Zitron URL: https://www.wheresyoured.at/let-ai-burn/ One-line: A polemic arguing the generative-AI industry is a financial bubble propped up by circular financing and manufactured hype, that its revenues are dwarfed by its costs, and that when it collapses there should be no bailout, tax break, or sovereign wealth fund to save it.

Key claims

• OpenAI and Anthropic together represent an estimated 89% of the revenue of the largest AI companies; strip them out and the rest of the AI industry pulls in roughly $20 billion in annual revenue. OpenAI had $5.7 billion and Anthropic a little under $5 billion in Q1 2026 revenue, much of it from customers burning tokens at prices that don’t reflect true compute cost.
• Generative AI is “mathematically certain to hallucinate” (per OpenAI’s own admission) and is not, in the author’s view, “autonomous” or “intelligent” in the sense the industry markets it as.
• Big Tech capex (Microsoft, Google, Meta, Amazon) — cited at $765 billion in 2026 and a projected $1 trillion+ in 2027 — is framed as compensating for a lack of organic hypergrowth ideas (“Google does not have the next Google Search… Amazon does not have the new AWS”), not as evidence of proven AI demand.
• Draws a contrast with the Dot Com bubble: internet demand in 2000 was real (people wanted to get online) even though the infrastructure buildout outpaced it; AI GPU demand, by contrast, is argued to be “decidedly artificial” outside of OpenAI and Anthropic’s own compute purchases, evidenced by NVIDIA reportedly agreeing to backstop cloud providers’ unused GPU capacity.
• Argues fears of China “catching up” in AI are overblown — cheaper open-source Chinese models (e.g. Zhipu’s GLM-5.2, cited as rivaling Claude Opus 4.8 on coding benchmarks “at a fifth of the cost”) have already happened, and “nothing” changed as a result. See China.
• Sam Altman’s floated idea of OpenAI handing 5% equity (~$42 billion) to a sovereign-wealth-style public vehicle is characterized as not a bailout in any meaningful sense — $42 billion is less than OpenAI’s projected 2026 compute spend, against a projected $852 billion burn through 2030.
• Distinguishes the AI moment from the 2008 financial crisis: OpenAI and Anthropic are “systemically irrelevant” (unlike AIG), have little debt, and a collapse would mostly hurt equity holders and venture capital vintages rather than the broader financial system — with the exception of SoftBank, called out as uniquely exposed via tens of billions in short-term loans tied to OpenAI.
• Concludes with an explicit policy demand: no bailouts, no handouts, no tax breaks, no CHIPS-Act-style support, no sovereign wealth fund — let the industry fail on its own economics.

Notable quotes

“No bailouts, no handouts, no special treatment, no tax breaks, no CHIPS act, and no sovereign wealth fund. It is time to tell the AI industry to go fuck itself, because it’s effectively done the same to the rest of society.”

“I estimate that at least 70% of the entire AI industry’s revenues are made up of OpenAI and Anthropic’s compute spend… the AI industry is, for the most part, venture capitalists funnelling money to hyperscalers so that they can funnel that money to NVIDIA or data center capex.”

“GPUs aren’t dark fiber! GPUs aren’t fucking railroads! GPUs are GPUs! They are used for basically one thing! And that one thing lacks meaningful demand outside of subsidized services and circular financing!”

“Every single person freaking out about ‘what if China does AI better than America’ is living in a child’s fantasy… The only thing that American labs are scared of is cheaper open source Chinese models offering similar performance to their premium products, something that has also already happened.”

“When the time comes, the AI industry must burn. It must be allowed to die.”

Embedded tweets

it’s time to call Kevin O’Leary his real name: Mr. Dogshit

— Ed Zitron (@edzitron) · May 12, 2026 · View on X

Recently, we purchased one of each Anthropic/OpenAI subscription plan and randomly ran long horizon coding tasks until we exhausted the weekly limit. It’s widely believed that a $200/month plan maxes out at ~$2000/month worth of tokens (assuming API pricing). However, we found

— SemiAnalysis (@SemiAnalysis_) · Jun 10, 2026 · View on X

It was $50 billion in September 2025. How exactly is anyone meant to afford this at this point?

↳ quoting: NVIDIA CEO JENSEN HUANG: 1GW AI FACTORY ON NVIDIA ARCHITECTURE COULD COST NEARLY $100 BILLION

— Ed Zitron (@edzitron) · Jul 05, 2026 · View on X

Meta may be close to deal with Anthropic to host private cloud compute. $META

— Sean D. Emory (@_SeanDavid) · Jul 05, 2026 · View on X

How it connects

• Ed Zitron — the author; this is his central thesis piece on the AI bubble.
• OpenAI — one of the two companies (with Anthropic) the piece argues dominates and distorts AI industry revenue figures.
• Anthropic — the other dominant company in the piece’s revenue analysis; also its stance on Chinese open models (Claude Haiku vs. cheaper Chinese models finding the same vulnerabilities) is cited.
• Nvidia — central to the capex/bubble argument; the piece discusses NVIDIA backstopping cloud providers’ GPU purchases as a demand-fragility signal.
• Sam Altman — his floated 5%-equity-to-the-public proposal is analyzed and dismissed as not a real bailout.