The Man Who Calls BS On AI: AI Is The World’s Greatest SCAM, And They All Know It! | Ed Zitron
with Ed Zitron
27 Aug 20266 min read1h 30m
TL;DR
Ed Zitron argues that generative AI is fundamentally a con — OpenAI lost $20.9 billion last year, users on a $200/month plan can burn $14,000 worth of compute, and the 'AI revenues' credited to Microsoft, Google, and Amazon mostly come from two unprofitable companies that require constant cash injections to survive. He contends that widespread AI adoption is largely non-consensual, driven by corporate forcing functions rather than genuine user demand, and that the trillion-dollar infrastructure build has no credible path to profitability.
Key Moments
Ed Zitron
“Amazon sent $50 billion to OpenAI this year. They sent $5 billion to Anthropic. Google sent $10 billion to Anthropic.”
Zitron is explaining how the biggest cloud companies are funding the very AI startups whose revenues prop up their own 'AI growth' narratives.
Ed Zitron is a tech industry commentator and PR veteran with over 15 years of experience, known for his sharply critical newsletter on the technology sector. He has become one of the most prominent skeptical voices on generative AI, arguing that the industry is built on unsustainable economics and misleading claims. His writing challenges the mainstream narrative around AI's transformative potential and the financial viability of companies like OpenAI and Anthropic.
Takeaways
1
AI subscriptions are massively subsidised by vendors Semi Analysis found that a $200/month ChatGPT subscriber can consume $14,000 worth of compute tokens, and even a $20/month user can burn $400. This means AI companies are effectively selling dollar bills for pennies to maintain user growth, with no disclosed path to closing the gap.
2
Enterprise AI 'adoption' collapsed when pricing turned real When OpenAI moved enterprises onto cost-reflective pricing in March 2026, Sam Altman admitted customers 'have a big problem with it.' Uber burned its entire annual token budget in three months. Actual willingness-to-pay at cost is far lower than headline adoption figures suggest.
3
Microsoft's AI revenue is largely circular money In fiscal year 2026, Microsoft generated ~$34 billion in AI-related revenue, but $24.1 billion came directly from OpenAI — a company Microsoft itself funds. That leaves roughly $10 billion of genuine external AI revenue against $115 billion in capex, with $175 billion planned for next year.
4
Big Tech AI revenues don't separate signal from noise Google, Microsoft, and Amazon refuse to break out AI-specific revenues clearly, instead using undefined 'annualised run rate' figures that can mean different things each quarter. When companies have good news, they disclose it — consistent opacity is itself informative.
5
Hallucinations remain a compounding problem in production While frontier models have reduced hallucinations on simple summarisation tasks, Zitron argues errors in complex workflows — code refactors, financial models, medical transcription — are multiplicative: each mistake builds on unchecked prior mistakes, especially as developers lose hands-on skills through over-reliance on LLMs.
6
AI infrastructure build-out dwarfs any plausible return The industry has already spent over $1 trillion in capex and plans another trillion next year, yet total external AI revenues outside of OpenAI and Anthropic — themselves cash-burn machines — amount to roughly $22 billion globally. Nvidia alone sold $215.9 billion in GPUs last fiscal year to support this.
7
Widespread AI use doesn't equal chosen AI use Gemini is surfaced by default in Google Search, Copilot in Word, and Rufus in Amazon — users encounter AI whether they opt in or not. Zitron argues this manufactured exposure is being counted as adoption evidence, distorting any honest demand signal.