Ed Zitron: The AI Bubble is Bleeding Cash, Here Are The Receipts
Speakers
Ed Zitron
— Author of Where's Your Ed at Newsletter, Host of Better Offline podcast
Summary
Ed Zitron delivers a comprehensive bear case on the AI industry, arguing that massive losses at OpenAI and Anthropic, combined with a data center overbuild and circular financing, make the sector a bubble. He advises shorting Nvidia, Microsoft, Broadcom, and the S&P 500, and avoiding any IPOs from leading AI labs.
- OpenAI lost $21B in 2025 with opaque accounting and no path to profitability.
- Microsoft's revenue is artificially inflated by OpenAI's unsustainable spending.
- Nvidia's sales depend on circular financing from unprofitable AI customers; a guidance miss could sink the stock.
- Broadcom is backstopping a $30B deal for Anthropic, creating direct exposure to an unprofitable AI company.
- Amazon and Google's cloud RPOs are heavily tied to Anthropic, posing a risk if it defaults.
- Both OpenAI and Anthropic IPOs are dangerous for retail investors given their broken economics.
- The largest S&P 500 companies face massive hidden costs from AI data center leases, threatening the index.
- Meta's AI strategy is incoherent but the company may outlast others due to Zuckerberg's unfireable status and cash flow.
Ideas
Avoid OpenAI IPO, massive losses.
OpenAI's financials reveal a $21 billion loss in 2025, bizarre accounting, massive sales and marketing costs that likely hide inference expenses, and no credible path to profitability. When the company goes public, retail investors will be the victims of a hype-driven IPO. Avoid participating.
Microsoft revenue inflated by OpenAI losses.
Microsoft's cloud revenue is heavily dependent on OpenAI, a deeply unprofitable company that lost $21 billion in 2025. A significant portion of Microsoft's Azure revenue comes from OpenAI spending (including credits and sales and marketing costs), inflating Microsoft's top line. When OpenAI runs out of money, Microsoft faces a material revenue hole.
Avoid Anthropic IPO, overhyped and unprofitable.
Anthropic suffers from the same broken economics as OpenAI. Its profitability was temporarily engineered through a discount from Elon Musk, its annualized run rate is exaggerated by a flawed calculation, and its cost structure worsens with revenue. An Anthropic IPO would be equally dangerous for public investors.
Amazon and Google clouds exposed to Anthropic.
Amazon and Google have massive remaining performance obligations from Anthropic (and indirectly OpenAI), which are used to justify their cloud buildouts. Since Anthropic is unprofitable and its revenue is overstated, these obligations are at risk of not being fulfilled, threatening the hyperscalers' revenue growth and balance sheets.
Broadcom backstops risky Anthropic deal.
Broadcom is backstopping a $30 billion chunk of a $35 billion deal that funds Anthropic's compute via Google. If Anthropic, an unprofitable company, fails to pay, Broadcom will be on the hook. The circular financing structure is foolish and exposes Broadcom to significant credit risk.
S&P 500 companies burdened by AI debt.
The largest companies in the S&P 500 have entered into hundreds of billions of dollars in lease commitments for AI data centers that lack real demand and will produce massive costs. As the AI bubble pops, these balance-sheet burdens will hurt the index, making it a short.
Nvidia's inflated sales will crash.
Nvidia's revenue is artificially inflated by circular financing from unprofitable AI companies like OpenAI and Anthropic. The company has set unrealistic expectations, and if its guidance fails to meet market hopes over the next few earnings calls, the stock will fall apart. Nvidia needs a trillion dollars in sales through 2027 to justify the bubble, which is highly unlikely.
