Ed Zitron warns OpenAI fragility

Ed Zitron’s latest warning is simple: OpenAI has become too expensive to run, too dependent on partners, and too committed to infrastructure to easily slow down. (youtube.com) (wheresyoured.at) He made that case in an April 2026 interview titled “How OpenAI ends and takes Oracle with it,” published by The Tech Report and linked to his April 24 premium post “How OpenAI Kills Oracle.” (youtube.com) (wheresyoured.at) The mechanics are not complicated. Frontier artificial intelligence models need massive clusters of chips and power to train new systems and answer user prompts, so growth can force costs higher even when revenue rises. (openai.com) (finance.yahoo.com) OpenAI’s own numbers show the scale of that machine. Chief Financial Officer Sarah Friar said annualized revenue passed $20 billion in 2025, up from $6 billion in 2024, while computing capacity rose to 1.9 gigawatts from 0.6 gigawatts. (finance.yahoo.com) At the same time, OpenAI has kept signing up more demand. Its 2025 enterprise report said more than 1 million business customers use its tools, ChatGPT message volume grew 8x, and API reasoning-token use per organization increased 320x year over year. (cdn.openai.com) Zitron’s argument is that this does not automatically create safety. He has written since July 2024 that OpenAI may need constant fundraising, major cost breakthroughs, and a stable Microsoft relationship just to keep the model working at current scale. (wheresyoured.at) The dependence piece is visible in OpenAI’s own infrastructure plans. On January 21, 2025, OpenAI, SoftBank, Oracle, and MGX announced Stargate, a project that said it intended to invest $500 billion over four years, with $100 billion to be deployed immediately. (openai.com) OpenAI also said Stargate would not replace Microsoft. The company said it would “continue to increase” its consumption of Azure, while Microsoft said key terms including revenue sharing and access to OpenAI intellectual property remain in place through 2030. (openai.com) (blogs.microsoft.com) That means OpenAI is trying to expand across several fronts at once: consumer subscriptions, enterprise software, application programming interfaces, Microsoft distribution, and a national-scale data-center buildout. Zitron’s criticism is that each layer adds obligations before the underlying economics are settled. (wheresyoured.at) (youtube.com) The company has also raised money on a historic scale. OpenAI said on March 31, 2025, that it closed a $40 billion funding round at a $300 billion post-money valuation, with SoftBank leading the financing. (techcrunch.com) (cnbc.com) Then came more infrastructure. On July 22, 2025, OpenAI said Oracle and OpenAI had agreed to develop 4.5 additional gigawatts of Stargate data-center capacity in the United States, bringing Stargate capacity under development to more than 5 gigawatts and more than 2 million chips. (openai.com) A separate pressure point is whether revenue can keep pace with those commitments. Ed Zitron cited reporting in April 2026 saying Friar had told colleagues OpenAI was not ready for an initial public offering in 2026 and was unsure revenue growth would support spending commitments. (wheresyoured.at) (economictimes.indiatimes.com) OpenAI’s public position is still expansion, not retreat. Its enterprise report says business revenue helps fund broad access to artificial intelligence, and its Stargate announcements frame the buildout as a long-term platform for future models and products. (cdn.openai.com) (openai.com) So Zitron’s “fragility” thesis is less about a sudden collapse than about a company that must keep selling, raising, and building at the same time. The more OpenAI grows, the more power, chips, capital, and partner cooperation it appears to need. (wheresyoured.at) (openai.com)