Is OpenAI putting the 'AI' in too big to fail? (www.morningbrew.com)

OpenAI’s CFO Sarah Friar ignited a debate by saying the company wants a government “backstop” on loans to help finance roughly $1.4 trillion in planned spending over the next eight years for data centers and AI infrastructure. CEO Sam Altman quickly denied OpenAI seeks to be “too big to fail,” and White House officials said taxpayers won’t underwrite bailouts — but investors still fled, wiping more than $820 billion off AI stocks in a single week. The controversy is compounded by OpenAI’s lack of sustained profitability, provocative product choices that have drawn criticism, and an absence of a near-term IPO or clear funding plan. The alarm goes beyond size to systemic risk: OpenAI’s buildout depends on billions in circular, interlocking investments and supply relationships with chipmakers and cloud partners (Nvidia, AMD, Microsoft, Oracle, CoreWeave), concentrated GPU supply, massive power and colocation needs, and tight global competition with China. That tight coupling raises contagion and moral-hazard concerns if public support is contemplated. Proponents including Nvidia’s Jensen Huang argue subsidies or government support for data centers are strategic; opponents worry about taxpayer exposure and market distortions. For the AI/ML community, the episode spotlights the industrial-scale capital, specialized hardware bottlenecks, and policy questions that will shape who governs, funds, and ultimately benefits from large-scale AI deployment.