Endowment chief warns OpenAI and Anthropic are spending into trouble

Scott Wilson, a top university endowment chief, is warning that OpenAI and Anthropic are headed for trouble. His argument: the frontier AI labs have taken on enormous spending commitments just as cheaper Chinese models close the performance gap, squeezing the economics of the industry's most prominent players.

The warning is a direct challenge to the prevailing narrative that the biggest AI labs are unstoppable. Wilson's point is that the combination of massive fixed costs and a rapidly improving low-cost alternative creates a classic margin squeeze, one that could force OpenAI and Anthropic to rethink their spending trajectories. For investors, this is not a distant concern; it is a live risk to the valuations that have been built on the assumption that frontier AI will remain a winner-take-most market.

For context, frontier AI models typically require vast amounts of computing power, data, and engineering talent. That translates into enormous capital outlays for training runs, data centers, and specialized chips. OpenAI and Anthropic have both made multi-year commitments to secure compute capacity, betting that their models will stay far enough ahead of the competition to justify the cost. Wilson's warning suggests that bet is looking shakier as Chinese labs deliver comparable performance at a fraction of the price.

The performance gap is closing faster than many in the industry expected. Cheaper Chinese models are not just catching up on benchmarks; they are also forcing enterprise buyers to ask a simple question: why pay a premium for a frontier model when a low-cost alternative gets the job done? That question is exactly what Wilson is pointing at. If the gap narrows further, the pricing power that OpenAI and Anthropic have enjoyed could erode, and their spending commitments would become a heavier burden.

This is a familiar pattern in technology. Incumbents with high fixed costs often dismiss cheaper challengers until the challengers are good enough. Then the margin compression happens quickly. For AI, the stakes are especially high because the spending commitments are so large and the revenue base is still relatively young. A slowdown in premium pricing could force labs to raise capital on worse terms, delay product roadmaps, or even consolidate.

For institutional investors, Wilson's warning carries extra weight because university endowments are major backers of venture funds that have poured money into AI. When a top endowment chief publicly flags trouble, it signals that even long-term, patient capital is starting to stress-test the AI narrative. That could ripple through private market valuations, especially for late-stage AI companies that have raised at eye-watering multiples.

For founders and operators building on top of AI, the implication is strategic. The moat is no longer just model quality; it is distribution, data access, and enterprise relationships. Companies that build their products exclusively on OpenAI or Anthropic may face pricing pressure or vendor lock-in as cheaper alternatives improve. Boards should ask their management teams how they would respond if the cost of AI inference dropped by an order of magnitude in the next 18 months.

Wilson's warning is not a prediction of collapse, but it is a clear-eyed reminder that the economics of frontier AI are not guaranteed. The labs have made enormous bets, and the competitive landscape is shifting under them. For anyone making decisions about AI budgets, vendor choices, or investment exposure, the message is simple: watch the gap, because it is closing faster than the spending commitments are being paid off.