Scott Wilson, chief investment officer of Washington University's endowment, told a room full of asset managers and venture capitalists that OpenAI and Anthropic are in major trouble, arguing that the companies have committed to enormous spending even as cheaper alternatives multiply. Wilson said he doubts OpenAI will ever turn a profit, predicting there will be many free alternatives that undercut the closed frontier labs. His remarks came amid debate over the companies' spending commitments.
An Unpopular Opinion at a Fort Worth Gathering
Wilson shared what he called his unpopular opinion at the Rock Yard Roundup, a Fort Worth gathering that drew roughly 100 asset managers, venture capitalists and technology founders, according to Business Insider. He argued that frontier companies are not worth the liabilities they have signed up for. Wilson had the opportunity to be an early investor in OpenAI himself but passed.
Wilson pointed to a broader shift he sees among portfolio companies heavily involved in AI, saying many are moving aggressively toward open-source alternatives instead of paying premium prices for closed models. That view lines up with market data showing Chinese-developed models are cheaper than U.S. frontier offerings. Chinese model makers including DeepSeek, Alibaba's Qwen, Zhipu AI and Tencent have narrowed the performance gap with U.S. labs while offering major cost savings, per the same Business Insider report.
Chinese Models Gain Ground, and Prices Keep Falling
The numbers back up at least part of the open-source narrative. More broadly, the share of businesses paying for model-serving platforms that offer open-source and Chinese-developed models climbed to 6.1% in July, up from 4.5% in January, according to Fortune. Open-weight models give companies more room to fine-tune and more control, and they generally cost less, the outlet reports.
The pricing pressure is not new. Stanford's 2026 AI Index separately estimated that the performance gap between the best U.S. and Chinese models had narrowed to just 2.7 percentage points, according to Truth on the Market.
The Economics of Closed Models
Wilson's view is a skeptical one: He argues that closed frontier labs face pressure from free and cheaper alternatives. His concern is whether those companies can justify the enormous spending they have committed to.
Part of the discussion around OpenAI's costs involves moves to control more of its own hardware stack. OpenAI and Broadcom unveiled a custom inference chip on June 24, with limited initial deployment expected late in 2026 and a ramp through 2027, according to HPCwire. The chip is intended to reduce OpenAI's reliance on Nvidia hardware.
Billions in Revenue, Questions About Returns
Anthropic is showing rapid growth in the near term. The company disclosed in May that its annualized revenue run rate had exceeded $47 billion, according to Schroders.
Wilson invested about $50 million in SpaceX almost a decade ago, according to Business Insider.
Investor concerns over whether these enormous AI investments will ultimately pay off have persisted even as analysts keep revising spending expectations upward throughout the year, Schroders notes. Anthropic held a 43.5% share of business AI-model spending in July, ahead of OpenAI's 39.7%, according to Fortune. OpenAI, Anthropic and Google DeepMind still control access to their frontier models through APIs, building applications, distribution deals and enterprise services around them — the exact structure Wilson suggested is now under threat from cheaper, open alternatives.
