The head of one of the largest US endowments says OpenAI and Anthropic are in trouble
The head of a major endowment says frontier labs won't be able to withstand competition from cheaper Chinese open-weight models.
TL;DR
- Scott Wilson, CIO of Washington University's endowment, believes frontier AI labs face significant trouble.
- He argues that massive spending by companies like OpenAI and Anthropic is unsustainable, especially with cheaper Chinese AI models rapidly improving.
- Wilson's view contrasts with investors like Vinod Khosla, who believes infrastructure control is paramount.
- Chinese models from DeepSeek, Qwen, Zhipu AI, and Tencent are closing the performance gap with U.S. labs while offering cost savings.
- Wilson notes that companies invested in by his endowment are increasingly switching to cheaper, open-source AI models.
- Data from OpenRouter shows DeepSeek leading in text-model requests, surpassing OpenAI and Anthropic.
- Khosla disagrees, emphasizing the value of controlling the entire tech stack, from chips to inference models, for cost efficiency in closed-source systems.