? Exponential View #568: The labs are rationing. Did you notice?

In the latest installment of Exponential View, the focus is on the hidden reality of AI labs rationing their resources, a trend that has gone largely unnoticed. As the demand for AI services surges, the compute crunch has become a critical issue for major tech companies, leading them to turn away business and impose limits on users.

Last year, AWS faced a significant setback when it lost a $10 million contract to host Fortnite due to insufficient compute capacity. A similar situation occurred for Microsoft, highlighting the growing strain on these companies' resources. This week, OpenAI's CFO revealed that the company is turning down opportunities because of the lack of available compute power. The rapid growth of AI applications has been evident in the surge of developers using tools like Codex, which expanded from 100,000 to 2 million users in just three months.

To manage the increased demand, Anthropic has tightened session limits for some users, affecting around 7% of its user base. This move comes as H100 rental prices reach an 18-month high, further exacerbating the compute crunch. Meanwhile, Alibaba's decision to close-source Qwen, its open-weight leader, signals a shift in strategy amid the intense competition for resources.

The compute crunch is not the only challenge facing AI labs. The rapid expansion of AI applications has also led to unintended consequences, such as GitHub Copilot injecting promotional content into code reviews, which has upset the developer community.

As the demand for AI services continues to grow, the need for sustainable and scalable solutions becomes increasingly urgent. The upcoming data edition will delve deeper into this trend, exploring the implications of the compute crunch and the challenges faced by AI labs in meeting the growing demand.

In addition to the compute crunch, the article also touches on the decade-long feud between OpenAI and Anthropic, examining the history of the split and the competing visions of AI that emerged from it. Despite the rivalry, both companies are adapting to the changing landscape, with Kuaishou's Kling AI reporting $300 million annualized video revenue in 2023, with expectations of doubling that figure in 2026.

As AI continues to transform various industries, the challenges faced by labs in managing resources and balancing growth with sustainability will remain critical issues. The compute crunch serves as a stark reminder of the need for innovative solutions to ensure the continued advancement of AI technology.