Open-weight models now handle a majority of tokens on Vercel’s AI Gateway. But Anthropic still takes 64% of the spend.

The surge in open-weight model token usage on Vercel's AI Gateway, now handling a majority, directly challenges the dominance of proprietary models. This reflects a clear preference for flexibility and cost-efficiency in certain production environments. However, Anthropic's 64% share of spending shows that premium closed-source models still capture high-value workloads. Developers are optimizing for different factors.

For Asia, this trend means increased opportunity for startups building on or fine-tuning open-source models. Companies in markets like India or Southeast Asia can lower development costs and accelerate deployment. This could also pressure regional cloud providers to enhance their open-source AI offerings, competing more directly with global players. The test for Asian AI companies is whether they can translate token volume into significant revenue.

The thing to watch is how quickly Anthropic's spend share declines over the next 12 months. If the 64% figure drops significantly, it would indicate a more profound shift in enterprise AI budgets. This would force major Asian tech firms to rethink their investments in exclusive model partnerships versus open-source strategies.