A recent Sequoia essay by partners Dean Meyer and Konstantine Buhler argues the United States may be winning the race for closed frontier models while quietly losing the contest for the open-model ecosystem, and points to China’s open weights as the force reshaping that balance.
From 1% to 69%
Citing the ATOM report, the essay notes that Alibaba’s Qwen rose from about 1 per cent of new open-model fine-tuning and adaptation in January 2024 to 69 per cent by February 2026. The US still holds the lead in closed systems from OpenAI, Anthropic and Google, but its startup ecosystem has become deeply dependent on Chinese open models as a base layer it cannot easily route around.
The open-model paradox
The core claim is that opening a model can erode one firm’s commercial barrier while deciding who builds the larger ecosystem. The authors compare it to open-source software: Linux never became a single company’s product, yet it underpins cloud and internet infrastructure. In mobile, Google’s open Android out-covered Apple’s closed iPhone. Meta’s Llama strategy is the same bet at the model layer, favouring ecosystem reach over per-user value.
The real risk
Sequoia’s sharper warning is that the US advantage was never a handful of model companies. It was an open research system, a mature venture culture and a global developer network. If the most advanced US models are locked behind commercial walls while China’s open models keep attracting developers worldwide, the long-term centre of gravity in AI infrastructure could move. The paradox is that protecting a short-term lead might forfeit the open ecosystem that built the lead in the first place.
Read the original report (Sohu IT)
Translated and adapted from Sohu IT (it.sohu.com).