U.S. AI Startups Rush to Develop Low-Cost AI Alternatives for China, But Still Face Funding Resistance

August 2nd, according to The Wall Street Journal report, as Chinese open-weight models such as Kimi, Qwen, and DeepSeek approach top U.S. models at a lower cost, Silicon Valley and Washington are increasingly concerned that Chinese models may long-term compress U.S. AI companies' profits. U.S. startups like Arcee AI, Reflection AI, and Poolside are actively developing local alternatives to meet the demand for low-cost, downloadable, and customizable models.

However, U.S. open-weight model companies are facing funding difficulties. Some investors question whether free open models can generate stable revenue and are also concerned that the related technology may weaken their investment value in OpenAI and Anthropic. In the first quarter of 2026, AI startups raised a total of $255.5 billion, with nearly two-thirds coming from three funding rounds for OpenAI, Anthropic, and xAI.

Arcee AI, with limited funding, used 2048 Nvidia Blackwell B300 chips to complete 33 days of pre-training and launch Trinity Large with a budget of approximately $20 million. The model is still inferior to top models and lags behind OpenAI and Anthropic in multiple benchmark tests, but the company plans to develop larger models through a new round of funding.

Nvidia has become a key supporter of the U.S. open AI ecosystem, not only developing the Nemotron series models but also investing in Reflection AI, Poolside, and Thinking Machines Lab. Industry insiders state that the U.S. open-weight ecosystem is still relatively small, with Chinese models maintaining an overall advantage.