Nvidia and CEO Jensen Huang have invested heavily to build an American open-source AI champion, but money alone has proven insufficient. Last August, Reflection AI co-founder Misha Laskin met Huang, who pivoted the startup from an AI coding tool to a new plan. Eleven months later, Reflection has yet to release its AI model and is falling behind rivals. Nvidia funds Reflection with $800 million to start, then 40 percent of its $2 billion round in October, and more this year at a $25 billion valuation. Nvidia also convenes the Nemotron Coalition of eight labs, lobbies with 22 other organizations backing open weight models on July 24, and allies with the Open Secure AI Alliance. The four leading open models globally are Chinese; the two American entries rank bottom, including Nvidia’s Nemotron 3 Ultra. Moonshot’s Kimi K3 shipped this month at 2.8 trillion parameters, the largest open weight model released, while Alibaba previewed Qwen3.8-Max at 2.4 trillion parameters. The largest American open model, Thinking Machines’ Inkling, is just under a trillion, and Nemotron 3 Ultra is 550 billion.