Nvidia Bets $6 Billion on Open-Weight AI to Counter China
Open-weight models let companies download, deploy and customize artificial intelligence systems, offering more control over data, costs and computing infrastructure than proprietary services from OpenAI and Anthropic. The rapid rise of Chinese developers such as DeepSeek has increased pressure on U.S. technology companies. For Nvidia, building competitive models would extend its influence beyond chips and reinforce the software ecosystem that drives demand for its AI hardware.
Nvidia reached a $6 billion licensing agreement with AI startup Poolside in August 2026 and recruited a large group of its engineers to accelerate the Nemotron open-weight model project. The investment, equivalent to about NT$191.1 billion, signals Chief Executive Jensen Huang’s push to build an open AI ecosystem. The initiative is aimed at countering Chinese open-source rivals while also challenging the closed-model strategies favored by OpenAI and Anthropic.
All Coverage
2 original reportsThe Backstory
The history behind this eventNvidia’s Huang Defends Open-Weight AI in Washington
Open-weight models allow companies to download, customize and deploy AI systems, widening access while raising policy questions around security, intellectual property and technological sovereignty. The debate intensified after Beijing-based Moonshot AI released Kimi K3 on July 16, challenging assumptions about the U.S. lead over China. Washington is now weighing whether restrictions on powerful downloadable models would curb security risks or instead slow adoption and weaken American competitiveness.
Nvidia Chief Executive Jensen Huang visited Washington on July 28, meeting Senator Mark Warner, the top Democrat on the Senate Intelligence Committee, other lawmakers from both parties and Commerce Secretary Howard Lutnick. Huang argued that open weights are vital to safe AI adoption across industries and U.S. technology leadership. Nvidia also highlighted plans to produce $500 billion of American technology over four years, as the Trump administration prepares to unveil an AI policy framework by Aug. 1.
Subscribe to Mark Radar Weekly
Every Friday, the week's strongest signals in your inbox. Unsubscribe anytime.
If you search news on Google, you can set Mark Radar as a preferred source—our coverage will show up more often in your results. Set as preferred source on Google →