Nvidia Vera Rubin processor enters production, accelerating next-generation CPU and GPU design cycles via EDA optimization.
Jensen Huang, the CEO who built a $3 trillion chip empire, made his debut on social media platform X on Friday with an open letter rather than a product announcement. Titled "Open Weights and American AI Leadership," the letter was signed by more than 20 companies including Meta, Microsoft, Palantir, IBM, Mistral AI, and the Linux Foundation. It argues that open-weight AI models strengthen security and accelerate innovation. Notably absent from the signatories: OpenAI and Anthropic, two companies that have advocated for tighter restrictions on open-source AI.
The timing proved consequential. Huang's post arrived just days after US Treasury Secretary Scott Bessent warned that Chinese AI firms could face sanctions for "industrial distillation attacks" on American patents, arguing that open-source AI should not become an "open hunting season" on US intellectual property. Huang publicly pushed back, framing distillation—where one model learns from another—as a normal part of AI development rather than a threat. His position aligned more closely with Beijing's perspective than with the harder line taken by OpenAI and Anthropic, creating tension with Washington's emerging stance.
The policy disagreement carries weight for Nvidia. Regulations restricting open models directly affect computing power demand and chip sales. The regulatory backdrop includes White House consideration of whether to restrict Chinese models like Moonshot AI's Kimi K3, which launched July 16 and ranks among the world's most powerful. White House advisor Michael Kratsios accused Moonshot of copying a US model through distillation.
Huang also addressed industry concerns about Nvidia's next-generation architecture. Research firms had warned of potential delays with the Vera Rubin platform, particularly around the Kyber rack solution critical to the Rubin Ultra rollout, with worst-case scenarios suggesting a full launch delay to 2028. Huang directly contradicted those claims at a developer conference, stating that Vera Rubin is "already in production" and promising "massive production volumes." This reinforces Nvidia's roadmap of annual chip releases, with Rubin eventually replacing Blackwell, which currently drives the majority of data center revenue.
Separately, Nvidia is deepening its push into government AI infrastructure. Alongside Naver and Brookfield, the company is expanding South Korea's national AI factory. Brookfield is contributing up to $9 billion in capital while Nvidia is investing $1 billion. The GAK Sejong data center will grow from 55 megawatts to 200 megawatts of capacity by 2028, powered by Nvidia's DSX platform, designed to deliver computing power for government AI models and autonomous AI agents.
Nvidia shares closed Friday at €182.00, down 0.80% on the day but up 13.55% year-to-date, sitting roughly 10% below their 52-week high of €202.50 from May 14. The coming week will test the stock's trajectory as three of Nvidia's largest customers—Microsoft, Meta, and Amazon—report quarterly earnings. For Nvidia investors, the single most important metric will be these companies' capital expenditure on AI infrastructure. Alphabet has already sent positive signals in its latest update. Should the hyperscalers confirm their spending plans, it would reinforce the demand story for both Blackwell and the incoming Rubin generation. Until Nvidia reports its own results in late August, the stock's direction will largely depend on how customers and competitors communicate their AI investment plans.