Nvidia Seals Multibillion Chip Deal with Murati’s Thinking Machines as U.S. Tightens AI

Nvidia’s new contract with Thinking Machines, the startup founded by former OpenAI chief technology officer Mira Murata, represents one of the few large‑scale chip deals that have been confirmed since the U.S. Commerce Department began tightening export controls on advanced AI hardware. According to the Financial Times, the agreement is valued in the “multibillion‑dollar” range and will give Thinking Machines access to Nvidia’s latest H100 GPUs, the same silicon that powers OpenAI’s ChatGPT and Microsoft’s Azure AI services. The deal was sealed just weeks after the company announced a series of licensing restrictions aimed at curbing the flow of high‑end GPUs to “high‑risk” end‑users abroad, a move that analysts at CNBC have described as “the most aggressive export curtailment in decades” (CNBC, Market Analysis).

The partnership is strategically significant for both parties. For Nvidia, securing a committed customer for its flagship AI accelerator helps offset the revenue hit expected from the new export regime, which the Silicon Shield report warns could shave up to 10 percent off the company’s projected 2024 AI‑chip sales if the restrictions are fully enforced (The Silicon Shield, cedarkeybeacon.com). For Thinking Machines, the contract resolves a supply bottleneck that had left the startup “scrambling for GPU access” in its early product development phase, according to the Financial Times. Murata’s team plans to use the H100s to train large‑scale foundation models for specialized enterprise workloads, a market segment that Nvidia has been courting aggressively after its consumer‑grade GeForce line proved insufficient for the compute demands of modern generative AI.

Regulators are watching the deal closely because it tests the limits of the new U.S. policy framework. The Commerce Department’s recent “Entity List” expansions target firms that are deemed to pose national‑security risks, and the agency has signaled that future licensing decisions will hinge on end‑use transparency and geographic destination (The Silicon Shield). Nvidia has reportedly filed a detailed compliance package with the Department of Commerce, arguing that Thinking Machines’ operations are confined to U.S. data centers and that the H100s will not be re‑exported to sanctioned jurisdictions. If approved, the contract could become a benchmark case for how U.S. AI chip exporters navigate the emerging regulatory landscape while still delivering the compute power that fuels the industry’s growth.

Industry observers note that the agreement underscores a broader shift toward vertically integrated AI ecosystems. By locking in a long‑term supply of Nvidia’s most advanced GPUs, Thinking Machines can focus on building proprietary software stacks without the uncertainty of spot‑market pricing or sudden supply chain disruptions. At the same time, Nvidia’s willingness to engage with a nascent player—rather than only the entrenched cloud giants—signals confidence that its hardware roadmap will remain indispensable even as the U.S. tightens export rules. As CNBC’s market analysis points out, “the real test will be whether other AI startups can secure similar deals under the new regime, or whether Nvidia will become the de‑facto gatekeeper for U.S. AI compute” (CNBC).

The deal also has implications for the competitive dynamics with China’s AI chipmakers, who have been accelerating their own H100‑class offerings in response to the export clampdown. The Silicon Shield report argues that the U.S. strategy aims to “choke off Nvidia’s global AI chip exports” to preserve a technological edge, but critics warn that such restrictions could drive foreign customers to domestically produced alternatives, eroding Nvidia’s market share abroad. By cementing a high‑profile domestic partnership, Nvidia may be hedging against that risk, ensuring a steady revenue stream from U.S.‑based AI ventures while the broader export environment remains in flux. The Thinking Machines contract, therefore, is not just a commercial transaction; it is a litmus test for how the U.S. AI ecosystem will adapt to a new era of heightened regulatory scrutiny and geopolitical competition.