Nvidia Corporation, the world’s leading maker of artificial intelligence (AI) chips, faces mounting risks in its China business even as it delivers record-breaking quarterly results. In its latest regulatory filing, the company for the first time flagged potential harm from U.S. curbs on Chinese open-source AI models and vehicle technologies, underscoring the stakes of escalating geopolitical tensions. Yet on a conference call with analysts, CEO Jensen Huang lauded former President Donald Trump’s decision to rescind a restrictive export rule, signaling confidence that U.S. policy shifts could ultimately benefit Nvidia’s global AI leadership.

Export Curbs Threaten Key Revenue Streams

Nvidia’s filing discloses that U.S. restrictions on the use of Chinese AI models—such as DeepSeek’s R1 series and Alibaba’s Qwen—risk curtailing the demand for Nvidia GPUs from developers who rely on these platforms. DeepSeek, which recently rolled out its upgraded R1-0528 reasoning model, and Qwen are among the fastest-growing open-source AI frameworks in China, collectively running on tens of thousands of Nvidia chips. Should new rules bar these models from deployment on U.S.-made hardware, Nvidia could see a further erosion of sales in one of its most dynamic markets.

Equally consequential are U.S. bans on connected-vehicle technologies of Chinese origin. After years of underperformance, Nvidia’s automotive division began to gain traction in China last year, leveraging its DRIVE platform for advanced driver-assistance systems. Yet the imposition of export controls on Chinese vehicle chips threatens to chop off this nascent growth area. By barring sales of critical vehicle-computing components, regulators could stall partnerships between Nvidia and local automakers, reversing progress that had lifted the company’s car business from the doldrums.

The practical toll of these curbs is visible in Nvidia’s financials. The company attributed $2.5 billion in lost H20 chip sales to sudden export bans and took a $4.5 billion inventory charge for H20 units it could not ship to Chinese data centers. Looking ahead, Nvidia projects an additional $8 billion hit in its current quarter, driven largely by a freeze on customized Hopper-class chips tailored for China. Although its data-center revenue surged 73% year-over-year to reach $39.1 billion, these sequential setbacks spotlight how policy shifts can ripple through Nvidia’s top line.

Blackwell Variant Offers a Possible Workaround

In response to restricted access to its H20 product line, Nvidia is reportedly developing a “Blackwell China Variant,” a reconfigured version of its next-generation Blackwell GPU that complies with U.S. export rules while retaining high performance. This strategy mirrors earlier workarounds in which Nvidia shipped generalized GPU models to China before enabling advanced features through software unlocks—although new rules may close that loophole. The success of any variant will depend on the fine print of future regulations and whether the Commerce Department imposes broader definitions that capture even innocuous hardware differences.

While praising Trump’s move to rescind the Biden-era “AI Diffusion Rule,” which would have tightened global chip exports, Huang warned that no clear replacement guidance has emerged. “The president has a plan. He has a vision. And I trust him,” Huang said, referring to Trump’s promise to craft export controls that strengthen U.S. platforms without shutting out partners. Yet he acknowledged that any successor rule “may impose new restrictions on our products or operations,” leaving Nvidia in limbo as it prepares for multiple regulatory scenarios. The absence of definitive policy language complicates Nvidia’s supply-chain planning, product rollouts and customer engagements.

Huang’s unusual public commendation of Trump stems from his belief that the former president’s approach—anchored in clear handles on AI export policy and robust trade deals—provides a more stable foundation for long-term planning. On the earnings call, Huang highlighted Trump’s deal-heavy Middle East tour, noting that infrastructure agreements, job creation and export-focused diplomacy align with Nvidia’s global ambitions. He contrasted this with the Biden administration’s more bureaucratic rulemaking process, which he said introduced greater uncertainty. “America wins when \[AI] models run best on American infrastructure,” Huang remarked, arguing that open collaboration with top Chinese developers offers insights into industry trends and fuels U.S. innovation.

Balancing Alliances and Adversaries

Despite policy risks, Nvidia remains bullish on non-China growth opportunities. Executives touted pending multi-billion-dollar deals in Saudi Arabia, the United Arab Emirates, and Taiwan, where governments are investing heavily in AI-enabled infrastructure. Saudi Arabia’s sovereign wealth fund plans to build an AI data center park powered by thousands of Nvidia GPUs, while Taiwanese cloud-service operators are preordering Blackwell-series accelerators. These partnerships, coupled with a robust U.S. federal procurement pipeline for AI research in quantum computing and defense applications, could offset some China shortfall—provided export licenses are granted.

Analysts warn that as U.S.–China technology competition intensifies, Nvidia must bolster its supply-chain resilience. The company sources advanced packaging equipment and memory modules from firms in Japan, South Korea, and Europe, but any extension of export controls to assembly tools or cooling systems could hobble production. To mitigate this, Nvidia is exploring dual-sourcing arrangements, stockpiling critical components and collaborating with foundries on new packaging methods less reliant on U.S. technology. Should export controls broaden, Nvidia’s ability to deliver on its roadmap for Blackwell, Grace CPU accelerators and future AI platforms could come under strain.

Domestic Development vs. Foreign Restrictions

In China, the government is doubling down on homegrown AI chip programs, intensifying efforts to reduce reliance on U.S. suppliers. State-backed firms are investing heavily in advanced logic design and chiplet integration, while open-source AI communities rally around indigenous hardware platforms. If Chinese developers pivot to domestic GPUs and accelerator architectures, Nvidia risks losing not only near-term sales but also influence over AI standards and toolchains. Huang warns that sidelining American platforms could drive China’s AI talent into the arms of rival ecosystems, possibly creating bifurcated global AI stacks.

As Nvidia charts its course through this geopolitical maze, the company’s dual thrust—advocating for pragmatic U.S. export policies while shoring up non-China growth—reflects a nuanced strategy. By publicly praising Trump’s decisive but flexible approach to AI rules, Jensen Huang signals a preference for clarity and speed in policy formulation. Yet Nvidia must prepare for scenarios where restrictions tighten, compelling further product redesigns, creative licensing workarounds, and accelerated diversification into new markets. How well Nvidia adapts will shape not only its own fortunes but also the broader landscape of global AI competition—determining whether U.S. platforms remain at the forefront of artificial intelligence innovation or cede ground to emerging rivals.

(Adapted from MarketScreener.com)