ANALYSIS: Qualcomm and Nvidia in China — The Semiconductor War Pushes the Boundaries of the Rules

The GPU is blocked, the CPU is open—for now

Nvidia’s approach is different but parallel to Qualcomm’s. Its most advanced AI GPUs remain under strict control. Sales of the H200 to ten Chinese companies authorized by Washington were blocked for months by Chinese authorities, who were promoting their own domestic manufacturers. But the Vera CPU—an 88-core ARM processor designed for agentic AI workloads—falls into a different regulatory category. Reuters reported in June that Nvidia was offering Vera to Chinese customers with a delivery target of August 2026, with approximately 300 pilot units ordered to test software compatibility before larger volumes.

Nvidia CEO Jensen Huang stated that his forecast of a $200 billion CPU market included China. In March 2026, Nvidia resumed production of the H200, which complies with export regulations, under a new mechanism: licensed sales, with 25% of the revenue remitted to the U.S. government. Members of Congress criticized this structure, arguing that it created a perverse incentive to approve more sales at the expense of national security—in a letter addressed to Howard Lutnick in January 2026, they pointed out that the H200 chips were six times more powerful than previously authorized chips.

The Issue of Reclassification

The central question surrounding Vera remains unresolved: Is it a general-purpose CPU or an AI accelerator that warrants stricter regulatory classification? If the Bureau of Industry and Security were to reclassify the category, Nvidia’s window of opportunity in China would close before the first shipments even materialize. This fundamental uncertainty is what makes the strategy risky, even for a company as well-capitalized as Nvidia. The precedent set by the H200 chips—approved and then blocked within a matter of months—is a warning that the industry has taken to heart.

China’s data center ecosystem is built on x86 architecture, and migrating to ARM entails significant software costs. Compatibility with existing AI software—trained and optimized for Nvidia’s CUDA GPUs—is not guaranteed on an ARM CPU. This is not a sure thing: it’s as much a technological gamble as it is a regulatory one, in a market where Huawei and Cambricon are making rapid progress on their own architectures.

Nvidia is selling a CPU in China because its GPUs are blocked. Qualcomm is selling ASICs below the threshold because its full-performance products cannot enter the market. These are not strategies for market domination—they are strategies for commercial survival in a market that cannot be abandoned but can no longer be served normally. The semiconductor war is producing some strange twists and turns.

Huawei, Cambricon, and the National Technology Wall

China’s goal of reducing its dependence on foreign AI technologies to 20%—backed by $295 billion in funding over five years—is not just wishful thinking. It is a direct result of the lesson learned from U.S. restrictions on chips: any technological dependence is a strategic vulnerability. Beijing’s response is systemic. Huawei Ascend, Cambricon, and the accelerators developed by Chinese internet giants such as Alibaba (Hanguang) and Baidu (Kunlun) benefit from guaranteed government contracts, tariff protections, and a rapidly developing software ecosystem under direct government oversight.

As early as late 2025, Beijing had banned the use of foreign AI chips in state-funded data centers. For Qualcomm and Nvidia, the segment still open to them is essentially the private sector—ByteDance, Alibaba Cloud, Tencent—where computing needs may still favor U.S. chips that are cheaper or more powerful than current domestic alternatives. But this window is narrowing. According to analysts, Huawei Ascend and Cambricon are expected to reach significantly higher production volumes by 2028, automatically reducing the market share available to foreign suppliers.

Howard Lutnick and the Next Wave of Restrictions

U.S. Secretary of Commerce Howard Lutnick stated in mid-June 2026 that “the era of the robotic arms race is coming,” signaling the imminence of new restrictions on robots manufactured in China. This statement is part of a trend toward a gradual tightening of the regulatory scope—robots, drones, chips, software—which makes U.S. companies’ compliance strategies structurally fragile. By June 2026, the Trump administration had already tightened the rules, requiring licenses for Chinese companies headquartered in China even when they operate abroad—a significant expansion of the regulatory scope.

Political pressure in the United States is structurally geared toward tightening regulations. In January 2026, members of Congress sent a letter denouncing the easing of controls on the H200 as a threat to national security. This bipartisan pressure creates an unpredictable regulatory environment for companies planning three to five years ahead. The instability of the rules is, in itself, a form of strategic risk—and it constitutes a comparative advantage for China, whose industrial policy on semiconductors is consistent, well-funded, and not subject to electoral cycles.

China is investing 295 billion in a coherent five-year strategy. The West changes its export rules every six to twelve months under pressure from political cycles. In a long-term technology war, strategic coherence is just as valuable as technological superiority. In this regard, the West’s deficit is real and rarely acknowledged.

Why Cutting the Cord Isn’t a Short-Term Option

This paradox encapsulates the entire Sino-American technological relationship in 2026: American companies need the Chinese market to fund the R&D that will allow them to stay technologically ahead, while China still needs American chips and tools to bridge its current technological gap. 46% of Qualcomm’s revenue comes from China. Abruptly cutting off this flow would undermine the company’s ability to invest at the very moment when global technological competition is intensifying. It is a dependency that slows decoupling without stopping it—and creates a commercial gray area where geopolitics and financial statements are in constant conflict.

On the Chinese side, domestic accelerators still lag significantly behind the best American chips in performance, especially when it comes to training workloads for large language models. This window of dependence is precisely what Qualcomm and Nvidia are trying to monetize before it closes—what the industry calls the “window of mutual dependence.” It’s a rational bet in the short term. But its lifespan is limited—commercially by the rise of Chinese national champions, and regulatory-wise by growing pressure from Washington.

The Structural Irony of the Technology War

There is a fundamental irony in this race: through their sales in China, American companies are indirectly funding a technological ecosystem designed to replace them. The revenue generated by ByteDance and other Chinese customers helps finance—through taxes and learning effects—the rise of the Chinese semiconductor industry. But turning down this market means ceding ground to international competitors who will find ways to enter it. There is no right option—only different trade-offs between national security and commercial competitiveness.

For the West as a whole, the central question is not whether Qualcomm should sell chips in China. It is deciding what level of technology transfer is acceptable with a declared strategic adversary, and establishing rules that are clear and stable enough for companies to plan five years ahead. The current regulatory instability—easing followed by tightening, licenses granted and then blocked—harms long-term U.S. interests just as much as it claims to protect them in the short term.

There is no simple answer here. To sell or not to sell—both options carry real and measurable geopolitical costs. What I find dangerous is the notion that simply adjusting computational performance thresholds is enough to resolve a fundamental strategic contradiction. You don’t win a generational technology war with regulatory decimal points.

Qualcomm Sets an Industry Precedent

The Qualcomm-ByteDance agreement of May 2026 is not merely a commercial contract. It is a model—a proof of concept that the entire U.S. semiconductor industry is watching closely. By showing that it is possible to secure significant data center revenue in China without violating export controls, Qualcomm has set an industry precedent. Other U.S. chipmakers, particularly those specializing in ASICs for AI inference, are now actively exploring similar contractual structures. Compliance by design is becoming a differentiating business competency, not just a legal obligation.

This precedent, however, has clear limitations. The ByteDance agreement covers inference chips—that is, chips that run pre-trained AI models, not those that create them. Training chips, which are far more computationally powerful, remain outside the scope of the current compliance model. This distinction is fundamental: China can purchase chips to deploy AI, but remains theoretically limited in its ability to train new state-of-the-art models using U.S. chips. Theoretically—because alternative supply routes remain well-documented and cause for concern.

The ByteDance-Qualcomm model looks elegant on paper: compliant inference chips, not training chips. In practice, a sufficiently distilled AI model can be retrained and refined on high-performance inference chips. The line between inference and light training is less clear-cut than a regulatory threshold would suggest.

From Chips to Robots: The Same Logic of Interdependence

Howard Lutnick’s statement on “the era of the robotic arms race” signals an expansion of the regulatory scope far beyond semiconductors. Robots manufactured in China—whose components often include sensors, embedded processors, and control algorithms with potential dual-use applications—are now in Washington’s crosshairs. This logical extension of the export control paradigm addresses a real concern: a country that masters advanced industrial robots and autonomous drones holds a potential military advantage that restrictions on chips alone cannot counterbalance.

For U.S. robotics and automation companies, this prospect creates a new area of uncertainty similar to that already faced by semiconductor manufacturers. Global supply chains for industrial robotics are deeply integrated with Chinese components and subsystems—a reality that the new restrictions will have to navigate without causing major economic disruption to U.S. manufacturing industries themselves. The technology war generates its own collateral costs for the U.S. economy, a reality that policymakers cannot ignore indefinitely.

What This Means for Qualcomm and Nvidia

For Qualcomm and Nvidia, the expansion of controls to include robots is not an immediate threat—their products are chips, not robots. But it sends a clear signal of policy direction: the trend in U.S. regulation is toward systematic tightening, category by category, until the architecture of technological dependence between the United States and China is fundamentally restructured. Each new category subject to restrictions narrows the scope of what U.S. companies can sell in China without regulatory friction.

This trajectory reinforces the case for long-term decoupling, which—even if it takes ten years rather than one—calls into question all revenue models dependent on access to the Chinese market. The $15 billion in data center revenue projected by Qualcomm by 2029—a significant portion of which is expected to come from China—is contingent on regulatory decisions over which neither the company nor its shareholders have control. This is the fundamental systemic risk of a business strategy built on the regulatory tolerance of a government that is sending clear signals of a gradual tightening of policies.

Robotics controls are the next chapter in a story whose arc we already know: restriction by restriction, category by category, Washington and Beijing are building two technological ecosystems that are becoming less and less compatible. Companies that have bet on continued mutual access will one day have to revise their business models. Not if, but when.

Manufacturing in Compliance via Taiwan

Qualcomm’s entire strategy rests on one element that years of discussion about computational thresholds tend to obscure: the chips have to be manufactured somewhere. The vast majority of ASICs tailored for ByteDance will pass through TSMC’s foundries in Taiwan—a geopolitically explosive reality in and of itself. TSMC is subject to U.S. export controls under current technology control agreements and must verify that the chips it manufactures for U.S. customers intended for Chinese buyers comply with regulatory thresholds. This is an additional layer of compliance that adds legal robustness to the structure—but also creates further dependence on a player whose own geopolitical security is uncertain in the face of pressure from Beijing.

China has no access to TSMC for the most advanced technology nodes—a direct consequence of U.S. and Dutch controls on EUV lithography equipment. This is one of the few restrictions that truly holds: China cannot manufacture 3- or 4-nm chips domestically today, regardless of how much capital is invested. This manufacturing bottleneck is the real Achilles’ heel of China’s AI ambitions—and that is why the 295 billion yuan includes a massive investment in alternatives to ASML equipment and in upgrading the capabilities of domestic foundries like SMIC. The race isn’t just about chips. It’s about the machines that make the chips.

The real control point in the semiconductor war isn’t U.S. law. It’s ASML in Eindhoven and TSMC in Taiwan. Two non-U.S. companies that hold the keys to the world’s most important technological lock. And they operate in a geopolitical neighborhood that Washington does not fully control. That is where the decisive battle will be fought.

A Precarious but Clear-Sighted Balance

The strategy that Qualcomm and Nvidia are deploying in 2026 is technically legal, commercially rational, and geopolitically risky. It relies on a regulatory nuance that could be called into question by a memo from the Bureau of Industry and Security, a bipartisan decision by Congress, or a unilateral restriction by Beijing on imports of foreign chips into critical infrastructure. There is room to maneuver—but it is narrow, contested, and temporary by nature. This is not a strategy for stable growth. It is a positioning strategy during a period of technological and geopolitical transition.

What is structurally clear in July 2026 is that the semiconductor war is no longer being fought solely in R&D labs or in Washington’s rhetoric. It is being fought in the technical specifications of chips designed to fall just below the legal threshold, in the contracts signed between ByteDance and Qualcomm, and in the decisions of Chinese customers to choose between a compliant American accelerator and a sovereign Huawei accelerator. The outcome of this war will be decided over ten years, not ten months—and no regulatory loophole strategy can replace a long-term industrial vision.

What This Means for the West

The lesson for the West is twofold. First, strategic consistency is just as important as technological superiority in a long-term competition. China has a five-year plan funded to the tune of 295 billion. The West holds elections every four years and has export rules that change with every administration. Second, companies cannot be the sole bearers of the burden of industrial policy and national security. Without a clear and stable collective vision, they optimize their short-term profits—which is their legal mandate—while long-term strategic competition remains without coherent guidance.

The semiconductor war will continue. Qualcomm will ship its first compliant chips in 2027, if the rules do not change. Nvidia will attempt to ship its first Vera CPUs in August 2026, if Beijing does not block imports. And in the meantime, China will press ahead with its 295 billion with a consistency that Western democracies struggle to match. The regulatory window exists. But it is not a strategy. It is a precious reprieve—provided we use it to build something more sustainable.

I don’t know if Qualcomm will succeed in its Chinese gamble. What I do know is that the West cannot win a generational technology war by playing within the limits of its own rules while its adversary rewrites new ones at a steady pace. Regulatory control is not an industrial policy. It is a delaying tactic—useful, but insufficient.

By Maxime Marquette, columnist