How the U.S. lost its chip war on China

When the White House quietly approved renewed exports of Nvidia’s H200 AI accelerators to China — with a 25% fee attached — it marked more than a policy adjustment. It marked the effective collapse of Washington’s semiconductor containment strategy. 

After years of escalating export controls, sanctions, and alliance pressure, the United States is now conceding what the chip war made clear: China cannot be technologically frozen, and U.S. monopoly control over advanced technology is no longer enforceable.

According to Bloomberg, the decision was driven by internal concern over Huawei’s accelerating progress in AI hardware and systems. The administration framed the move as a way to preserve U.S. “tech stack dominance,” but the reality is more revealing. 

Washington is retreating from an unwinnable attempt to enforce technological monopoly control, while trying to slow China’s exit from U.S.-controlled software ecosystems, especially Nvidia’s CUDA platform.

This was never about national security. It was always about preserving imperialist monopoly power.

From “Pivot to Asia” to technology war

The semiconductor war did not emerge overnight. It is the culmination of a decade-long imperialist containment strategy aimed at China’s rise. The Obama administration’s 2011 “Pivot to Asia” laid the groundwork, repositioning U.S. military forces around China while signaling the end of unconditional economic engagement. What began as military encirclement soon expanded into economic and technological warfare.

Under Trump, that shift became explicit. Tariffs were imposed. Chinese technology firms like Huawei and ZTE were sanctioned. Access to critical components was cut off. Under Biden, the same strategy was deepened and systematized. Export controls widened. Military alliances such as AUKUS and the U.S.–Japan–South Korea pact fused technological restrictions with strategic encirclement. The semiconductor industry became the central front.

Publicly, these measures were justified as defensive. In practice, they were aimed at preserving the monopoly foundations of the imperialist system. For decades, U.S. power rested on its ability to dominate new technologies long enough to extract enormous super-profits before competitors could catch up. China’s rapid movement into advanced manufacturing, telecommunications, renewable energy, and AI threatened to close that monopoly window.

The chip war was launched to keep it open.

Weaponizing the global supply chain

Washington’s strategy relied on turning the global semiconductor supply chain into a weapon. Because the U.S. historically dominated chip design, software, and key intellectual property, it believed it could enforce obedience far beyond its borders.

Export controls targeted advanced AI processors such as Nvidia’s H100 and H200. The Foreign Direct Product Rule asserted U.S. control over foreign-made products that rely on U.S. technology, forcing companies like TSMC and ASML to comply or lose access to critical tools and markets. Allies were pressured to abandon profitable Chinese markets. Scientific collaboration was restricted. Investment flows were blocked.

The globalized system of production — once celebrated as efficient — was transformed into a gated hierarchy with Washington as rule-maker and enforcer. But this strategy carried an internal contradiction. The same interdependence that gave the U.S. leverage also made it vulnerable to blowback.

That blowback came quickly.

China’s response: planning, scale, and technological sovereignty

China did not respond to the chip war with panic or retreat. It responded with long-term planning tied to production. The confrontation exposed a fundamental clash between two systems: monopoly-finance capitalism and state-led socialist development.

In the United States, AI became a speculative asset. Investment was driven by hype, stock prices, and Pentagon contracts. In China, AI was treated as infrastructure — something to be integrated into manufacturing, logistics, energy systems, and national planning.

Huawei’s progress illustrates this difference. Its Ascend 910C chips are not replicas of Nvidia’s best products, but they are increasingly competitive. The CloudMatrix 384 system compensates for efficiency gaps through scale and coordination, deploying 384 chips in tightly integrated clusters. Huawei has successfully substituted quality (Nvidia’s superior individual chips) with quantity (CloudMatrix’s massive chip clusters), resulting in a system performance that approaches Nvidia’s best in key workloads.

This is not engineering shaped by quarterly profit targets. It is capacity built through long-term state planning and coordination.

U.S. officials reportedly concluded that Huawei could produce millions of Ascend accelerators within a few years. That realization stripped export controls of their force. Instead of stopping China’s advance, containment sped the drive toward domestic production.

The second China shock

The consequences extend far beyond semiconductors. China’s advance represents a second China shock. The first, beginning in the 1990s, followed China’s integration into global supply chains, when U.S. and multinational capital reorganized production internationally — entire industries relocated, millions of jobs disappeared, and working-class communities in the United States were devastated. This second phase marks a break from that pattern. It is centered on China’s own advanced industrial and technological development, not on serving as a manufacturing platform for Western capital.

Chinese firms now lead or dominate in key sectors once assumed to be permanent strongholds of imperialist capital. Huawei in telecommunications. BYD in electric vehicles. CATL in batteries. DJI in commercial drones. Tongwei in solar manufacturing. They strike directly at the monopoly profits that sustained Western dominance.

According to data released on Dec. 1 by the Australian Strategic Policy Institute, Chinese institutions now lead research output in 66 of the 74 critical technologies it tracks — nearly 90% of the fields assessed. The United States leads in only eight. China’s research dominance spans areas central to modern industrial power, including nuclear energy, synthetic biology, small satellite systems, and cloud and edge computing.

This represents a reversal from the early 2000s. At that point, U.S. institutions led the overwhelming majority of advanced research fields, while China accounted for only a small fraction. Over the past two decades, that balance has flipped. China’s lead in cloud and edge computing, in particular, reflects the priority placed on deploying AI at scale — integrating research directly into production, logistics, and infrastructure rather than treating it as a standalone laboratory exercise.

This is why the chip war mattered. It was never just about semiconductors. It was about whether the United States and other imperialist powers could still decide who builds the most advanced technology, who gets access to it, and who does not.

The answer is no. The ability to plan, scale up production, and put technology to work now lies outside their control.

Blowback at home and abroad

The attempt to weaponize global production inflicted serious damage on the United States itself. Reshoring initiatives faltered. TSMC’s Arizona fab — marketed as a symbol of “tech sovereignty” — became a case study in dysfunction, plagued by delays, disputes with U.S. unions over staffing, training, and work practices, and soaring costs. Engineers ultimately had to be flown in from Taiwan to retrain U.S. workers on basic fabrication protocols, exposing the lack of a trained industrial workforce after decades of deindustrialization.

Allied governments and corporations were forced into impossible positions. Companies in South Korea, Japan, and Europe were compelled to sacrifice profits and market access in China for a strategy that primarily served U.S. geopolitical aims. Rather than consolidating control, Washington imposed real economic costs on allied states and corporations, forcing them to absorb losses in markets, supply chains, and investment.

Globally, countries began diversifying away from U.S.-controlled supply chains. The open weaponization of technology made clear that dependence on U.S. systems carried political risk. Claims of a “rules-based order” rang hollow when rules were rewritten at will.

Inside the U.S., the policy fueled a growing military-digital complex. Government money flowed to tech monopolies and defense contractors while social needs went unmet. Even Biden warned of the emergence of a new military-tech complex, in which Big Tech is fused with the armed and intelligence apparatus, concentrating technological and coercive power.

The chip war did not revive U.S. industry. It exposed its fragility.

Nvidia, CUDA, and a strategic retreat

This is the context in which the Nvidia decision must be understood. Allowing H200 exports to China is not a clever compromise. It is a retreat shaped by failure.

Washington is trying to keep Chinese AI firms tied to Nvidia’s CUDA software, slowing the shift toward domestic and open-source alternatives such as Huawei’s CANN. By blocking access to Nvidia’s newest Blackwell chips while allowing sales of the H200, the U.S. is holding back the most advanced hardware while maintaining dependence on U.S.-controlled systems.

Yet even this goal is tenuous. Chinese firms have already demonstrated the ability to train highly capable AI models with fewer resources. Systems such as DeepSeek matched the performance of leading U.S. models by emphasizing training efficiency and better use of available hardware rather than sheer computing scale, undercutting the assumption that restricting access to top-tier chips would halt progress. As domestic hardware continues to improve, software dependence will erode as well.

U.S. export controls were meant to slow China’s access to large-scale AI computing power long enough to give U.S. firms a decisive head start. But China continued advancing anyway — through domestic chip development, scale, and more efficient use of computing resources. As that gap narrowed, the value of strict controls became increasingly uncertain. Faced with the prospect that the restrictions might not stop China but would certainly cut off U.S. corporate profits, Washington chose to reopen exports. In doing so, it accepted a weaker, less durable form of influence in exchange for continued market access.

The end of monopoly enforcement

The chip war’s outcome is now clear. The United States did not fail because of a single mistake. It failed because the strategy itself was flawed. Imperialist monopoly capitalism cannot outplan a system organized for long-term development. Coercion cannot substitute for production. Sanctions cannot replace planning.

By attempting to freeze China’s development, Washington accelerated it. By weaponizing interdependence, it undermined its own position. By prioritizing monopoly profits, it weakened its industrial base.

The reopening of Nvidia exports is not a reset. It is an acknowledgment that the old model of technological domination no longer works. The era when the United States could dictate the terms of global technological development through choke points and monopolies is ending.

What comes next will not be decided by chips alone. It will be decided by which social system can organize production, labor, and technology to meet real needs over time. On that terrain, the chip war has already delivered its verdict.