China’s Chip Cheating
Unpacking a web of semiconductor smuggling, shell companies, and shameless subterfuge
By John Mac Ghlionn, researcher and cultural commentator
President Trump recently announced that chip tariffs would arrive “very soon.” He’s right to act. Trump deserves real credit for recognizing something that too many past presidents ignored: the fact that semiconductor supremacy isn’t just a matter of economics. It is, in many ways, a matter of survival.
For decades, politicians and elites in the American establishment have treated technology like a luxury good—something to buy, sell, trade, and outsource. They have failed to acknowledge a deeper truth: that microchips, the component that powers our technologies, are the lifeblood of America today. Without reliable access to the technology, we cannot lead in the future of technology, defend our country, and thrive as a nation.
But recognizing is only the first step. That clarity must now be matched with decisive action. Because, while America slept, its chief geopolitical nemesis, China, quietly recognized the importance of chips.
The United States and its allies—the leading producers of the most advanced semiconductor technology—have long agreed to limit the export of advanced chips and chipmaking tools to certain adversarial nations, especially those that might weaponize them. These restrictions are enforced through licensing rules, blacklists, and multilateral controls, such as those coordinated through the Wassenaar Arrangement, an international agreement that limits the export of dual-use technologies to authoritarian regimes. The goal is simple: prevent hostile nations from using U.S. and allied technology to harm us and our interests.
However, one regime has found a way to exploit the system, and it resides in Beijing.
A recent report from Rest of World, a nonprofit newsroom tracking technology’s impact beyond the West, reveals a sobering reality about the chip market: China isn’t just involved, despite U.S. efforts, it is pushing the technology forward at an alarming pace.
They do it not through explicit defiance, but through subtler, more dangerous tactics: smuggling, shell companies, cloud workarounds, and technological ingenuity. In other words, they aren’t breaking down doors; they’re slipping through the cracks. Smuggling operations are widespread and well-organized.
The evidence is inescapable. Supposedly locked out of Chinese markets, Nvidia’s high-end AI chips have been pouring into the country through Malaysia, Hong Kong, Singapore, and other regional hubs. These chips often reach Chinese firms indirectly, routed through opaque supply chains, re-labeled shipments, and middlemen who know exactly how to exploit regulatory blind spots.
China’s is a nationwide, all-hands-on-deck approach; even students returning from studying abroad carry GPUs stuffed into their suitcases. Business travelers disguise them inside consumer electronics shipments. Shipping records, customs declarations, and baggage inspections are manipulated or bypassed altogether.
Rest of World estimates that more than 12,500 restricted GPUs are smuggled into China annually—enough to equip hundreds of AI labs, defense contractors, and state-owned tech giants. That number is staggering when you consider that the most advanced GPUs are designed by U.S.-based firms like Nvidia and AMD, or manufactured by close allies such as Taiwan and South Korea, countries that are supposed to be fully aligned with American export control policy. In theory, these chips should be tightly tracked and difficult to reroute. In practice, the global semiconductor supply chain is so vast and fragmented that even the strictest controls to date have failed to prevent backdoor access.
In a world where access to a few thousand cutting-edge chips can tip the balance of AI breakthroughs, these smuggling routes directly threaten America’s technological lead.
An important element of this effort is the use of shell companies and corporate mutations. When the U.S. government blacklists Chinese companies, they often simply morph and mutate. Take Sugon, for example. Once a rising star in China’s supercomputing industry, it was blacklisted in 2019 for ties to the Chinese military. Within months, Sugon’s former executives founded a new firm—Nettrix—essentially a spin-off that continued to sell servers equipped with Nvidia and Intel chips, thereby bypassing restrictions on selling them to Chinese military research labs, surveillance firms, and state-backed universities—precisely the types of end users that the original ban was intended to block.
Inspur, China’s largest server manufacturer, employed the same strategy. After being sanctioned, it quietly spun off shell companies and rebranded divisions to keep buying and integrating restricted U.S. technologies. Without aggressive follow-up and enforcement, blacklisting has become a minor speed bump rather than a truly effective barrier to keeping U.S. technology out of the hands of a hostile foreign power.
Cloud services also provide a powerful back door. When hardware smuggling gets risky, Chinese firms often simply pivot to renting computing power abroad. This method reduces traceability, avoids export red flags, and allows them to evade physical controls. They buy access to American cloud infrastructure—Amazon Web Services, Google Cloud, Microsoft Azure—all powered by the very Nvidia chips they are prohibited from physically owning. By renting GPU time instead of importing hardware, Chinese companies circumvent export restrictions at low cost.
American corporations earn billions from these deals. Cloud providers like Amazon, Google, and Microsoft may not directly sell chips to Chinese military-linked firms, but by offering rental access to their GPU-powered infrastructure, they provide a convenient workaround—one that’s become impossible to ignore. At best, these companies are turning a blind eye. At worst, they’re tacitly enabling the violation of America’s trading rules by its chief geopolitical rival.
The scale is vast, with contracts worth hundreds of millions, potentially billions, routed through opaque intermediaries, shell firms, and “clean” third-party vendors. Chinese front companies have become increasingly sophisticated, often registering abroad or using local affiliates to mask their true affiliations. Some rent compute in bulk through resellers or smaller U.S.-based startups acting as brokers. Others rely on seemingly independent entities that serve as de facto proxies for PLA-linked institutions or AI research hubs. Given the scale of these cloud contracts and the sophistication of China’s front companies, it’s difficult to believe that the tech giants are entirely unaware. And yet, the revenue keeps flowing.
As a result, China continues to train its AI models. Global corporations pad their pockets. Everyone benefits—except American national security.
If the United States hopes to keep China from accessing our semiconductors, it will require serious changes to our existing, permeable system of export controls. Doing so is critical to preventing China from accessing cutting-edge computing power in the race for AI and strengthening their surveillance systems. Right now, they build that future on American servers, right under our nose.
Of particular danger where chips are concerned is Beijing’s AI ambitions. And the AI race between the United States and China is nail-bitingly close. According to credible reports, China is rapidly closing the AI gap, with state-backed companies developing military-grade models for surveillance, warfare simulation, and autonomous weapon systems. Given the potential AI holds to revolutionize so many aspects of life, falling behind to an autocratic regime is a risk too large to tolerate.
The United States hoped to slow China’s AI progress by restricting access to advanced GPUs. However, those controls mean little if Chinese firms can simply rent time on American infrastructure to train the same models. It’s ineffective to enforce hardware limits while offering unrestricted access to the software pipeline, which is just as important.
And China’s ambitions are constantly adapting to America’s attempts at denial. While American firms discard five-year-old GPUs as outdated relics, China sees them very differently. Constrained by sanctions, Chinese engineers have gotten remarkably good at extracting unexpected power from ‘obsolete’ chips to advance their tech.
As many readers likely know, a Chinese AI startup called DeepSeek recently shocked the tech world by releasing cutting-edge large language models. However, they might not realize that these models were trained not on Nvidia’s latest H100s but on earlier-generation hardware long dismissed by the West. They demonstrated that even ‘inferior’ chips can lead to significant technological breakthroughs with more innovative optimization techniques.
Announcing upcoming tariffs was the first necessary blow in the fight against Chinese chip cheating. It signals to both Beijing and American corporations that business as usual is over. Tariffs raise the cost of complicity and begin to realign incentives, penalizing not just direct trade but also the gray-market networks that enable China’s end-runs around export controls.
However, this war cannot be won with tariffs alone. China’s strategy is comprehensive; it combines smuggling, cyber theft, industrial espionage, trade manipulation, diplomatic coercion, and domestic innovation. The American response must acknowledge this ‘holistic’ approach and counter it appropriately.
A more comprehensive strategy must begin by getting tougher with those who enable China’s chip black market—namely, third-country smuggling hubs and complicit U.S. corporations. As mentioned, Malaysia, Hong Kong, and Singapore have become key transit points for restricted semiconductors. The United States can no longer afford to treat them as neutral intermediaries. Reciprocal tariff negotiations offer one clear path forward. If these governments want access to U.S. markets, they must prove they are not laundering high-end chips into China. That means customs transparency, active cooperation with U.S. intelligence, and accountability for domestic ports and firms that knowingly handle illegal shipments.
The same pressure must be applied at home, to American cloud providers like Amazon, Microsoft, Google. That could include mandatory KYC (Know Your Customer) protocols for GPU rentals, real-time audit trails of compute usage, and a licensing system for high-performance compute clients to verify end use and national affiliation. No Chinese shell company or proxy entity should ever be permitted to rent GPU time on U.S.-based infrastructure through intermediaries or resellers. If companies cannot or will not self-regulate, enforcement should be swift—and public. This isn’t about punishing profit. It’s about restoring national control over our most critical strategic technology.
Limiting only the most advanced chips while leaving older-generation GPUs unrestricted is a mistake. Even legacy chips can be clustered to power military-grade AI, and the Chinese military is aware of this. Export restrictions should be broadened to cover all meaningful semiconductor hardware, not just cutting-edge tech, but anything with computational value that could enhance China’s surveillance or weapons programs. Meanwhile, the cloud loophole must be slammed shut.
Finally, no export control regime will fully address the chips issue unless the United States rebuilds its own semiconductor ecosystem from the ground up. That means full-spectrum industrial revival. That level of independence cannot be outsourced and will require serious congressional investment, long-term regulatory clarity, and cooperation between national security agencies and industry leaders. If the Trump administration wants to reshape America’s technological future, this is where it must begin: with a deliberate dismantling of the global system that allowed China to cheat and a bold commitment to build something more substantial in its place.