Winning the AI Arms Race Against the Chinese Communist Party

The Stakes

Chairman Mast, Ranking Member Meeks, and other distinguished members of the committee, I’m privileged to have the opportunity to testify today. In the next handful of minutes, I’d like to dispel a few myths that have begun to mislead U.S. policymakers toward a losing path in our competition with China on artificial intelligence (AI).

But first, a word about what’s at stake.

Imagine, if you will, that the question before you today were whether the United States should sell China our advanced propulsion systems that make our nuclear submarines quiet and stealthy. Or imagine that we were debating whether to sell China a prototype of our F-47 next generation fighter for them to reverse engineer. Or whether we should assist the Chinese military’s biological warfare program. Or help Beijing map out the best way to paralyze our power grids through cyberattacks.

We would laugh at the premise there’s any debate.

Yet, the United States has in essence taken a step toward all those damaging outcomes—and more—by relaxing our export controls and selling China the advanced chips they need to surpass us in AI.

The security implications of licensing Nvidia to sell its H200 AI chips to China are clear. It will help China supercharge its military modernization by enhancing its capabilities in nuclear weapons, intelligence gathering and surveillance, cyber warfare, autonomous drones and vehicles, electronic warfare, precision strike systems, biowarfare, command-and-control networks, space warfare, and advanced simulations for training and operational planning.

It will also help China’s heavily subsidized companies compete with American AI developers and cloud-service providers, putting at risk hundreds of billions of dollars in existing or planned U.S. investment and allowing China to eat into America’s market share of what might be the most valuable market in human history at precisely the moment when transformational capabilities are beginning to emerge.

In other words, AI isn’t just a pivotal technology. It is the pivotal technology that will determine whether the United States and its allies, or an axis of totalitarian regimes, control the means to lead in virtually all technologies, military and commercial.  And this is a pivotal moment in its development.

President Trump is aware how important AI is, as many of his recent comments and his official AI Action Plan indicate. He wants to win the competition. He may not be aware, however, of how damaging his administration’s decision to sell China H200s is to that goal.

The first task, then, should be to obliterate the faulty arguments that were made to the Trump administration that put U.S. policy on the wrong track.

Myth #1: China’s Chip Industry is only “Nanoseconds” behind the U.S.

AI leadership boils down to a handful of key factors. Whichever companies, countries, and alliances can marshal the best teams of engineers, acquire the most fulsome reams of data, provide the cheapest and most plentiful electricity, disseminate the most creative and convenient AI applications, and—crucially—harness the most powerful computers will lead the AI revolution and reap the greatest benefits.

We must acknowledge that China is making great strides in most of these areas. Their companies have brilliant engineers (many of them U.S. trained), plenty of data, massively subsidized fleets of coal-fired power plants, and cleverly designed AI apps that they give away for free.

Luckily, the United States has an ace card. In fact, we hold what could be a straight flush. We have privileged access to the most powerful computer chips in the world. American companies like AMD and, especially, Nvidia design those chips. And friendly countries like South Korea and, above all, Taiwan produce the chips and other cutting-edge components that go into them. This privileged access is thanks in no small measure to President Trump: He instituted groundbreaking export controls in his first term that the Biden administration expanded.

China’s lack of access to advanced semiconductors and its inability to produce them at scale constitute the central bottleneck preventing China from overtaking the United States in AI.

Some influential voices have argued that China is catching up so quickly in its indigenous chip production that the U.S. advantage is rapidly vanishing. These voices suggest we may as well lift our export controls and sell China the advanced U.S. chips they need.

“China is going to win the AI race,” Nvidia CEO Jensen Huang said in a November 2025 interview with the Financial Times. He revised his remark shortly after, saying “China is nanoseconds behind America in AI.”

But China’s political leaders and AI developers are far less confident than Mr. Huang in either of those statements. Their recent rhetoric and actions, including by top leader Xi Jinping, show that Beijing is struggling mightily to address a large—and growing—American advantage in AI computing.

In 2024, Xi identified technology as “the main battleground” of geostrategic competition. Last April, he pointedly instructed the Politburo of the ruling Communist Party to “face up to the gap” on core AI technologies. In October, Xi called for China to take “extraordinary measures” to achieve “decisive breakthroughs” in semiconductors. In December, Beijing set aside nearly $70 billion to subsidize China’s semiconductor industry. This is on top of an estimated $145 billion funneled into the sector by the Communist regime since 2014.

These aren’t the efforts of a competitor who is only nanoseconds behind the leader. They are signs of Xi’s deep frustration—even desperation—to address what Chinese AI developers openly acknowledge: China is falling further behind America in computing power.

Listen to what they are saying. A few days ago, on January 10, Tsinghua University hosted a summit of China’s leading AI technicians. Lin Junyang, a top AI scientist at Alibaba, forecast that because of China’s compute gap, there was at most a 20% probability that the world’s leading AI company will be Chinese within five years.  At the same event, Tang Jie, the chief scientist at Zhipu AI, noted that “the gap may not be narrowing—it might even be widening.” Yao Shunyu, chief scientist at Tencent (and a former OpenAI researcher) noted that “the main bottleneck is production capacity,” including advanced equipment and software China needs to make its own chips.

Analysis of Chinese tech behemoth Huawei’s own roadmap points to the same conclusion: the U.S. advantage in computing is poised to keep growing, dramatically: According to Chris McGuire, a former U.S. official who worked on AI policy, “The best U.S. AI chips are currently about five times more powerful than Huawei’s best offerings. By 2027, that gap will widen to seventeen times.”  He added: “Huawei’s strategy of compensating for inferior quality with higher quantity is also failing.”

Myth #2: America Can Induce China to Become “Addicted” to U.S. Technology

Another myth you’ll hear around Washington is that if we sell China enough advanced AI chips, Beijing will become “addicted” to them and give up their efforts to indigenize chip production. Setting aside the faulty logic that America’s national interest would be well served by giving Beijing the means to leapfrog America in AI, Beijing has utterly dismissed the premise that it would allow Chinese companies to remain dependent on Nvidia.

“Even if we introduce Nvidia chips, our determination to pursue independent innovation will remain unwavering,” said Wei Shaojun, vice chairman of the China Semiconductor Industry Association, in an 8 January interview with Chinese state media. “The purpose of importing [technology] is to catch up better, and catching up will eventually lead to running alongside or even taking the lead.”

In other words, Beijing views the import of Nvidia chips as a stopgap measure to help Chinese chipmakers catch up, overtake and eliminate Nvidia as a supplier and competitor. This is consistent with Beijing’s longstanding playbook. And the China Semiconductor Industry Association’s views can be taken as authoritative on this point given that its purpose, according to its charter, is to “uphold Marxism-Leninism, Mao Zedong Thought. . . and Xi Jinping Thought on Socialism with Chinese Characteristics for a New Era.”

China, an economy less than two-thirds the size of the United States, has already spent at least four times what the U.S. Congress appropriated to support our own semiconductor industry with the 2022 CHIPS Act. It is pure fantasy to assume Chinese leaders will cut their losses now in order to “addict” themselves to American chips.

Myth #3: We’ve Already Lost (Recommendations for Winning)

The third myth is the most dangerous: that China’s momentum in the AI race is unstoppable and America should accept its own decline. This is defeatism masquerading as “realism.” America still holds the high ground in the most important technology race of our time. Our companies design the world’s most advanced chips and our companies dominate global cloud infrastructure.

Here’s how to make sure it stays that way:

First, let the Chinese Communist Party struggle to reinvent the wheel we have already created. This could be our Reagan “Star Wars” moment. We rattled the Soviets in the 1980s by merely researching strategic missile defense. This time, we have a real, hardware-based advantage. If China’s rulers want to spend hundreds of billions only to stay in second place, they should be our guest. President Trump would never sell his Golden Dome missile-defense program to the Iranians or the North Koreans. He shouldn’t sell China an advantage in AI.

Second, press the administration to reverse the disastrous decision to let American firms sell advanced AI chips to China. Some at Commerce are trying to stem the bleeding by putting terms and conditions into the export regulations—a valiant effort, but not nearly enough.  Congress needs to step in, reverse the policy, and put durable guardrails in place so the mistake can’t be repeated. To maximally protect U.S. advantages, we also need to close the loopholes that China exploits to make, buy, or rent AI computing power or illicitly exploit U.S. AI technology. If nothing else, reach out to President Trump. In my experience, he always welcomed different viewpoints made in good faith.

Third, get our allies to “level up.” Our allies are indispensable in the global semiconductor supply chain. The Netherlands and Japan are home to companies that make equipment no one else in the world can—including the machines that print circuits on advanced chips. But gaps remain: Dutch and Japanese firms can still sell China lithography machines and service equipment in Chinese fabs in ways American companies cannot. We need to expand our existing controls and push our allies to level up.

 Thank you for the opportunity to testify today. I look forward to your questions.