The AI Race We Can’t Win With Chinese Parts

Commentary

While U.S. tech executives and government officials speak urgently about the need to dominate artificial intelligence (AI) to compete with China, the infrastructure needed to build that dominance remains critically dependent on the Chinese regime.

From May 13 to May 15, a delegation including America’s top tech CEOs and leaders in AI traveled with U.S. President Donald Trump to Beijing to meet with Chinese leader Xi Jinping. The visit was framed around innovation, investment, and the shared future of technology.

No doubt the conversation touched on AI’s importance to American competitiveness. What likely wasn’t discussed—or, if it was, wasn’t public—is the fact that the United States cannot currently build the data centers required for the desired AI dominance without parts supplied by Beijing.

According to Bloomberg reporting from April, nearly half of all U.S. data center projects planned for this year are being delayed or canceled. That’s a striking figure in and of itself, especially considering the massive public backlash against data centers. But this situation has nothing to do with public opinion. It is also not because of a lack of funding.

AI giants are already spending $650 billion on AI infrastructure in 2026 alone, according to an analysis by Bridgewater Associates. It also isn’t because of a lack of talent or computational chips. The bottleneck is far more basic: electrical transformers, switchgear, and batteries.

The numbers are stark. U.S. imports of high-power transformers from China surged from fewer than 1,500 units in 2022 to more than 8,000 units in 2025. China accounts for more than 40 percent of U.S. battery imports and nearly 30 percent of certain transformer and switchgear categories. Lead times for these components have stretched from 24 to 30 months before 2020 to as long as five years today. For data centers that need to deploy within 18 months, this is a catastrophe.

The contradiction is so fundamental that it deserves to be spelled out in plain language: We are telling ourselves that we must build massive AI infrastructure to compete with China while at the same time depending on China for the basic electrical components that make that infrastructure possible.

Kevin O’Leary, the Shark Tank investor behind the proposed 40,000-acre Stratos data center campus in rural Utah, argued during a recent interview with Tucker Carlson that data centers are essential to American AI dominance over China. He isn’t wrong. The urgency of AI competition is real. The dependency is also real. But the two drivers are in direct tension.

For decades, the United States has outsourced electrical equipment manufacturing. When the AI boom arrived with its voracious demand for power infrastructure, there was no domestic capacity to meet it. Reshoring efforts have been announced by four of the world’s largest manufacturers of electrical power equipment companies, but new U.S. transformer factories won’t be producing at scale until 2027 or 2028, which is years too late for the companies trying to compete right now.

So what happens in the meantime?

To compete in a race that they claim they must win, U.S. tech giants are forced to import more from China, using parts from the competitor they claim they must beat.

I can’t help but see the irony in this, but there are several ways to read this situation.

One such way is as a strategic contradiction that reflects poor planning and the speed of the AI boom outpacing manufacturing capacity. This simple irony suggests that it is a problem to be solved through investment in domestic production and supply chain diversification.

Another possibility is more unsettling. It suggests that the current arrangement—in which U.S. tech giants invest heavily in AI infrastructure while depending on Chinese suppliers for essential components—creates a structural incentive for continued engagement with China rather than competition. It creates mutual dependency disguised as competition. U.S. companies need Chinese parts. Chinese companies benefit from American orders. Both sides have reasons to maintain the relationship, even as they publicly discuss the need to “decouple” from China.

A third option could consider whether the people who recently traveled to Beijing understood the full implications of this dependency. Did they discuss transformer supply chains? Did they negotiate on component exports? Or was the visit framed, as it appeared publicly, as a gesture of innovation and investment without addressing the fundamental contradiction?

This narrative seems neither realistic nor transparent.

The resolution will not come from more investment announcements or more visits to Beijing. It will come from domestic manufacturing capacity that takes years to build or from a genuine decoupling from Chinese supply chains that would require even more time and far greater cost. In the meantime, the United States is competing in a race it cannot currently win, using infrastructure it cannot currently build, powered by components it cannot currently make.

Yet while communities across the nation debate data centers, arguing about water usage, electricity bills, and land use, the central reality remains hidden: We cannot build the infrastructure we claim that we urgently need. We are dependent on Chinese supply chains for the essential components that make AI dominance possible.

So why are we not being told this?

Why does every public conversation about data centers focus on local impacts while the strategic contradiction that justifies building them in the first place goes unmentioned?

One possibility is that they are hoping that communities will be so focused on debating whether data centers should exist in their backyards that no one will ask whether they can exist at all.

Another is that those in power are still scrambling to find a solution and cannot publicly admit that the premise of the entire endeavor—speed, dominance, and competition with China—has already collided with reality.

Either way, we deserve to know what we are actually building and what we are actually dependent on to build it.

Views expressed in this article are the opinions of the author and do not necessarily reflect the views of The Epoch Times.