Communist China has spent years quietly embedding itself inside the hardware that keeps American artificial intelligence running.
Now the Trump administration is doing something about it.
And what security experts found buried inside the supply chains powering Big Tech’s data center empire is going to make your blood boil.
The Problem Nobody Wanted to Talk About
The Trump administration has moved to tighten scrutiny of Chinese-made equipment powering America’s data centers, citing national security risks as the electricity demand from AI systems surges to levels the grid was never built to handle.
This did not happen in a vacuum. In April, President Trump issued a Presidential Determination naming transformers, transmission lines and conductors, substations, high-voltage circuit breakers, and power control electronics as “essential to the national defense.” That language matters. It is the kind of designation that opens the door to restrictions, sourcing requirements, and enforcement action.
Big Tech has been racing to build data centers at a pace that strains credulity. According to S&P Global, US data center capacity is forecast to grow from 62 gigawatts in March 2026 to 152 gigawatts by 2030. That is not a modest expansion. That is a near-tripling of capacity in four years, and every watt of it has to come from somewhere.
The problem is where a lot of the hardware has been coming from.
What the Experts Are Actually Saying
Yury Dvorkin, an associate professor at Johns Hopkins University, put the exposure in plain terms. “The mid-term exposure is concentrated in grid connectors: transformers, switchgear, and batteries,” he said. He added that “China’s share of certain transformer and switchgear categories runs near 30%, and it accounts for over 40% of U.S. battery imports.” And the vulnerability does not stop at the data center fence line. Dvorkin noted that “our analysis shows that there is also a deeper exposure upstream: copper, electrical steel, and battery cathode materials.”
That upstream exposure is the part that tends to get glossed over. People focus on finished components and miss the raw material layer entirely. Communist China controls enormous portions of the global supply of the inputs that go into the things that go into the things that power American AI. That is a dependency problem stacked on top of another dependency problem.
Ben Boucher, senior analyst for supply chain at Wood Mackenzie, said the dependence on Chinese imports runs across multiple parts of AI infrastructure. “There are a few key areas that the AI infrastructure buildout is increasingly dependent on Chinese imports,” he said, pointing specifically to “substation transformers, which hyperscalers typically have on site to step transmission voltage down.”
Wood Mackenzie’s own research found that power transformers are already in an estimated shortage of 15% and substations are at 8% in 2026. Restrictions on Chinese-made units will make that worse before it gets better. Boucher wrote in an August blog post that “a bulk of the impacts will be centered around data centers who have been using Chinese units to minimize lead times,” and that “the 100 MVA+ segment is where the shortage is already most acute, and it is precisely where data centers have been turning to Chinese manufacturers to manage lead times.”
And then there is the optical transceiver piece, which does not get nearly enough attention. Zhongji Innolight and Eoptolink lead the world in data center optical transceiver revenue. Chinese firms together account for roughly two-thirds of global optical transceiver unit supply, according to an August report by Counterpoint analyst Neil Shah. These are not peripheral components. Optical transceivers are how data moves through the fiber-optic cables inside these facilities at the speeds AI workloads require.
Laveena Iyer, a senior analyst at The Economist Group, told CNBC that Washington’s focus has broadened considerably. “Scrutiny over China’s presence in the U.S. data centre power stack has only risen over the past eight months or so, this is a shift from the previous focus that centred solely on the compute stack,” she said.
That shift is significant. For years, the policy conversation fixated on chips. Semiconductors. Nvidia. The compute layer. What Iyer is describing is a dawning recognition that the compute layer sits on top of a power and connectivity infrastructure that is itself deeply compromised by Chinese supply chain penetration.
Trump’s Tariffs Are Forcing the Reckoning Big Tech Avoided
A White House spokesperson told CNBC that “reshoring manufacturing that’s critical to our national and economic security has been a top priority for President Trump, and the Administration continues to deliver with a robust and nimble agenda of tax cuts, tariffs, and deregulation.” The spokesperson added that “trillions in investments across key sectors — from steel to semiconductors to autos — prove that the Administration’s strategy is paying off.”
Some companies have started responding. Hitachi Energy announced in September 2025 a $1 billion investment to build critical grid infrastructure manufacturing in the US, including $457 million for a new large power transformer facility. Siemens Energy said in February it would invest $1 billion in US production of grid and gas turbine equipment for AI infrastructure and data centers.
But these announcements come against a backdrop of years of neglect. Big Tech spent the last decade optimizing for speed and cost, which meant buying from whoever could deliver fastest at the lowest price. Communist China was very good at being that supplier. The same companies now scrambling to find American-made transformers are the ones that spent years building supply chains that ran straight through Beijing.
US optical technology firms Lumentum and Coherent stand to benefit from any restrictions placed on their Chinese competitors. Nvidia invested $2 billion each in the two companies in March — a signal that at least someone in the AI supply chain is starting to think seriously about where the hardware actually comes from.
The problem is the timeline. Neil Shah’s August Counterpoint report noted that “Western competitors like Coherent and Lumentum possess advanced photonic designs, but currently lack the cleanroom capacity, automated packaging infrastructure, and yield scale required to absorb Innolight and Eoptolink’s volume within a 12-to-24-month horizon.” In other words, even if restrictions came tomorrow, American companies could not immediately fill the gap. That is what a decade of offshoring to Communist China actually costs.
And Iyer noted what happens if Washington tries to ban Chinese optical transceivers outright: US hyperscalers would have to look for alternatives locally, potentially pushing up costs. Those costs do not stay inside the data center. They get passed along — to businesses, to consumers, to the working families already watching their electricity bills climb because these same data centers are consuming power at a rate the grid was never designed to sustain.
The 5G comparison is instructive here. Washington spent years allowing Chinese telecommunications equipment to embed itself in American networks, then discovered that ripping it out was enormously disruptive and expensive. The lesson from that episode was supposed to be: do not let Communist China get so deeply embedded in critical infrastructure that removal becomes painful. That lesson was not learned. The AI buildout repeated the same mistake, just in a different layer of the stack.
Trump’s tariffs and the Presidential Determination are the kind of trade enforcement tools that force the reckoning that the market, left to itself, would keep deferring. Big Tech will complain about costs and timelines. Those complaints deserve exactly as much sympathy as the companies that chose to build their supply chains through a geopolitical adversary while knowing the risks.
The same companies now warning about disruption from Chinese supply chain restrictions are the ones that spent years censoring conservative Americans on COVID policy, questions about the 2020 election, and the events of January 6. The idea that these corporations deserve a pass on national security accountability because restricting Chinese hardware might slow their expansion is not a serious argument.
What Trump is doing here is exactly what an America First trade policy is supposed to do: use the leverage of the American market to force investment back into domestic production, protect critical infrastructure from a known adversary, and stop letting Communist China profit from building the backbone of American AI while simultaneously working to undermine it. The short-term disruption is real. The long-term alternative is worse.
Source: Breitbart, “Security Experts: AI Data Centers Contain Hidden Risks from China,” September 7, 2026; CNBC, “Hidden China risks are emerging in America’s multibillion-dollar AI data center boom,” September 3, 2026
