Palantir’s defence chief says the AI race raises the risk to Taiwan

Palantir defence chief Mike Gallagher said the AI race makes it more tempting for Xi Jinping to seize Taiwan, home to TSMC, before China falls behind for good. He warned that Beijing could then hold the world economy hostage. US intelligence said in March that an invasion would carry a high risk of failure.

Mike Gallagher, Palantir’s head of defence, has warned that the race for AI makes it more tempting for Chinese leader Xi Jinping to seize Taiwan, Bloomberg reported. The island is home to TSMC, the world’s top maker of advanced AI chips, and Nvidia calls it the epicentre of AI.

“The logic of Xi Jinping sacrificing enormously to take Taiwan is getting more compelling,” Gallagher said at a defence policy event in Indiana on Wednesday.

“If we run away with AI, now is his moment to stop us from doing that, because he may not be able to catch up later on,” he added.

Gallagher, a former Republican congressman who led a House committee on China, joined Palantir in 2024. Last month, he called Taiwan the “single point of failure” for AI in a Wall Street Journal opinion piece and urged the US to arm the island heavily.

“If China takes Taiwan, they will control TSMC or disable TSMC,” he said. Beijing would then be able to “hold the rest of the world economically hostage,” he added.

China claims Taiwan as its territory and has not ruled out force. At his White House summit with President Donald Trump in September, Xi urged Washington to oppose Taiwanese independence. In May, Trump put a $14B arms package for Taiwan on hold, calling it a negotiating chip.

US intelligence has played down the near-term risk. A March assessment found that an invasion “would be extremely challenging and carry a high risk of failure.” Still, China has stepped up coast guard patrols off Taiwan’s east coast.

Palantir is a key Pentagon supplier. Its Maven AI system helps US forces with surveillance and targeting, and NATO uses it too.

Gallagher also pointed to talk in Silicon Valley of AI that can improve itself. If that is real, he said, it would be a game changer on par with nuclear weapons in the 1940s.