Alicia García-Herrero, chief APAC economist at Natixis, told Bloomberg on September 24, 2026 that the hardest file on the US-China table right now is artificial intelligence, ahead of Taiwan and trade, according to The Hill. She said so as Bloomberg reported the same day that the two countries extended their trade truce as Trump and Xi met.
The setup matters. According to Reuters, Reuters reported on September 23, 2026 that the US and China agreed to extend their trade truce by two months and work on a bigger deal, per Bessent. The Hill reported on September 24, 2026 that the deal was extended through January as Xi arrived for a meeting with Trump, according to Bloomberg. According to The Guardian, The Guardian reported on September 23, 2026 that Trump greeted Xi Jinping on arrival, in what it described as a rare move. That is the backdrop against which García-Herrero was asked to rank the pressure points.
Why AI Ranks Above Taiwan in Her Read
Her answer inverts the usual hierarchy. García-Herrero said, “I think Taiwan is an easier point for him and for China versus Trump than AI. Trump is loud and clear and very outspoken on the importance of reaching AI before China.” Most observers would put Taiwan at the top of any escalation ladder. She puts AI there, because the administration has staked an explicit commitment to arriving first, and a commitment that public is difficult to trade away at a summit.
Per our editor’s segment, she expects that commitment to translate into continued pressure on hyperscalers to keep competing and keep building the physical footprint that AI requires.
Her Analysis: The Economics Turned
Then comes the part of her argument that is contested analysis rather than settled fact, and it should be read as her view. García-Herrero said US hyperscalers “realize China is ahead of all of that. They don’t have the problem of building data centers and the ultimate costs for both training is much lower.” Her framing is that the pushback in Washington now reflects a changed domestic political environment around data centers and the cost of energy, alongside AI safety concerns, according to Natixis.
Her conclusion is unambiguous. García-Herrero said, “China does not want any restraint on AI currently. Why would it, according to Natixis? It is quite clear that in the current circumstances they can go faster than U.S. on compute, costs and that is basically potentially where we are now.”
A Development That Sits Awkwardly Beside Her View
On the same day she spoke, CNBC reported on September 24, 2026 that China confirmed the first AI talks with the United States have taken place, and hinted at a trade truce extension, according to Bloomberg. That leaves her thesis intact. Talking falls short of conceding, and Beijing confirming that a meeting occurred falls short of offering restraint. But it does sit awkwardly beside “why would it,” and readers should hold both facts at once rather than resolve the tension prematurely.
Tradeoff, and What Investors Should Take From It
Per our editor’s segment, García-Herrero’s framing points to a hard choice: accept Chinese AI leadership or escalate technology competition further than markets can comfortably absorb. She also noted Europe remains squeezed between the two superpowers without leverage to shape the outcome.
Two calendar items sharpen the window. Per our editor’s segment, Taiwan elections are scheduled for later in 2026, and US midterm elections in November 2026 will influence Trump’s negotiating flexibility.
The takeaway for an investing audience is conceptual. If the cost of building and powering data centers has become a competitive variable between countries rather than only between companies, then energy availability, permitting, and siting have become geopolitical facts. That reframes what a summit on AI is actually about (we profiled seven of the power, cooling, and networking suppliers riding that buildout in a free report). For the primary source language on federal AI infrastructure posture, the executive order on removing barriers to American leadership in artificial intelligence is the document to read alongside her argument.