President Donald Trump has confirmed that artificial intelligence will be a central topic when Chinese President Xi Jinping visits Washington on September 24, framing the encounter as a defining moment in what he called a race that “whoever wins that race is probably going to win. Period.” The confirmation arrives alongside striking new polling data suggesting that Trump’s core message, that America is out in front and intends to stay there, is failing to convince a large share of the American public.
According to a Pew Research Center survey released Thursday and reported by the South China Morning Post, 36% of Americans now believe China is more advanced than the United States in artificial intelligence. Only 12% say the US is ahead. A further 18% consider the two countries roughly equal, while 33% told pollsters they were not sure. It is the first major US public opinion survey to show China clearly leading American perception of the AI race, and it lands at a politically inconvenient moment for the Trump administration, which has spent 2026 trying to reassure allies, investors, and voters that American frontier models remain the global benchmark.
Trump’s framing: “We want to keep it that way”
Speaking to reporters after confirming the September 24 summit, Trump insisted the US retained its edge. “We’re leading China. We want to keep it that way,” he said, before adding the more apocalyptic assessment that whoever wins the AI race will effectively win everything downstream of it. The remarks are consistent with the administration’s messaging throughout the summer, that AI dominance is a zero-sum contest with national security, economic, and ideological stakes, but they also acknowledge that the contest is close enough to be worth staking a presidential summit on.
The Xi visit will be the first face-to-face meeting between the two leaders in Washington during Trump’s second term, and the AI portion of the agenda will build on the working-level dialogue that both governments confirmed earlier this month. As EastFrontier reported when the two sides scheduled their first formal AI talks for September, the diplomatic track has been rapidly overtaken by events on the ground — from Moonshot’s Kimi K3 release to Alibaba’s Qwen 3.8 preview, that have narrowed the perceived capability gap from years to weeks.
The partisan split — and the “unsure” plurality
The Pew data reveals a notable partisan divide. Some 43% of Democrats say China is ahead in AI, compared with 30% of Republicans. Republicans are almost twice as likely as Democrats to say the US is still leading (17% versus 9%). But in both camps, the plurality view is either that China leads or that respondents simply do not know, a finding that suggests the administration’s rhetorical framing of American dominance is not filtering down into settled public conviction.
The 33% “unsure” figure is arguably the most politically significant number in the survey. It represents an enormous bloc of voters who are open to being persuaded in either direction, and it helps explain why the White House has been so aggressive in publicizing enforcement actions against Chinese labs, including last week’s accusation that Moonshot distilled Anthropic’s Fable model to build Kimi K3 — while simultaneously talking up domestic infrastructure buildouts.
Why the numbers have shifted
Analysts have offered several explanations for the shift in public sentiment. The most obvious is the cascade of high-profile Chinese model releases in 2026: DeepSeek R1 and R2 earlier in the year, MiniMax and Z.ai’s Hong Kong debuts, Alibaba’s Qwen 3.8 preview claiming to trail only OpenAI’s Fable 5, and Moonshot’s Kimi K3 topping several open-weights benchmarks. Coverage of these releases has been overwhelmingly framed around the “gap is closing” narrative, and Chinese labs have consistently paired capability announcements with open-weight releases that generate downstream buzz among developers.
A second factor is compute and infrastructure. Z.AI’s reported earlier this month completion of a 1-gigawatt data center built entirely on Chinese-made chips has become a symbol of Beijing’s self-sufficiency push. Combined with Huawei’s move into DRAM, Moore Threads’ 256-GPU system, and CXMT’s imminent $8.6 billion Shanghai IPO, the picture presented to the American public is of a Chinese AI stack that no longer depends on Western silicon.
A third factor is narrative asymmetry. American AI labs have been forced by the Trump administration into a defensive posture — Anthropic was ordered to suspend Fable 5 and Mythos 5 last month over national security concerns, and CAISI, the NIST unit tasked with evaluating Chinese model bias, has quietly stopped publishing findings after director Chris Fall departed after just three months. Chinese labs, by contrast, have been shipping. When the visible signal is one side pausing and the other releasing, public perception follows suit.
What the September 24 summit will actually cover
Officials familiar with the preparatory work have indicated that the AI portion of the Trump-Xi agenda will likely cover four issues: frontier model export controls, chip trade, adversarial distillation, and open-weight model access in the US market. The last item is particularly sensitive. Andreessen Horowitz has estimated that roughly 80% of US AI startups now use Chinese base models, and 200 Silicon Valley founders wrote to the White House this week, warning against a broad ban. On the Chinese side, MOFCOM is separately consulting Alibaba, ByteDance, and Z.ai on possible export controls covering Chinese model weights and training data, a mirror-image response that could reshape the terms of any deal.
Treasury Secretary Bessent has publicly stated that the administration “supports open-source models” while threatening sanctions against adversarial distillation and has hinted that US companies may eventually be required to disclose their use of Chinese models. That disclosure regime, if it materializes, would give Washington its first real map of how deeply Chinese AI has already embedded itself in the American economy, and it would almost certainly influence how the Pew numbers move in the next survey cycle.
For now, Trump’s confirmation of the September 24 date does two things at once. It elevates AI to the highest level of US-China diplomacy, effectively conceding that the technology has become the organizing issue of the bilateral relationship. And it hands Xi Jinping a public stage on which to advance the “AI for all” vision he laid out at WAIC in Shanghai, a vision that, according to Pew, a growing share of Americans already suspect is winning.