The Trump administration is intensifying pressure on China across multiple fronts, from AI model investigations to fresh tariff threats, in what analysts describe as a deliberate strategy to extract maximum concessions before President Xi Jinping’s scheduled September 24 visit to Washington. But a growing chorus of experts, industry leaders, and even administration insiders are warning that the approach could undermine the very diplomatic outcomes it aims to achieve.
The Squeeze Strategy Takes Shape
President Trump made no secret of his framing during a press event this week. “We’re leading China. We want to keep it that way,” he said, adding that “whoever wins that race is probably going to win. Period.” The remarks came alongside his confirmation that AI would be on the agenda for the Xi summit, framing the bilateral meeting as a high-stakes negotiation rather than a diplomatic dialogue.
The pressure campaign is multi-pronged. The Bureau of Industry and Security (BIS) has formally opened an investigation into Moonshot AI and other Chinese firms over alleged export control violations related to chip usage. Treasury Secretary Scott Bessent has threatened sanctions against Chinese AI developers. The White House accused Moonshot of distilling Anthropic’s Fable model to build Kimi K3. And the administration is actively weighing a ban or restrictions on foreign-made open-source models, a measure that would effectively cut off American developers from the Chinese AI tools many now rely on daily.
According to SCMP’s analysis, this coordinated escalation is designed to arrive at the September summit with Beijing already on the back foot, having absorbed weeks of economic and technological punishment.
Why Analysts Say It Could Backfire
The core risk, according to Paul Triolo of DGA-Albright Stonebridge, is that the pressure may be too effective — not at extracting concessions, but at destroying the conditions for productive negotiation. “Depending on the number of Chinese companies targeted… the retaliation has the potential to scuttle both the AI dialogue and the September 24 meeting between Presidents Trump and Xi,” Triolo told Reuters.
The bilateral AI safety talks scheduled for September, themselves the product of months of careful diplomacy following the May 2026 Trump-Xi summit in Beijing, are now at risk. Beijing has already signaled it is considering restricting overseas users’ access to Chinese AI models as a retaliatory measure, a move that would directly impact the roughly 60% of token traffic on the OpenRouter platform that currently flows through Chinese models.
The irony is not lost on observers. Washington’s leverage depends partly on the assumption that China needs the summit more than the United States does. But a Pew Research Centre survey released this week found that 36% of Americans now believe China is more advanced than the US in AI, while only 12% say the US is leading. That public perception gap, with 43% of Democrats and 30% of Republicans viewing China as ahead — complicates the administration’s narrative that it negotiates from strength.
A Divided Administration
The leverage strategy is encountering resistance from within the administration’s own ranks. AI czar David Sacks publicly stated that “the Kimi Panic needs to stop. As long as we don’t sabotage ourselves with unnecessary rules, the US will continue to win.” That message directly contradicts the hawkish posture of OSTP director Michael Kratsios, who on July 22 decried Chinese distillation practices and backed the BIS investigation.
The split extends to corporate America. A coalition of 179 startups organized by the Little Tech Association wrote to the administration on July 22 urging against a ban on Chinese open-source models. Two days later, Microsoft, Mistral, Perplexity, and Meta jointly called for a “legal and commercial framework rather than broad restrictions.” Nvidia CEO Jensen Huang went furthest, telling Axios that American companies should “absolutely” be allowed to use Chinese AI models, calling them “excellent.”
These voices represent a significant constituency that views the squeeze strategy as potentially self-harming. As one legal analysis noted, Chinese models account for approximately 61% of tokens processed on the OpenRouter platform, and Alibaba’s Qwen has surpassed 1 billion downloads while forming the base of roughly 40% of new derivative models on Hugging Face. Cutting off access would impose immediate costs on American developers.
The Acceleration Paradox
Perhaps the most significant risk identified by analysts is that maximum pressure accelerates precisely the outcome Washington seeks to prevent: full Chinese AI self-sufficiency. The Wall Street Journal reported this week that Vice Premier Ding Xuexiang, Xi Jinping’s closest confidant, has adopted the same “all-out” approach used for China’s 1960s nuclear weapons program, warning Chinese AI companies that anyone who resisted using domestic chips was “a traitor.”
Huawei expects to ship approximately 1.5 million AI chips this year, roughly doubling 2025 volume. China’s memory chip makers CXMT and YMTC are riding the AI boom to new power, with CXMT posting 719% year-over-year revenue growth in Q1 2026. Each new restriction has historically fueled rather than hindered China’s domestic ecosystem development.
The historical pattern suggests that using technology restrictions as summit leverage creates a perverse dynamic: if Beijing believes the restrictions will come regardless of diplomatic outcomes, it loses the incentive to make concessions. And if restrictions are lifted after the summit as a reward for cooperation, they validate China’s strategy of building without Nvidia while accepting temporary access when offered.
What Comes Next
The two months between now and September 24 will test whether Washington’s gamble pays off. Beijing’s response so far has been calibrated, retaliatory but not escalatory, maintaining the fiction that the summit remains productive. But each new action narrows the diplomatic space available.
The AI gap has narrowed to a record low of approximately 6% by some measures. Kimi K3 ranks third globally on Artificial Analysis benchmarks. And China’s open-weight models are becoming core infrastructure for global tech. The leverage Washington once held, technological superiority so vast that access to American AI was irreplaceable, is diminishing with each passing quarter.
The question is no longer whether Trump can squeeze concessions from Xi. It is whether the squeeze itself is creating the conditions for its own irrelevance.