On September 24, the presidents of america and China met in Washington and agreed to construct what their joint language calls a “constructive relationship of strategic stability.” On November 18-19, Donald Trump and Xi Jinping are anticipated to fulfill once more, on the sidelines of the APEC Financial Leaders’ Assembly in Shenzhen. Two summits in three months is a uncommon prevalence in China-U.S. relations.
Within the September summit, already their second of 2026, one key space stood out as the main focus: synthetic intelligence. The dangers of AI now have a devoted channel for dialogue. There’s not but a corresponding China-U.S. platform for speaking in regards to the economics of the AI race.
Throughout Trump’s state go to to Beijing in Might, the 2 leaders agreed to restart government-to-government AI dialogue. In September, on the eighth spherical of financial consultations in New York and Washington, U.S. Treasury Secretary Scott Bessent, U.S. Commerce Consultant Jamieson Greer, and Chinese language Vice Premier He Lifeng held the primary AI dialogue inside that framework. They agreed to determine a channel for communications on AI incidents, and scheduled a second spherical of talks earlier than the top of November.
Shortly thereafter, on the presidential summit, Xi described China and america as two AI powers that ought to “draw on one another’s strengths, not guard towards one another.” He added, “The 2 sides can proceed their dialogue on AI, alternate views on its dangers and advantages, and collectively stop the misuse and abuse of AI.” Trump famous that discussions on “urgent points” associated to (amongst different subjects) AI, which he calls “superintelligence,” might “promote peace and prosperity for many years and many years to come back.” Each leaders thus supplied the brand new dialogue platform political cowl.
That’s real progress, with a revealing restrict. All of the steps to date – the exchanges on dangers and advantages, the incident channel, the push for lab self-regulation – concern the risks AI may create. Nothing but addresses the looming financial downside of the AI sector.
The Rising AI Bubble
In america, current developments are fairly worrying. The enterprise mannequin for U.S. frontier AI companies assumes that compute demand will continue to grow exponentially, mannequin functionality will convert reliably into income, and capital will keep affected person. All three assumptions are trying more and more dangerous.
The capex wanted to stay on the frontier is heavier, relative to the trade’s age, than something know-how has carried earlier than. Utilities are signing energy contracts for gigawatts of capability that don’t exist.
In the meantime, Wall Avenue has began asking a query it ignored a yr in the past: the place, precisely, is the return? In July, environment friendly Chinese language open-weight fashions confirmed that frontier-level efficiency may very well be reached with a fraction of the compute. AI-linked U.S. equities offered off arduous; analysts stepped again from value targets constructed on the idea of everlasting shortage.
China has the mirror-image downside. Chinese language corporations try to shut the frontier hole beneath overseas constraints whereas having to supply returns giant sufficient to justify the spending at dwelling. In a September research, the Rhodium Group calculated that the mixed free money circulate of Chinese language tech giants Alibaba, Tencent, and Baidu swung from constructive 170 billion yuan in 2025 to detrimental 16 billion yuan within the first half of 2026.
The 5 American hyperscalers noticed their very own fall from $191 billion to $14 billion and raised $163 billion in long-term debt in six months to maintain constructing.
Clearly, neither the U.S. nor the Chinese language AI sector can fund this race from operations anymore. Each nations’ capital markets are bored with narratives with out precise returns. Endurance on either side is carrying skinny.
The China-U.S. competitors exacerbates these financial dynamics. Every new spherical of U.S. export controls accelerates the Chinese language substitution course of, which shrinks the market accessible for American corporations. On the identical time, deeper substitution raises the funding China should take in, which worsens China’s return-on-capital downside.
In different phrases, additional escalation within the AI race may drain away the capital each methods is dependent upon. This isn’t an ethical argument, however an actual monetary dilemma.
Washington and Beijing have already conceded, of their September dialogue, that they face some frequent dangers from AI dangers. What’s not clear is whether or not the financial downside is included on that agenda.
The Probability at Shenzhen
The November APEC summit presents an unusually sensible setting for addressing this subject. The APEC CEO Summit will result in 1,000 enterprise leaders to Shenzhen with AI on the formal agenda and a Shenzhen Initiative to be launched. The host metropolis itself, lengthy nicknamed the Silicon Valley of the East, holds China’s densest cluster of AI and robotics corporations.
Unofficial channels are additionally in place. Tsinghua’s Middle for Worldwide Safety and Technique and the Brookings Establishment have hosted 14 rounds of China-U.S. AI dialogue since 2019, with Microsoft’s former chief analysis and technique officer Craig Mundie serving as a co-chair.
Three deliverables needs to be on the agenda for the APEC assembly.
First, China and america ought to add the lacking piece of the puzzle in Shenzhen. The official AI dialogue can be held in November; on the APEC summit, either side can begin a standing China-U.S. AI economics dialogue mechanism supported by assume tanks and trade associations, the “monitor two” format lengthy utilized in China-U.S. exchanges.
Each official AI dialog to this point lacks the individuals who management the cash: finance officers, inventory market regulators, and executives. The primary such dialogue might happen on the margins of the APEC assembly, convened by assume tanks and trade associations fairly than ministries.
In Shenzhen, the Guangdong Synthetic Intelligence Trade Affiliation – which operates the Guangdong AI & Robotics Abroad Service Middle, the province’s platform for corporations going international – might co-convene, alongside established assume tanks, a working session between U.S. buyers, analysts, and executives and their Guangdong counterparts, on the trade’s personal floor.
Why begin with the “monitor two” format? As a result of the U.S. and Chinese language governments would discover it arduous to host this dialog. No U.S. administration can brazenly entertain the thought that its AI increase could also be overextended; no Chinese language ministry can candidly examine burn charges with a overseas energy. Finance officers can discuss to economists, and economists can discuss to one another. An unofficial venue is the one why this specific discuss can occur.
Second, China and america ought to set a ground beneath export controls and market entry in Shenzhen. Neither aspect can afford to be seen conducting negotiations over such basic points, however they will quietly alternate purple strains in a personal setting. Misjudgment, not malice, triggers most crises. The most cost effective deterrent is a sentence spoken early.
Third, it’s time to extend collaborations on the periphery. Requirements for AI deployment in third nations – issues like information governance, security analysis, labor displacement – don’t at all times pit U.S. and Chinese language pursuits towards one another. In actual fact, that occurs far much less typically than both aspect assumes. Shenzhen has already hosted APEC-track discussions on accountable AI beneath the Coverage Partnership on Science, Know-how and Innovation framework. A joint sign from Washington and Beijing supporting shared regional norms would price little, match the CEO Summit’s personal agenda, and preserve third events from taking advantage of dueling requirements.
Time Is Brief
By greenlighting discussions on the financial aspect of AI, Washington can be shifting a burden, not making a concession. Trade leaders and educational can be the interlocutors, which means the administration doesn’t must make extra guarantees even whereas the Treasury good points extra perception into AI’s monetary stability.
For its half, Beijing would reinforce its longstanding declare that Chinese language AI is an financial mission fairly than a navy one. The talks would additionally give China’s capital-hungry sector a discussion board to argue towards escalation utilizing industrial – not diplomatic – logic.
The dialogue wouldn’t launch a communiqué or joint assertion. This is able to be a recurring dialog that, over time, builds a shared vocabulary for what sustainable AI competitors might appear like.
The capital markets should not ready; they’re repricing each nations’ AI sectors now. There’s nonetheless an opportunity for China and the U.S. to get forward of the bubble, however the window is closing. After the U.S. midterms, coalition politics in Washington will harden, and each unmanaged month afterwards would strengthen the case for constraint in Beijing.