U.S.-China Relations
NextFin News - The world's most powerful artificial-intelligence technology is advancing faster than the rules meant to contain it, and the two figures framing the debate at the United Nations this week want opposite outcomes. UN Secretary-General António Guterres opened the 81st General Assembly's high-level week by calling for binding international guardrails on AI, warning that "the world cannot afford a race to the bottom on AI safety." President Donald Trump, addressing the Assembly on Tuesday, carried an AI policy built on a different premise: the United States should "win at AI, while dismantling regulatory barriers." The gap between those two lines is the central tension in the AI trade today.
The 81st session of the UN General Assembly opened its high-level week in New York on Tuesday, September 22, 2026, with President Trump second on the speaker list after Brazil's Luiz Inácio Lula da Silva. Six days earlier, in his traditional pre-Assembly press conference, the Secretary-General had devoted nearly half his remarks to artificial intelligence — a signal that AI had moved from a technical sidebar to a top-tier geopolitical issue alongside the wars in Ukraine and the Middle East and the climate crisis.
Guterres framed AI as one of "three existential threats" facing the world, alongside climate change and "the intolerable deepening of inequalities." His argument was specific: development is outpacing understanding. "A growing number of those building it are sounding the alarm – warning that development is racing ahead of our understanding of the risks," he told reporters on September 16. His prescription was equally specific:
"We need guardrails to build trust — and that make AI safe, transparent, accountable, with human dignity at the center," Secretary-General António Guterres said at his pre-Assembly press conference.
But he was blunt about what will not work. "Voluntary pauses that are isolated, unverifiable and unevenly applied will not address the problem," Guterres said — a line that reads as a direct answer to the private-sector slowdown pledges that have circulated through 2026. His alternative is a coordinated international effort, ideally a summit: "I think it makes full sense at the right moment to have a serious international summit on artificial intelligence."
The counterweight took the same podium on Tuesday. President Trump's address continued a policy posture the White House has been building since taking office: the AI Action Plan, released July 23 alongside a "Winning the AI Race" summit, contains more than ninety recommendations organized around accelerating innovation, building infrastructure, and exporting what the plan calls the "American AI stack." At that summit, Trump distilled the philosophy in one line: the plan would help the United States "win at AI, while dismantling regulatory barriers."
The two positions are not merely different tones. They are different theories of how a dual-use technology gets governed. One says the risk is transnational and therefore requires transnational rules. The other says the race is national and therefore requires national speed.
And the market has been watching. As the speech began, U.S. equities were coming off Monday's session, when the S&P 500 closed at 7,778 — up 1.67% from the prior session and about 0.5% below the all-time high of 7,816.70 set in August 2026. The AI trade has been the engine of that rally, and the AI trade is exactly what a regulatory fork threatens.
The Mechanism: Why AI Governance Fractures Along Power Lines
The first-order read of the Guterres-Trump juxtaposition is simple: the UN wants rules, Washington wants speed. The second-order question is why the gap keeps widening even as the technology's risks become more visible.
The mechanism is a race dynamic that turns cooperation into a collective-action problem. AI capability is both an economic asset and a military one. For any single country, slowing down unilaterally means ceding ground to a rival that does not slow down. Guterres acknowledged this directly:
"The world cannot afford a race to the bottom on AI safety," the Secretary-General said. "I am aware of the geopolitical divides and the temptations they might generate."
That sentence contains the whole problem. A "race to the bottom" is what happens when each player's rational move — go faster, regulate less — produces a collectively worse outcome. The Secretary-General's guardrails are an attempt to change the payoff structure so that restraint does not mean defeat.
The Trump administration's position rests on the opposite calculation. Its AI Action Plan explicitly frames the contest as a race with China: "America is in a race to achieve global dominance in artificial intelligence," the document declares. The plan's international pillar recommends that the U.S. advocate against "burdensome international AI governance standards" that would impose requirements on American systems beyond what domestic rules demand, while encouraging partners to adopt U.S. controls and exports. In other words, the preferred governance vehicle is not a multilateral treaty — it is the diffusion of the American technology stack.
This is not an accident of rhetoric. It is a coherent strategy: if American models, chips, and cloud infrastructure become the default abroad, then American norms travel with them. Governance happens through market share rather than through a conference room in New York.
The personal dimension sharpened the week before the Assembly. On September 14, as Guterres was preparing his remarks, President Trump posted on his social network that existential-risk warnings about AI were a hoax:
"I am the Hoax Buster, and I'm right now breaking another Hoax -- That AI is going to take over, consume, and destroy the World, and that Robots will be marching into our Cities, and getting rid of us all!" Trump wrote on September 14. He added: "There is a SICK conspiracy going on against AI and Data Centers, and the only one that is happy about it is China. WHOEVER WINS AI, WINS!"
The timing mattered. Guterres's call for guardrails rested in part on a proposal from Anthropic CEO Dario Amodei to coordinate a slowdown among leading technology companies and governments. Trump's post singled Amodei out by name: "Dario (Anthropic!), who is now pretending to be a 'perfect little angel' — and we will continue to do so." The administration's answer to "why no new rules" was equally direct: "We already have tremendous CRIMINAL and REGULATORY power over these companies!" and the control the technology needs is "a STRONG AND SMART (High IQ!) PRESIDENT, and the U.S.A. has that, in spades!"
So the two poles are fully drawn. The Secretary-General wants a summit with verification. The President wants unilateral American speed.
Cyclical or Structural: This Is a Regime Shift, Not a Policy Cycle
The key judgment for investors is whether this regulatory divergence is cyclical — a disagreement that narrows when the political cycle turns — or structural, a durable feature of the AI era.
The evidence points to structural. Three tests separate the two:
First, a cyclical divergence would revert when leadership changes. But the AI race dynamic does not depend on who holds office; it depends on the existence of a capable rival. The Biden administration also balanced three objectives — leading China, promoting adoption, and mitigating risk — and under Trump the first two are being promoted "at the expense of the third," as one Council on Foreign Relations analysis put it. The weighting changed, but the race itself did not.
Second, the technology's dual-use nature makes the tension permanent. A model that can accelerate drug discovery can also be repurposed. A chip that powers a data center can power a weapons system. That duality means every governance decision is also a security decision, and security decisions are the least likely to be harmonized across rivals.
Third, the institutional architecture is being built along national lines, not multilateral ones. The UN did establish an Independent International Scientific Panel on AI in 2025 — a 40-member body meant to ground policy in evidence. Its preliminary report in July warned of mental-health harms, the "potential use as a destructive tool," and "challenges associated with controlling the technology." But a scientific panel is not a regulator. It can inform; it cannot enforce. Enforcement still lives in export controls, investment screens, and domestic law — all national instruments.
The structural call matters because it changes what investors should price. If this were cyclical, the play would be to wait out the regulatory noise. If it is structural, the play is to assume a permanently higher compliance cost, a permanently fragmented global market, and a permanently privileged position for the incumbents who can afford to navigate both.
The Second-Order Effect: Fragmentation Becomes a Moat
Here is the consequence most market commentary misses. Regulation, when it arrives, will not hit all players equally. It will consolidate the field.
The first-order fear is that rules slow innovation. The second-order reality is that rules raise fixed costs, and fixed costs favor scale. A company that already spends billions on safety testing, compliance staff, and government liaison can absorb a new guardrail regime. A startup cannot. The same dynamic applies across borders: a firm with the legal and engineering capacity to maintain separate model versions for the U.S., the EU, and China is playing a different game than one that cannot.
This is why the "American AI stack" strategy is commercially coherent, not just diplomatically convenient. Exporting the stack means exporting the compliance surface. Partners who adopt U.S. hardware and models inherit U.S. governance expectations as a package deal. The countries that cannot afford the full stack — and there are many — become dependent on whichever power is willing to finance it. The AI Action Plan explicitly targets emerging markets, where Chinese providers have been making deep inroads and where "the Chinese option is sometimes the only one on the table," as one analysis noted.
The irony is sharp: a world of competing governance blocs is less safe in the Secretary-General's terms, but more profitable for the companies that can operate inside more than one bloc. Fragmentation is a moat for the incumbents.
The Counter-Thesis: Speed Is the Safer Bet
The strongest argument against Guterres's position deserves its due. It runs like this: the country that moves fastest captures the technology's gains first — in productivity, in military capability, in economic growth. A pause, even a coordinated one, hands the initiative to actors who do not honor it. And the historical record of international technology governance is mixed at best: treaties work when verification is possible and stakes are symmetric, and AI offers neither easy verification nor symmetric stakes.
There is also a democratic argument. Slowing deployment means delaying the benefits — in medicine, education, climate modeling — that AI could deliver now. From that vantage point, the Trump administration's deregulatory stance is not recklessness; it is a bet that American openness produces better outcomes than a negotiated minimum standard that locks in the lowest common denominator.
This counter-thesis is not fringe. It is backed by the administration itself. Vice President JD Vance said the industry's own push for rules is suspect: "I feel a little bit weird about the fact that you have so many frontier AI tech companies kind of coming to the government and begging the government to regulate them." And there is a point about enforcement: a guardrail that cannot be verified is a promise, not a rule. Guterres himself conceded the verification problem when he rejected "isolated, unverifiable" pauses.
But the counter-thesis has a hole. It assumes the race is only between responsible actors. It does not fully account for the possibility that the technology itself develops capabilities that no single company or government can reliably control — the scenario the 40-member scientific panel flagged when it warned of systems that may become difficult to control at scale. If that risk is real, then even a speed-advantaged winner is winning a race toward a cliff.
Conclusion: What to Watch, and What Would Prove This Wrong
The structural-fragmentation judgment can be tested. Watch two observable signals:
First, whether the U.S. and China engage substantively on AI safety inside a multilateral forum within the next twelve months — a working-level dialogue on frontier-model risk with shared incident reporting would signal that cooperation is possible and that the fragmentation call is too pessimistic.
Second, whether the U.S. Congress passes a frontier-model licensing or preclearance regime within eighteen months. The executive order signed June 2 declined to impose licensing; if Congress imposes it anyway, the deregulatory path has reversed and the fragmentation thesis weakens materially.
The immediate implication is that the AI trade now carries a governance discount that did not exist a year ago. Investors priced AI as a pure growth story; they must now price it as a growth story with a regulatory overlay that differs by jurisdiction.
Who benefits: the hyperscalers and chip incumbents with the balance sheets to absorb compliance costs and the lobbying capacity to shape them. Who is exposed: smaller model developers, open-source labs without commercial safety infrastructure, and the countries that become battlegrounds for stack adoption rather than customers with real choice.
The time-horizon split:
- Short term (sentiment/liquidity): volatility around UN week and any new regulatory announcement. Headlines will move names, but the direction depends on whether the market reads a rule as a speed bump or a barrier.
- Medium term (fundamentals): compliance costs rise, margins compress for players who cannot spread them over scale, and consolidation accelerates.
- Long term (structural): two or three governance blocs emerge — an American-led stack, a Chinese-led stack, and an EU-style rights-based regime — and the companies that can operate across all three become the durable winners.
The base case is continued divergence with no binding global treaty before 2028. The upside case is a limited U.S.-China safety dialogue that contains the most dangerous capabilities without slowing commercial deployment. The downside case is a verified AI-enabled harm at scale that forces emergency regulation and triggers a sharp repricing of the entire sector.
Guterres ended his press conference with a line that captures the stakes: "At a time when the world faces three existential threats – runaway artificial intelligence, the climate crisis, and the intolerable deepening of inequalities – international cooperation is more necessary than ever." The counterpoint, delivered from the same podium, is that cooperation without speed is its own kind of risk.
The market's verdict will not come from a speech. It will come from the companies that can navigate both.
The AI race is not being slowed by regulation — it is being split by it, and the winners will be the few who can run in more than one lane.
Explore more exclusive insights at nextfin.ai.
