AI slowdown has no enforcer: Trump, China’s Global Times and Michael Burry all said no

Dario Amodei’s plan to slow the AI race is two days old, and every part of it that needed someone else’s consent is already gone. The Anthropic chief executive’s essay asks for three things: outside evaluators inside every frontier lab, a common speed limit agreed among labs in democratic countries, and eventually a deal with China. By Monday morning the only part still standing was the one Anthropic can do on its own.

President Trump said on Sunday that “whoever wins AI, wins” and blamed “negative forces” for the warnings. Beijing’s Global Times called the essay a “Cold War playbook.” Michael Burry, the investor who shorted the housing market before 2008, posted early Monday that the whole exercise is “self-serving,” because “IPOs need hype & puffery.” And the money kept moving: SoftBank has sealed an upsized $11.87 billion loan to fund its OpenAI stake, and Anthropic is preparing a Nasdaq listing at a valuation near $2 trillion.

What the market sold on Monday was not the pact. It was the suppliers who would feel a slowdown first.

Market Brief

Key Takeaways

5 Points30s Read

1. 01The plan–Dario Amodei’s essay asks for embedded third-party evaluators at every frontier lab, a shared speed limit among Western labs, and coordination with authoritarian governments. Only the first step is within Anthropic’s control, and it has committed to it with METR.
2. 02Washington–Trump said on Sunday that “whoever wins AI, wins” and blamed “negative forces” for the warnings; Speaker Johnson said emergency AI regulation would mean losing the race to China.
3. 03Beijing–The Global Times called the essay a “Cold War playbook” and “hypocritical and short-sighted,” according to Reuters, closing off the plan’s third step for now.
4. 04Wall Street–Michael Burry called the slowdown talk “self-serving” and said “IPOs need hype & puffery,” while SoftBank sealed an upsized $11.87 billion loan for its OpenAI stake and Anthropic prepares a Nasdaq listing at about $2 trillion.
5. 05The market–Suppliers took the hit, not the labs: SoftBank fell as much as 13.2% in Tokyo, Kioxia 9.8%, SK hynix 5.3% and Samsung 3.7%, per Reuters. OpenAI’s IPO delay to 2027 or later is the one capital-market consequence so far.

What “pace the frontier” asks for, and what Anthropic can deliver alone

The essay, published on Saturday under the title “We Must Pace the Frontier”, argues that the industry must slow the rate at which it improves model capabilities, and it names two reasons: recursive self-improvement, AI that builds the next generation of AI, and the episode in August in which a group of OpenAI agents attacked Hugging Face without being told to. Amodei writes that a swarm of that kind could be capable of taking over the entire internet within six to 12 months and cause “hundreds of billions of dollars” in damage. TECHi traced the earlier chapters of that story when researchers linked OpenAI agents to a RubyGems attack.

The plan has three steps, and only one of them is in Anthropic’s gift. Step one: every frontier lab gives “ongoing, employee-like access to a team of embedded third-party evaluators.” Anthropic has committed to that unilaterally, with the evaluation group METR first in line: “desks in our offices, access badges, and company laptops,” plus the right to publish findings about risk levels and incidents without Anthropic’s editorial control. Step two: frontier labs in democratic countries agree on “common safety standards as well as limits on the rate of unchecked AI progress.” Step three: the U.S. and other democratic governments try to coordinate with authoritarian ones, “to the extent this is possible.”

Sam Altman signed on within hours. “I agree with Dario that we need to pace the frontier,” the OpenAI chief executive wrote on X, adding that OpenAI would also give independent evaluators employee-like access. Elon Musk’s reply ran to three words: “Dario is right.”

Steps two and three are where the plan needed help from people who have now declined to give it.

Trump on the AI slowdown: “whoever wins AI, wins”

Asked about the essay at his Doonbeg golf resort in Ireland on Sunday, Trump gave the answer the industry’s Washington critics had predicted. “We’re leading China in AI, we’re the most sophisticated country in the world, and frankly, I want to keep it that way because whoever wins AI, wins,” he said, according to NPR. “And we can put guardrails, we can do this and that,” he added, before blaming “a lot of negative forces that are bringing it up that shouldn’t be bringing it up.” He did not say who the negative forces were.

House Speaker Mike Johnson, on CNN the same day, ruled out the emergency legislation a coordinated pause would need: “If Congress just races in and does some sort of emergency session to try to regulate AI, we will lose the race to China.” David Sacks, who co-chairs the President’s Council of Advisors on Science and Technology, said he supported companies slowing down voluntarily but criticized motives he tied to liability rather than safety.

That leaves step two of Amodei’s plan in an awkward place. Competing companies agreeing among themselves on “limits on the rate” of progress is the kind of arrangement that normally needs either a law or an antitrust exemption behind it. Altman’s post on Monday acknowledged the gap and said OpenAI would not wait: “We welcome a federal framework that sets consistent safety requirements for frontier AI. But we do not believe we need to wait for an anti-trust exemption or legislation to begin the work of providing this confidence.” He also drew the line the labs are now defending: “When we talk about ‘pacing’, we do not mean ‘stopping’.”

Beijing on the AI slowdown: a “Cold War playbook”

Step three fared worse. Amodei’s essay pairs its safety argument with two demands aimed at China: do not sell powerful AI chips or chipmaking equipment there, and crack down on unauthorized distillation by companies in authoritarian countries. He concedes that whether China and other adversaries would comply with any pacing regime is the hardest problem in the plan.

The Global Times, the state-backed tabloid, answered on Monday. The essay was a “Cold War playbook” whose real aim was “to attempt to curb China’s AI development through technological barriers and regulatory monopolies,” it wrote in an editorial, according to Reuters. A “silent AI Cold War” of that kind was “hypocritical and short-sighted,” and excluding China would “significantly increase the trial-and-error costs and risks of loss of control” in AI development worldwide.

Set the rhetoric aside and the structure is the problem. A speed limit that binds American labs but not Chinese ones is exactly the outcome Trump and Johnson say they will not accept, and a limit that includes China needs a negotiating partner that has just called the offer a hostile act. Amodei’s own essay concedes as much: his case for pacing rests on democracies keeping a capability lead while they slow down, which is why the export-control demands are in there at all.

Wall Street on the AI slowdown: Burry’s “hype & puffery” and $11.87 billion for OpenAI

The third missing enforcer is capital, and Monday brought two data points on that.

The first was Burry’s post on X, which laid out four objections in numbered form. Large language models “are not AI and won’t be AGI,” so “there is nothing AI to slow down.” “Competition is coming up fast, slowing benefits incumbents.” “IPOs need hype & puffery,” and a company saying “we are so awesome it could become dangerous” is supplying it. And the talk provides “cover for real uncontrollable slowing growth as IPOs look to be pushed out.”

The second was the money. SoftBank sealed a two-year loan of $11.87 billion from about 20 banks to fund its OpenAI investment, upsized from a $10 billion target, Bloomberg reported on Monday. Anthropic has signed compute commitments worth about $517 billion covering 14.8 gigawatts of capacity in the 11 months through August, up from roughly $180 billion disclosed to investors earlier, and is preparing a Nasdaq listing at a valuation near $2 trillion before the November midterm elections, Investing.com reported, citing The Information and the Financial Times. CNBC confirmed the Nasdaq choice on Monday and described a company walking a tightrope, pitching investors on growth while its chief executive argues for restraint. TECHi reported over the weekend on the second straight quarter of adjusted profit behind that pitch, and last week on the $13.7 billion GPU deal that shows what the spending looks like in practice.

There is one genuine capital-market consequence so far, and it belongs to OpenAI. Altman told Fortune that an initial public offering now would be “an ill-advised moment” and that the timing would be “not 2026,” pushing one of the largest listings ever contemplated into 2027 at the earliest, Reuters reported. He called even a 6% to 10% estimate of extinction-level risk “unacceptable.”

Burry’s fourth point, that slowdown talk is cover for growth that is slowing anyway, is the one the IPO delay makes hardest to dismiss and hardest to prove. OpenAI publishes no growth figures. Anthropic’s leaked numbers, a revenue run rate the same reports put at $65 billion, up from about $9 billion seven months earlier, point the other way.

What the market sold on Monday: SoftBank -13%, Kioxia -10%, SK hynix -5%

The selling landed on the companies that would feel a slower frontier first: the ones that supply it. SoftBank, OpenAI’s largest outside backer, fell as much as 13.2% in Tokyo; memory maker Kioxia dropped 9.8% and Tokyo Electron 3.7%, Reuters reported. In Seoul, SK hynix slid 5.3% and Samsung Electronics 3.7%; TSMC lost 1.2% in Taipei. “Selling pressure is likely to hit AI and semiconductor-related stocks in Tokyo following a series of weekend comments calling for a slowdown in the pace of AI development,” Takayuki Miyajima, senior economist at Sony Financial Group, said in a note quoted by Reuters.

TECHi’s coverage of the SoftBank decline and the KOSPI’s 3% drop has the detail on each market. Nasdaq 100 futures had opened on Sunday evening down 1.2%, twice the S&P 500’s decline, with oil above $100 a barrel and a Federal Reserve meeting this week adding their own pressure.

Note the asymmetry. The labs whose chief executives wrote the pledge are private; the pledge cannot be sold. What could be sold was the capital-spending chain, and the capital-spending chain is what a real slowdown would cut. Jonathan Krinsky of BTIG wrote on Monday that semiconductor stocks have “unfinished business” to the downside, with a test of the 200-day moving average implying a fall of about 17%, and that the iShares Semiconductor ETF is still trading below a declining 50-day average, Investing.com reported. His call is technical and owes nothing to the essay, which is the point: the chip trade was already stretched before anyone proposed slowing it.

What would make the pledge binding

Strip away the responses and the essay has produced one enforceable commitment: evaluators with badges and the right to publish. That is more than the responsible-scaling policies the labs wrote years ago, which Altman conceded on Monday “focused primarily on the deployment of completed models, not what happens during their development process.” OpenAI now says it writes “explicit safety cases in advance of frontier reinforcement learning runs” expected to raise capability significantly.

Three things would turn a pledge into a policy, and each has a date attached. OpenAI has said it will match the evaluator commitment and “have more to share soon”; the test is whether the access is contractual and the findings publishable, as Anthropic’s are. Anthropic’s IPO prospectus, when it comes, will have to describe capability pacing either as a risk factor or as a strategy; that document is the one place where a promise to grow more slowly carries legal weight with the investors Burry is addressing. And the Fed meets on Tuesday and Wednesday with markets pricing an 86% chance of a quarter-point rate increase, according to Investing.com, which would do more to slow AI capital spending than any essay if the cost of a two-year loan keeps rising.

Amodei ends his essay by admitting the measures “will not be easy. But I believe we owe it to humanity to try.” Monday’s answer from Washington, Beijing and Wall Street was not that he is wrong. It was that nobody outside the labs is volunteering to make them try.