Donald Trump spent part of his weekend in Ireland dismissing the warnings that now come from inside the artificial-intelligence industry. Speaking on September 13, the president said the doom talk around advanced AI “won’t happen,” and that he has “none” of those concerns. He blamed the push to slow down on what he called “negative forces” that should not be raising the matter.
The comments were a direct answer to the essay Anthropic chief executive Dario Amodei published over the weekend under the title “We Must Slow Down the Frontier.” By Monday that essay was moving markets from Tokyo to New York. SoftBank fell more than 11 percent in Tokyo, its worst one-day decline since July 17, after dropping as much as 13.2 percent intraday. Chip and memory suppliers led the losses across Asia and Europe.
The selling was broad. Korea’s KOSPI closed down 3.37 percent, with SK Hynix off 5 percent and Samsung Electronics down 3 percent. Foreign investors sold more than 1.1 trillion won of Korean shares in a single day. In Europe, ASML fell more than 4 percent and Infineon dropped 6 percent. US futures pointed lower, with Intel down nearly 6 percent, Micron off about 5 percent, and Nvidia down more than 2 percent in premarket trading.
The companies hit hardest were the memory and equipment makers that had been the tightest part of the AI supply chain. KB Securities noted that memory inventory at Samsung and SK Hynix now covers less than ten days of supply. Analysts said the repricing is aimed at the pace of capital spending rather than at AI demand itself. The underlying business, they argued, has not changed.
Trump’s reply to all of this was the same line, repeated several times: “Whoever wins AI wins.” The administration’s economic officials have taken the same position in calmer language. Treasury Secretary Scott Bessent said AI research and development should not be paused. He has returned repeatedly to a pair of figures: the United States held roughly 55 to 60 percent of the world’s computing power last year, and he wants that share at 80 percent by 2028.
Bessent’s argument is that slowing the frontier would surrender that lead. SoftBank, meanwhile, was raising money rather than pausing. The Japanese conglomerate finalized an $11.87 billion two-year syndicated loan the same day, up from a $10 billion target, to fund its investment in OpenAI. About twenty banks participated, and Masayoshi Son plans to have committed nearly $65 billion by October.
The essay itself is the latest turn in an argument that has been running for months. Researchers who have trained and tested frontier models have argued that the systems are advancing faster than anyone can fully assess. Investors and builders have responded that the warnings are overblown and that pausing would cede ground to rivals overseas.
Congress has also declined to take up the alarm. House Speaker Mike Johnson said in a weekend interview that there is “no need for everyone to panic right now.” He added that companies have reached no consensus on standards or guardrails, which he offered as a reason for caution rather than urgency. The speaker’s comments leave the safety argument without an obvious sponsor in the House.
The administration’s confidence rests on a particular theory of the race. Bessent’s arithmetic treats computing power as the measure of national advantage, and treats any slowdown as a giveaway of that advantage. It is a view that frames AI as an industrial competition first and a safety question second.
Not everyone on Wall Street is dismissing the warnings, however. Short sellers and some analysts have begun to question the “adjusted” profit figure Anthropic has been presenting to investors. Their point is that excluding items such as stock-based compensation flatters a business that still spends heavily. The dispute shows the slowdown debate has moved from labs and legislatures into the market itself.
The two camps have stopped talking past one another and begun arguing in public. Researchers who have seen the systems up close are warning about risk. Washington’s answer, for now, is that the race itself is the policy.
Analysts said the episode resembles earlier moments in the AI debate, when warnings from researchers generated headlines but no law. A viral essay and a round of official statements can drive a news cycle without producing legislation. The difference now is that the argument has reached the president, the Treasury, and the leadership of Congress at the same time.
The split inside the administration is mirrored across Washington. Some lawmakers have called for new rules, while the leaders of both chambers have so far declined to move. The result is a policy that answers the safety debate with confidence rather than caution, betting that the downside of pausing is greater than the downside of pressing on.
What the weekend established is that the safety case has escaped the labs and reached the political class in a way it had not before. The question is whether that case can be converted into law, or whether it fades the way earlier alarms did. For the moment, the people who set the agenda in Washington have decided that the only acceptable answer to a warning about the frontier is to keep running.
