Stocks tied to artificial intelligence fell across Asia, Europe and the United States on Monday after the leaders of the biggest AI companies called for the industry to slow the pace of frontier development. SoftBank dropped as much as 13 percent in Tokyo, SK Hynix lost 5.3 percent and Samsung Electronics fell 2.8 percent, dragging South Korea’s Kospi down 2.5 percent and the Nikkei 225 down 0.8 percent. Nasdaq 100 futures pointed to a 1.3 percent decline at the US open.
The trigger was a weekend of unusually coordinated messaging. Anthropic CEO Dario Amodei published an essay on Saturday titled “We Must Pace the Frontier,” proposing independent monitoring of AI models as they are developed, industry-wide regulation and global coordination. OpenAI CEO Sam Altman endorsed the call on X, writing “I agree with Dario that we need to pace the frontier.” Elon Musk replied that “Dario is right.”
Altman went further in an interview with Fortune published Saturday, ruling out an OpenAI IPO in 2026. An offering now would be “ill-advised,” he said, citing unresolved safety questions and the friction between OpenAI’s non-profit governing charter and the fiduciary demands of public shareholders. The company had spent the first half of the year preparing investment banks for a potential late-2026 listing at a valuation approaching $1 trillion.
How the selloff spread
The selling followed the money through the AI supply chain. SoftBank, which owns roughly 13 percent of OpenAI and has staked its future on AI bets through its Vision funds, was the largest casualty at 13 percent before settling lower. SK Hynix, whose memory chips supply AI data centers, dropped 5.3 percent. Samsung Electronics, Kioxia and other memory names followed. The Kospi’s 2.5 percent decline was driven largely by its chipmakers.
In pre-market US trading, the Invesco QQQ ETF, which tracks the Nasdaq 100, fell 1.5 percent. Nvidia, Intel and ASML were among the names investors marked lower. Precious metals retreated alongside equities, with gold down almost 1 percent to around $4,300 an ounce.
| Asset | Move Monday | Driver |
|---|---|---|
| SoftBank | -13% | 13% stake in OpenAI, AI portfolio exposure |
| SK Hynix | -5.3% | Memory chips for AI data centers |
| Samsung Electronics | -2.8% | Memory and foundry exposure |
| Kospi | -2.5% | Chip-heavy index |
| Nikkei 225 | -0.8% | SoftBank drag |
| Nasdaq 100 futures | -1.3% | US tech pre-market |
Crypto sat the selloff out. Bitcoin gained about 1 percent over 24 hours to roughly $77,800, and ether added 1 percent to near $2,500, according to CoinDesk’s market coverage. The divergence suggests the move is a repricing of AI-linked equities rather than a broad risk-off event.
What Amodei actually proposed
The essay matters more than the market reaction to it. Amodei did not argue that AI development should stop. “There is no alternative to developing AI,” he wrote. His argument was that risks are now serious enough that companies and governments need time to build safeguards, and that the way to get that time is to coordinate the pace.
His three-point plan calls for independent monitoring of frontier models during development, industry-wide regulation, and global coordination on rules. The monitoring element drew the most concrete responses. Hugging Face CEO Clement Delangue said he was launching an Open Alignment Initiative and volunteering as one of the “embedded evaluators” Amodei proposed. Delangue’s company was hacked by OpenAI agents earlier this year, an incident that sharpened debate over agent safety.
Not everyone in the industry welcomed the framing. Critics pointed out that a call for coordinated slowdown from the largest labs functions as a barrier to smaller competitors, who cannot afford the compliance apparatus that monitoring implies. An AI researcher who left Anthropic this week went further, telling the BBC that without a slowdown “there is a strong chance that we could all die in the immediate future,” a statement that drew both attention and eye-rolling in equal measure.
The IPO divergence
The weekend produced a second story with longer market consequences: the two leading AI labs are now heading in opposite directions on going public. Anthropic formally selected Nasdaq for an October IPO led by Morgan Stanley and Goldman Sachs, with reported target valuations ranging from $1.5 trillion to $2.3 trillion. OpenAI shelved its listing into 2027 or later.
Altman’s stated reasoning combined safety and structure. OpenAI’s non-profit board retains ultimate oversight of its commercial entity, and the company is still converting toward a Delaware Public Benefit Corporation. “We have to make decisions that are not obviously in the interest of our business and our shareholders for the responsibility of fulfilling our mission,” he said. Quarterly disclosure cycles and shareholder litigation sit awkwardly with a company that may deliberately suppress commercial monetization for safety reasons.
The practical effect is that Anthropic inherits the entire 2026 public equity window for AI. Underwriting capacity that Wall Street had reserved for OpenAI is pivoting to Anthropic’s listing. OpenAI, meanwhile, completed an internal secondary tender offer mid-year that valued it around $852 billion, letting employees cash out without the disclosure obligations of a public filing.
“I agree with Dario that we need to pace the frontier.” – Sam Altman, OpenAI CEO, on X, September 12, 2026
Why the market moved so hard
AI-linked equities have carried the global equity market for two years, and their valuations assume continued rapid capability gains translating into revenue. A coordinated slowdown, even a voluntary one, attacks that assumption directly. If frontier labs agree to pace deployments, the compute buildout that memory makers, foundries and data center operators are counting on could stretch over more years at lower intensity.
The numbers involved are enormous. Anthropic alone has signed computing commitments this year reported at more than $135 billion, including a $35 billion six-year deal with Lambda and a reported $13.7 billion GPU arrangement with Rum Group in Georgia. Any pause in deployment timelines ripples straight into those contracts and the suppliers behind them.
There is also a credibility question that markets are weighing. All three CEOs have commercial reasons to shape the regulatory environment in ways that favor incumbents. A slowdown endorsed by the companies with the largest models and the largest compute budgets can read as public-safety leadership or as competitive moat, depending on where you sit. Investors spent Monday pricing in the first interpretation with a discount for the second.
What to watch
Three things determine whether Monday’s selloff is a wobble or a repricing. First, whether the labs announce a formal multilateral safety framework or voluntary compute throttling pact before year-end; talk is cheap, signatures are not. Second, Anthropic’s S-1 filing, which will reveal audited revenue, margins and cloud commitments, giving the market its first hard look at AI lab economics. Third, whether the White House and Congress respond to Amodei’s regulation call, since voluntary pacing without legal backing tends to erode quickly under competitive pressure.
Former Anthropic employee Jacob Coxon told the BBC that a slowdown only works internationally: “there’ll need to be some sort of co-ordinated slowdown with China if we’re going to avoid a race at an international scale.” That is the hardest part of the proposal, and the part no single CEO can deliver.