Investors in the ever-hotter AI stock rally must suddenly consider a risk with the potential to be even more damaging than high valuations and big spending: Politics getting in the way.
The US administration’s move to block foreign nationals from accessing Anthropic PBC’s most advanced models over security fears was an unprecedented intervention in the affairs of a leading AI lab. In response, Anthropic disabled all access to the two models for everyone.
Washington has for years policed the chips that train artificial intelligence, but its reach for models themselves is a wake-up call: The risk of government intervention is a real threat to AI firms’ global presence and subsequently to their earnings. Those are the very profits that must flow to justify elevated stock prices for technology companies, chipmakers and manufacturers of electrical equipment.
The episode transforms the narrative around AI from a contest to see who is the best and smartest, to one about national security and ring-fencing the most advanced technology. Model availability is suddenly an operational risk, both for providers and users, especially if sanctions appear out of the blue and at random.
Goldman Sachs Group Inc. partner Bobby Molavi framed it bluntly — frontier AI is becoming “state-supervised strategic infrastructure.” What happened was “an example of a kill switch of sorts being used to restrict access,” he said. For Molavi, the question for equity investors is now whether AI model providers will end up becoming “hybrid consumer, enterprise and state-funded defense contractors.”
Market chatter suggests investors took note, although asset pricing did not. For now, AI sentiment is about as bullish as it gets: The early-June collapse in chip stocks has reversed, SpaceX has soared after its blockbuster listing despite concerns, and the Iran war deal has aided the risk-on mood. Futures on the Nasdaq-100 Index were up 1.7% on Friday morning before the open of US exchanges, signalling the tech-heavy gague would rebound following a two-day drop.
The only trade that actually could show a direct impact was the thin over-the-counter market in Anthropic’s pre-IPO stock. A Hyperliquid proxy contract slipped about 3.7% to around $1,627 — a rounding error against the trillions in AI market cap that can’t be traded on the government intervention headlines.
But it would be wrong to dismiss the threat as an idle one. Anthropic already sits on the Department of Defense’s supply-chain risk list after refusing to let its Claude model power mass domestic surveillance and fully autonomous weapons. Investors might have concerns ahead of a planned IPO with a reported valuation near $965 billion.
Given OpenAI is also considering selling shares to the public, the question is whether investors are tempted to rethink valuations now or once pure-play AI firms trade publicly. What’s for sure is that post IPO, this political tail risk stops being theoretical and becomes something markets can choose to discount directly. That also suggests “platform risk” now runs through the Magnificent Seven names with stakes in, or contracts with, AI frontier companies.
Geographically, the implications extend beyond the US. The European-sovereignty trade was already taking shape, with shares in OVH Groupe, the French cloud champion, climbing to their highest since 2023.
French Prime Minister Sebastien Lecornu said Tuesday the US government’s June 12 move was evidence his country needed to develop “strategic autonomy” in AI. It backs the case for France dropping data tools from American AI giant Palantir Technologies Inc. in favor of homegrown alternatives.
European contender, Mistral AI, is in funding talks that would almost double its valuation to €20 billion ($23 billion). An impressive figure, but the gap in financing scale with its US counterparts is brutal. Mistral has raised around $4 billion all-in against roughly $186 billion for OpenAI and $161 billion for Anthropic, and its “sovereign” stack still trains on Microsoft Corp.’s Azure and runs on Nvidia Corp. chips.
China is increasingly positioning itself as the alternative. Beijing’s strategy combines lower operating costs, state backing and a willingness to make advanced models widely available overseas, even as it maintains strict oversight at home through licensing requirements.
After Anthropic curbed access, Beijing-based Knowledge Atlas Technology JSC — better known as Zhipu — moved within hours to unveil its most advanced open-source model yet. Its stock surged 33% on Monday. DeepSeek, the startup that once before rattled the AI world, has made a permanent 75% price cut on its latest model and now ranks among the most-used AI providers on the OpenRouter platform, as measured by token consumption.
Anthropic is in talks with Washington to lift the curbs, and if the jailbreak concerns prove as narrow as the company insists, access could be restored within weeks and the episode will become a footnote.
More likely is that a precedent just got set. “Who gets the best model” is now a decision made by governments rather than companies and political risk becomes a permanent line in every AI valuation — discounted into the IPOs, shaved off multiples, and paid as a premium for anything that touches on sovereignty.
The incident poses an “obvious disruptive risk to the entire AI growth story,” according to Rosenblatt Securities analyst Barton Crockett. “The concern is that we may be entering a period when future advancements in AI frontier models can be seen as unsafe by some in the security arena and spark security-driven holdups by the government,” he said.
Photo: Photographer: Gabby Jones/Bloomberg
Copyright 2026 Bloomberg.
Interested in Political Risk?
Get automatic alerts for this topic.
