Anthropic has launched the roadshow for its long-awaited initial public offering, setting a price range of $70 to $78 per share. At the top end, the AI lab would be valued at up to $1.4 trillion on a fully diluted basis—a measure that counts not just shares outstanding but every share that could exist, including reserved stock—though that figure falls well short of the $2 trillion valuation some news outlets had reported just weeks ago.

The company filed confidentially in June and published its prospectus on Oct. 26, with trading in its shares able to begin as early as Wednesday or Thursday next week. Anthropic is offering 825 million shares on the Nasdaq under the ticker "ANT," a sale that would raise as much as $64.4 billion at the top of the range. During a roadshow, a company and its underwriters pitch the deal to institutional investors and gauge demand before the final offer price is set, so how much appetite exists within the $70-to-$78 range should become clear in the days ahead. Seven strategic partners, including Google and Nvidia, have also committed to buying $26 billion of non-voting stock at the IPO price—shares that carry economic exposure to the company without a corresponding say in its governance. For comparison, SpaceX, which listed in June, had sought a valuation of about $1.75 trillion.

Tuesday's amended filing also showed Anthropic's annualized revenue run rate climbing to roughly $81 billion as of the end of October, up from about $71 billion in late September. A run rate extrapolates the most recent period's revenue across a full year, a yardstick commonly used to track the trajectory of fast-growing private companies between reporting periods.

The debut lands at a turbulent moment for the broader AI industry. After the Democrats' major gains in the midterms, the sector has been hit by a sharp sell-off, with investors spooked by the prospect of stricter regulation and tougher scrutiny—particularly around the data center buildout. The Nasdaq fell 5.5% last week and erased nearly $2.5 trillion in market value in a single session, and some analysts have warned the market is tipping into correction territory, conventionally defined as a decline of at least 10% from a recent peak.

Nvidia, one of Anthropic's key strategic partners, has added to the market jitters. The chipmaker's shares closed down more than 4% on Monday after its board opened an independent investigation in the wake of reports that CEO Jensen Huang had been the target of a Chinese intelligence-gathering effort. Those reports followed news that Huang had been romantically linked to a woman named Alice Huang, a former executive at Singapore-based cloud computing and data center company Megaspeed, whom U.S. intelligence agents believe to be an asset of China's main intelligence agency. Beijing has called the accusations baseless, while Nvidia has said Jensen Huang fully supports the review.

The nerves over the midterms and Nvidia's future compound longer-running fears—often voiced by Anthropic itself—about the safety, and potential existential risk, of the AI systems the labs are building.

Anthropic's prospectus carries a fresh warning on the company's progress toward recursive self-improvement (RSI)—the point at which AI systems become capable enough to substantially build their own successors, with each generation speeding up development of the next. RSI has long been a safety concern because that feedback loop could, in theory, push AI capabilities forward faster than humans can test, understand, or control them. Prospectus risk factors are the formal disclosures a listing company puts in front of prospective shareholders, and Anthropic's filing uses that channel to acknowledge the danger, warning that using AI to build more capable AI "could also pose a considerable safety risk, particularly if it sharply accelerates the pace of progress."

In the new amendment, Anthropic said internal estimates about future models indicate it "may imminently meet" its Responsible Scaling Policy's definition of "dramatically accelerated progress attributable to AI." That policy pegs the threshold at roughly a doubling in the pace of AI progress, driven substantially by AI automating research. Anthropic said no current model has crossed it and that it is considering policy changes "to keep humans in control of frontier model development."

This story was originally featured on Fortune.com.