Anthropic priced its stock at $83 per share last night, five dollars above the top of its range, and sold 980 million shares. The offering raised about $81 billion, making it the largest IPO ever and topping SpaceX. Before the first trade, market makers were bidding in the $100 to $105 range. At $104, that put the company at a $1.9 trillion valuation on paper. The stock later ran from the $80s up to a peak of $111 before settling around $95.
The reaction split the tech world. One side sees the numbers as proof that AI is real. Anthropic generated $4.5 billion last year and $33 billion in the first nine months of this year, with an October pace pointing to $81 billion over the next 12 months. The other side points to roughly $900 billion in compute commitments, some stretching to 2045, and asks whether the bills can ever be covered. A Financial Times analysis says Anthropic would need $602 billion a year in sales just to earn a fair return. Meanwhile, a group of AI researchers, doomers, accelerationists, and co-founders signed a public statement declaring that Anthropic is not a safe AI company. Anthropic’s own banker has projected $685 billion in sales by 2030. The question is whether this becomes the next Facebook—which lost half its valuation after its IPO, and more than half of that in under four months—or whether it keeps running.
What the IPO Numbers Actually Show
Anthropic listed under the ticker ANT. The $81 billion raise is the largest on record. Yet the first-day trading, which peaked at $111 and settled near $95, still looks overinflated to some observers given the underlying commitments. The company moved from roughly $5 billion to an $81 billion run rate in a short period—more than 10x growth. That growth is what bulls highlight. At the same time, the $1.9 trillion paper valuation sits against an $81 billion run rate.
For context, Walmart carries a $1 trillion valuation on $736 billion in revenue. Anthropic is valued at roughly 1.5 times Walmart while generating about one-ninth the revenue. The market is not pricing today’s business; it is pricing a future in which AI captures large portions of other industries. That is the core bet.
The Compute Math and the $602 Billion Threshold
Anthropic’s IPO prospectus outlines more than $800 billion in compute infrastructure commitments through 2045. Some estimates put total computer commitments near $900 billion. One detailed calculation shows that earning a 20 percent return on roughly $1 trillion in capital, at a 40 percent operating margin and 17 percent tax rate, requires about $602 billion in annual revenue. That figure is nearly 10 times the current projected run rate. It also exceeds the combined revenue of Alphabet and Apple and equals about two-thirds of the $908 billion earned by all 76 IT and software companies in the S&P 500.
The company has so far been spending at a ratio well above 2-to-1. Reaching the $600 billion revenue level under that pattern would imply closer to $1.6 trillion in spending. The $80–90 billion raised in the IPO, while record-setting, is not expected to last indefinitely. Blackstone has warned that Anthropic is likely to debut with a double-B credit rating, placing it in junk territory and reversing earlier expectations of investment-grade status. A sub-investment-grade rating raises borrowing costs and limits access to large pools of institutional capital at the exact moment the company faces massive infrastructure outlays.
Blackstone itself is participating in Project Northstar, a roughly $63 billion financing lease covering about 1.5 gigawatts of Google’s AI chips. That structure includes more than $42 billion in senior loans and $18 billion in junior debt, with Blackstone taking about $9 billion of the junior tranche at yields above 10 percent.
The Safety Statement and Concentration of Power
A broad coalition signed a statement arguing that Anthropic portrays itself as an AI safety company while racing to build superintelligence through recursive self-improvement. The statement notes that Anthropic warns recursive self-improvement could concentrate power and escape human control, yet continues to race to build it first. It accuses the company of using safety branding to court investors, employees, and the public while quietly weakening safety pledges and lobbying for self-serving regulation. Signatories included researchers from Harvard, Hugging Face, and an XAI co-founder, among others.
One original founder who signed explained that he did not fully agree with every word—particularly “reckless”—and did not know the statement’s title in advance, but he agreed with the underlying concern. He argued that Anthropic is on track to become the most powerful company on earth and therefore needs a level of transparency and openness closer to that of a sovereign government. He also noted that many safety groups and individual entrepreneurs, executives, and researchers privately agreed with the concerns yet felt unable to sign because of risks to relationships, funding, or access to models. That reluctance itself is cited as an early signal of concentrated power.
The practical effect is straightforward: once a company controls critical tools, public criticism carries the risk that those tools can be withdrawn. Teams racing against deadlines can find their workflows disrupted overnight. Several observers have responded that Anthropic’s policies do not match its name’s implication of acting “for humanity,” and that “Misanthropic” and “closed AI” would be more accurate labels. Nvidia CEO Jensen Huang said one month earlier that nobody is building more compute today than the people asking to be slowed down. The distinction drawn by some is between treating safety risks as engineering problems that can be solved and approaching the entire field with an extinction mindset.
Bull, Bear, and the Post-IPO Path
Oversubscription occurred, though reports of six-to-seven-times demand were overstated; the actual multiple was closer to 1.5 times at most. The stock’s first-day pop of roughly 15 percent was supported by partner investments and a relatively small float, giving structural price support in the short term. Longer-term views diverge. One analyst argues that heavily touted oversubscribed IPOs tend to underperform over the subsequent 12 months as true price discovery replaces scarcity. A prediction of a 20–30 percent decline within six months has been floated for that reason.
Other risks include the sudden creation of large numbers of wealthy employees. Historical patterns at companies such as Microsoft show that major IPOs often trigger departures once lockups ease. Engineers who just became rich leave, and the company loses talent at the moment it needs to execute. In addition, once the market sees the full scale of spending required to reach the projected revenue levels, pressure to raise prices—potentially by four times or ten times—becomes more likely. Higher prices would reduce usage. At the same time, improving open-weight models running on local hardware are expected to capture a share of demand that currently goes to cloud frontier systems.
Local alternatives continue to advance. New Microsoft Surface devices and tools from Apple are increasing the practical power available on-device. Companies that prefer not to send customer data to external clouds are already shifting toward open-weight models running entirely on their own infrastructure. That shift reduces reliance on any single provider’s pricing and availability.

Where the Numbers Leave the Business
The technology itself can be real while the stock remains a high-stakes bet on future dominance. Explaining away $81 billion in annualized sales requires dismissing clear revenue growth. Explaining away $900 billion in commitments requires assuming those bills will be covered at scale. The $602 billion revenue threshold, the junk-credit outlook, the long-dated compute contracts, and the safety-related concentration concerns all sit in the same set of numbers. Whether Anthropic converts today’s investor enthusiasm into sustainable cash flows, or whether the 2-to-1 spending pattern and post-IPO employee exits create lasting pressure, will be settled by execution rather than by the size of the initial raise.




