Anthropic’s 48-hour funding clock turns $900B into a railway-mania test
The last time investors tried to buy a real infrastructure revolution this fast, Britain got railway mania: tracks that changed the economy, shares that ran ahead of the tracks, and a crash that punished anyone who forgot the difference. Anthropic, the AI lab behind Claude, is now asking investors to make a similar judgment in 48 hours: whether a company building genuinely useful AI infrastructure is worth close to $900 billion before public markets can test the price.
The clock is the new pressure point. TechCrunch reported that Anthropic is asking investors to submit allocations within 48 hours for a funding round expected to close within two weeks. TechCrunch reported a day earlier that the company had received multiple preemptive offers to raise around $50 billion at an $850 billion to $900 billion valuation.
The railway comparison is not decoration. The Federal Reserve Bank of New York’s Liberty Street Economics blog describes the 1840s railway boom as a period when plentiful money and falling bond yields pushed speculation into railway shares before the commercial crisis of 1847. The technology was real. The overpricing was real too. AI has the same uncomfortable shape: real demand, real data-center bottlenecks, and private prices moving faster than ordinary evidence can catch up.
Anthropic has stronger operating facts than most companies caught in that kind of frenzy. TechCrunch reported that its annualized revenue run rate has surpassed $30 billion, up from roughly $9 billion at the end of 2025, and that one person familiar with its finances said the figure is closer to $40 billion. It also reported that Anthropic raised its last round at a $380 billion valuation in February. A new round near $900 billion would more than double that mark in a matter of months.
The pressure does not stop with Anthropic’s cap table. Fortune reported that Alphabet’s first-quarter profit jumped 81 percent to $62.6 billion, with about $28.7 billion coming from marking up private-company stakes, primarily Anthropic. Marking up means Alphabet records the increased value of its Anthropic shares as profit, even though it has not sold them. Fortune also reported that Amazon’s first-quarter net income included a $16.8 billion pretax gain from Anthropic investments, more than half of Amazon’s pretax income for the quarter.
The accounting is legal. It is also why the 48-hour deadline matters. Robert Willens, an accounting expert cited by Fortune, described the loop plainly: the more Alphabet and Amazon invest in Anthropic, the more profit they can report, without Anthropic paying them cash.
Amazon’s version is especially tight. Anthropic said this month that Amazon is investing $5 billion now, with up to $20 billion more possible, on top of the $8 billion Amazon previously invested. In the same announcement, Anthropic committed to spend more than $100 billion on Amazon Web Services over 10 years to secure up to 5 gigawatts of computing capacity for Claude, its family of AI models. That is not fake demand. It is a real cloud commitment. It also means investor, supplier, and customer relationships are starting to overlap in ways that make clean valuation signals harder to read.
Alphabet has a similar paper-profit exposure. Fortune reported that Alphabet holds an estimated 14 percent stake in Anthropic. TechCrunch reported on April 24 that Google had discussed investing up to $40 billion in Anthropic in cash and compute. If the next round prices Anthropic higher again, Alphabet’s stake becomes more valuable on paper, and the gain can lift earnings even if the operating business did not suddenly sell billions more in search ads or cloud services.
The skeptical case should not erase the builder case. Frontier AI is expensive because training and serving advanced models requires huge data centers, specialized chips, and long-term power commitments. Anthropic may need this much capital because Claude demand is actually growing that quickly. A $900 billion valuation would also put Anthropic above OpenAI’s February round, which TechCrunch said valued OpenAI at $852 billion after the new money was included.
But some investors closest to the lowest entry prices appear to be choosing patience over another markup. TechCrunch reported that some early backers, particularly those who invested in 2024 or earlier, are skipping this round and waiting to potentially cash out during Anthropic’s anticipated initial public offering later this year. Reuters reported that the IPO could launch as soon as October, citing Bloomberg.
That is the real test behind the 48-hour one. Private investors may decide within days whether to bless Anthropic’s next price. Public investors may get their turn within months. If they treat Claude’s revenue and cloud commitments as durable infrastructure, the round will look early. If they decide the paper profits ran ahead of the tracks, the railway analogy stops being a doorway and becomes the warning label.