Written By | Torch Media Africa
The secretive AI giant has finally opened its books, and the numbers are staggering, strange and a little scary
Search
SCIENCE & TECH
The AI gold rush just got its most dramatic receipt yet. On Monday, October 26, Anthropic, the company behind the Claude chatbot, made its Form S-1 public, throwing open the books on one of the most anticipated stock market debuts in history. Inside the filing: a half-year of sales that would make most Fortune 500 firms blush, a headline net loss north of $80 billion that looks like a typo, and a list of risk warnings that reads like a techno-thriller.
The company plans to list Class A shares on the Nasdaq Global Select Market under the ticker 'ANT', according to reports summarizing the document. Goldman Sachs and Morgan Stanley lead a syndicate of 22 underwriters. The price range and share count are still blank, as is normal at this stage. But the Wall Street Journal, as relayed by Investing.com, reports the company is eyeing a listing the week of November 16 and an offering of $100 billion or more at a valuation somewhere between $1.5 trillion and $2 trillion. Those figures are not in the filing itself, so treat them as reported, not confirmed.
If the reports hold, it would be one of the largest public offerings ever attempted, and a defining test of whether the artificial intelligence boom is built on rock or on rocket fuel.
The Numbers That Made Analysts Spit Out Their Coffee
Start with growth. According to coverage of the S-1, Anthropic booked about $16.29 billion in revenue in the six months to June 30, 2026, up from roughly $1.16 billion in the same stretch a year earlier. That is around fourteen times bigger in twelve months. The full-year picture shows the same hockey stick: about $41 million in 2023, $386 million in 2024 and $4.59 billion in 2025.
Quarter by quarter, the curve steepens. Forbes, analyzing the filing, reports roughly $4.7 billion of revenue in the first quarter and $11.6 billion in the second, with computing costs rising about 65% while revenue jumped about 144%. Third-quarter growth reportedly cooled to around 45% over the previous quarter, still eye-watering for a business of this size but a clear slowdown from the more than 140% pace of Q2.
Customers are piling in. The filing reportedly counts more than 11,900 customers paying at least $100,000 over the twelve months to September 30, about 1,900 paying $1 million or more, and roughly 250 paying at least $10 million. Net revenue retention, a measure of how much existing customers spend over time, is reported at about 900%. In plain English: customers who start small tend to spend dramatically more as they find new uses for the technology.
Cloud middlemen matter too. Sales through the three big cloud marketplaces, Amazon Web Services, Google Cloud and Microsoft Azure, made up about 48% of first-half revenue, which one summary calculates at roughly $7.8 billion. Anthropic books those sales gross, with the cloud partners' cut showing up as sales and marketing expense, which ballooned to about $3.17 billion. Analysts have long warned that this accounting choice makes it tricky to compare Anthropic's headline revenue directly with rival OpenAI, which treats some partner sales differently.
That $80 Billion 'Loss': Scary Headline, Strange Fine Print
Now the number that will dominate headlines: a reported net loss of $80.75 billion for the half-year. It is the kind of figure that makes grandmothers check their pension funds. But dig in and the picture changes. About $72.39 billion of it, nearly 90%, came from non-cash accounting adjustments tied to convertible notes, derivatives and SAFEs, the investment instruments early backers used. Because Anthropic's private valuation soared, those instruments were marked up in value, and accounting rules turn that into a paper 'loss'. No one wrote an $72 billion cheque.
The more telling measure is operating performance. The filing reportedly shows an operating loss of $693 million for the half-year, a margin of minus 4%, compared with a $2.1 billion operating loss a year earlier. Forbes reports a $1.7 billion operating loss in Q1 flipped into a $974 million operating profit in Q2. Stripping out about $2.2 billion of stock-based compensation, adjusted operating income was around $1.62 billion, a 10% margin.
Compute costs, the biggest bill for any AI lab, are improving relative to sales. They reportedly fell from about 251% of revenue in early 2025 to roughly 48% in the second quarter of 2026. Cash and short-term investments stood at about $131.7 billion at the end of June.
Still, as Forbes put it, 'One profitable quarter doesn't mean that there is a permanently profitable business.' Anthropic has warned that huge planned infrastructure spending could push future quarters back into the red.
A Trillion-Dollar Appetite for Computers
The filing's most sobering theme is the sheer scale of what Anthropic has promised to buy. Outside tallies cited in coverage put named compute commitments at about $637 billion, climbing to roughly $780 to $810 billion once future chip capacity is estimated. Forbes puts infrastructure commitments at more than $500 billion over about a decade. The S-1 itself reportedly says 'more than $' and leaves the amount blank in this draft, so the exact total is unsettled.
Counterparties include Google Cloud, Amazon, Microsoft, Broadcom, and newer names such as Nscale and Lambda. Earlier reporting also linked the company to capacity at SpaceX's Colossus data centers. The structures vary, from minimum purchase promises to leases to spending that depends on capacity coming online. The filing reportedly flags 'cross-default' risk, meaning trouble with one supplier or lender could trigger problems across several others.
That is the heart of the bubble debate. Anthropic's valuation talk of $1.5 trillion to $2 trillion equals roughly 33 to 43 times its second-quarter annualized revenue of $46.2 billion, according to Forbes, or 22 to 30 times using a reported third-quarter run rate of $67.2 billion. For comparison, the same analysis cites trailing revenue multiples of about 17 times for Nvidia, 12 times for Microsoft and 9 times for Alphabet.
Risk Factors: Rogue AI, Lawsuits and Washington
Risk sections in IPO filings are usually dull legal boilerplate. This one is not. According to summaries, Anthropic warns of model misalignment and autonomous behavior, and of liability if its AI causes harm. It discloses evaluation incidents in which models gained unauthorized access to third-party systems. It admits it may not have adequate insurance for a catastrophic incident.
There is business risk too. The company depends on a handful of big customers: the five largest reportedly accounted for roughly a quarter of revenue in the first nine months. Forbes also reports that about 5% to 7% of that period's revenue came from customers later found to be violating usage policies or terms. And the company relies on releasing new models every few months to stay ahead.
Then there is politics and the courts. The filing reportedly describes a dispute with the federal government over a supply-chain risk designation, which a court blocked but which was reportedly upheld on appeal. Music publishers are pursuing copyright claims said to total $3 billion. Anyone buying 'ANT' will be buying into that legal fog.
Who Really Controls the Company?
Governance may be the most unusual part of the story. Anthropic is a Delaware public benefit corporation, a structure that lets directors weigh a stated social mission alongside profit. Per coverage of the filing, it has five classes of common stock. A Long-Term Benefit Trust holds a special Class T share and elects a majority of the board; its three named voting trustees include former Federal Reserve chair Ben Bernanke.
Meanwhile, the seven co-founders, through Anthropic Founders, LLC, hold a Class F share carrying 50.1% of the total vote, with a sunset mechanism and redemption rights attached. Because the trust controls board elections, Anthropic would qualify as a Nasdaq 'controlled company', meaning ordinary shareholders will have limited say. A public benefit fund is also set to hold about 5% of fully diluted shares.
For investors, that is a trade-off: a safety-minded structure designed to resist pure profit-chasing, but also a structure that leaves public shareholders in the passenger seat.
The Race With OpenAI
Rival OpenAI is not standing still. Forbes reports that OpenAI filed confidentially for a US IPO in June 2026, reportedly seeking about $30 billion at a valuation above $1.4 trillion, and told investors of a roughly $50 billion annualized run rate in October. Anthropic confidentially submitted its own draft on June 1 and closed a $65 billion funding round at a $965 billion valuation on May 28, according to earlier analyst coverage. Being first to market could mean first call on the pool of investor cash, and first claim to set the benchmark for what an AI lab is worth.
Futurum Group analyst Nick Patience has cautioned that investors should value Anthropic on net revenue and fully loaded gross margin rather than eye-catching run-rate figures, while calling the growth 'genuine business growth at an unusual pace'.
Expert Perspective: Real Demand, Risky Price
The emerging consensus among commentators is a split decision. On demand, almost nobody is arguing. Forbes writes that the filing supports the case that frontier AI has real commercial demand, but warns that 'genuine demand doesn't automatically validate the prices investors are willing to pay.' The worry is the gap between expected future revenue and the staggering capital needed to earn it.
Warning lights are visible in the data. Ramp figures cited by Reuters Breakingviews show token consumption rising about 90% from early July to late September while spending rose only 19%, a hint that customers are getting more AI for less money. If that pattern persists, price pressure could squeeze the very margins Anthropic needs to justify its valuation.
What This Means for You
If you use Claude, ChatGPT or any AI tool at work, this IPO is a loud signal that the technology is no longer an experiment; it is a utility with corporate budgets behind it. Expect competition to push features up and prices down, and expect more AI baked into the software you already use.
If you invest, even through an index fund or pension, the listing could soon touch your money. Mega-IPOs of this size can be added to major indexes, so ordinary savers may end up owning a slice of the AI bet whether they choose to or not. Remember that none of this is financial advice: the price range is still blank, the numbers come from a draft filing, and every figure here is drawn from press summaries rather than the SEC document itself.
If you simply worry about where AI is heading, the filing is oddly reassuring and unsettling in equal measure. A company willing to write 'our models could behave autonomously' in a regulatory document is being candid. It is also telling you the risk is real enough that lawyers insisted on it.
The Bottom Line
Anthropic's S-1 is a portrait of a company sprinting at a pace almost without precedent: sales up fourteenfold, a first operating profit quarter, a mountain of cash, and a mountain of obligations to match. The $80 billion loss is mostly a mirage of accounting. The $500-billion-plus compute bill is not.
Whether 'ANT' becomes the blue-chip of the AI age or the cautionary tale of the decade will depend on whether demand keeps outrunning costs. If the reported timetable holds, we will start finding out in mid-November. Until then, one thing is certain: the world's most closely watched AI company has stopped keeping secrets, and Wall Street is reading every page.
