#AnthropicIPOFiling:HighGrowth,HighLosses

Anthropic Filed for Its IPO, and the Numbers Behind Claude Are Almost Hard to Believe

I have been following this story since the first reports in early June, and now that the key figures from the filing have surfaced, the picture is clear. Anthropic is the five-year-old AI lab behind Claude, founded by researchers who left OpenAI after disagreements over governance and safety. Filing for an IPO means the company has formally submitted its paperwork to the SEC so it can list on a public exchange and offer shares to outside investors. That is the moment a private research lab becomes a business the whole market gets to price, and very few companies arrive at that moment with numbers like these.

The headline 2025 figures are the ones everyone is quoting. Revenue came in at $4.59 billion, up from $386 million the year before, which is growth of roughly 1,088 percent, or about 12 times, in twelve months. Operating loss widened to $8.06 billion from $2.98 billion in 2024. The GAAP net loss was $41.97 billion, but around $34 billion of that was a non-cash accounting charge tied to the rising estimated value of financing that could later convert into Anthropic shares. That slice is a paper entry, not money walking out of the building. Strip it out and the real operating loss is the $8 billion figure, and that is the number genuinely worth arguing about.

Then you look at the trajectory and the growth story gets louder. Annualized revenue run rate went from $87 million in January 2024 to $1 billion by December 2024, $9 billion at the end of 2025, $14 billion in February 2026, $19 billion in March, $30 billion in April, $47 billion in May, and $65 billion by the end of July. From the end of last year to the end of July this year that is more than seven times, and roughly $18 billion of annualized revenue added in just two months, which works out to about a 38 percent jump in 60 days. In the second quarter of 2026 alone, revenue came in above $11.5 billion against $787 million a year earlier, growth of about 1,360 percent.

The most important line in this whole story is not the revenue, it is what happened underneath it. Anthropic told shareholders its adjusted operating income was positive for a second consecutive quarter, a measure that strips out items like stock based compensation. That would make it the first frontier AI lab to post a quarterly operating profit, and the reported path runs from $4.8 billion of revenue in the first quarter of 2026 to roughly $10.9 billion in the second, a 130 percent sequential increase, with a first operating profit around $559 million. I want to be careful here, because some of those figures are projections the company shared with investors rather than audited results.

What supports that story is margin expansion, and it is dramatic. Gross margin went from roughly negative 94 percent in 2024 to somewhere between 44 and 60 percent in 2026 depending on how training costs are treated, and by September reports indicated margins above 80 percent before revenue sharing with distribution partners and before model training costs. Compute cost per dollar of revenue reportedly fell from $0.71 in the first quarter of 2026 to about $0.56 in the second. A business with an infinite burn rate is a story. A business where each dollar of revenue costs steadily less to produce is a company. The direction of that single metric tells you which one Anthropic is turning into.

The product engine behind the curve is just as impressive. Claude Code went from launch to $1 billion in annualized revenue in about six months, the fastest any enterprise software product has crossed that milestone, then $2.5 billion by February 2026 and around $8 billion by May, up from roughly $500 million in September 2025. That is a climb from $500 million to $8 billion in eight months. Enterprise use now accounts for more than half of that revenue, business subscriptions have quadrupled since the start of the year, and the number of customers spending more than $1 million annually doubled from 500 to over 1,000 in under two months.

Adoption is broad underneath that revenue. Anthropic reports more than 300,000 business customers, with 1,000 plus accounts spending over $1 million a year, and roughly 70 percent of the Fortune 100 are customers. Claude reached about 245 million monthly active users across web and mobile, with monthly active app users up around 640 percent year over year against roughly 62 percent for its closest consumer rival. The API serves over 87,000 developer organisations, up about 142 percent year over year, with more than 25 billion API calls a month and 45 percent of them coming from enterprise platforms. Claude Code alone has around 4.2 million weekly active developer users across 1,400 plus enterprise engineering organisations. Anthropic also holds roughly 40 percent of the enterprise LLM API market and around 54 percent of the AI coding market.

Now the cost side, because this is where the loss debate lives. Anthropic spent $7.33 billion on compute and infrastructure in 2025, about three times the prior year, and that single line is more than half of total operating expenses of $12.65 billion. In other words, for every $1 of revenue the company brought in, it spent roughly $2.76 running the business. It has also committed $518 billion to future cloud, compute and infrastructure obligations, with reported plans for around $80 billion of cloud infrastructure costs through 2029. Half a trillion dollars sounds terrifying until you remember what it buys, because guaranteed access to chips, data centres and power is the real moat in frontier AI.

The balance sheet is stronger than the loss column suggests. Anthropic held $20.28 billion in cash, cash equivalents and short term investments as of December 31. It has raised $118.15 billion across nine primary rounds, and its largest backers are Amazon at about 21 percent and Alphabet at about 15 percent, which means two of the biggest companies on earth are effectively long this story while also supplying the cloud it runs on. The filing itself landed just four days after the company closed a $65 billion Series H on May 28 at a $965 billion post money valuation, and the S-1 followed on June 1. Two days later it named Morgan Stanley, Goldman Sachs and JPMorgan Chase as lead underwriters. OpenAI filed on June 8 and is expected to list by early 2027.

The valuation path has moved as fast as the revenue. Anthropic was valued at $380 billion in February 2026, then $965 billion in May, a gain of about 154 percent in three months. Secondary market pricing has put it near $1.32 trillion, and the IPO is being discussed at more than $2 trillion, which would be more than double the May private mark. At $965 billion the company traded at roughly 15 times its $65 billion run rate. If full year 2026 revenue lands in the $100 billion to $120 billion range that investors are projecting, and the reported 2028 internal forecast of $190 billion to $200 billion holds, that multiple compresses quickly. That is the whole bull case in one sentence: the price looks extreme against today's revenue and reasonable against the revenue two years out.

The market backdrop matters too. The closest recent comparison is SpaceX, which listed at a $1.77 trillion valuation, priced at $135 and jumped 19 percent to $160 on day one, and still trades around $147 today. That is a useful reminder that mega listings can absolutely work, but the tape is far more selective than it was earlier this year, and AI and chip stocks have already sold off once. Anthropic's listing is expected after the November US midterm elections, and it stands to cap one of the strongest years for US IPOs since 2021.

The risks are right there in the filing and I will not pretend otherwise. Nearly a quarter of 2025 revenue came from just two customers, and many of the largest clients are not locked into long term contracts and could cut or stop spending. Cumulative net losses since the 2021 founding are estimated in the $10 billion to $15 billion range on a GAAP basis, and the company has been explicit that planned data centre spending could push profitability back negative in later quarters. Add elevated interest rates, economic uncertainty and a genuine debate about how much capital the AI trade can keep absorbing, and the caution is understandable.

What I am watching from here comes down to four things. One, the run rate heading into the fourth quarter, because if it keeps compounding at this pace the consensus full year range starts to look conservative. Two, compute and infrastructure spending as a share of revenue, since that ratio separates a growth investment from a structural problem. Three, customer concentration, specifically whether the top two customers fall below 20 percent of revenue. Four, the actual IPO price against the $965 billion May mark, because that single number will set the benchmark for every AI company queuing up behind them, OpenAI included.

One more angle for anyone reading this from the crypto side. AI equity news has become one of the strongest sentiment drivers for AI related narratives in digital assets, so a listing of this size is not only a tech story. It is a macro event that tends to spill into how the market prices anything tagged with the AI theme, which is why I am tracking the pricing date as closely as the revenue numbers.

The whole picture in one line: Anthropic's business is growing faster than almost anything we have seen in enterprise software, and its losses are high because AI infrastructure, computing power, research and global expansion all cost enormous amounts of money. Those two facts are not a contradiction, they are two sides of the same strategy, and the margin trend suggests the second one is starting to catch up with the first. The debate was never whether Anthropic can grow, because the numbers already answered that. The debate is whether the market will pay a trillion dollar plus price for a company that is only now touching profitability. I am not calling the price at this stage, but I am certainly not looking away.