Anthropic Publicly Files IPO Prospectus: First-Half Revenue of $16.286 Billion, Net Loss of $80.748 Billion

Anthropic, the company built on its Claude series of large models, has for the first time laid the financial groundwork of the AI arms race before the public.

At 9 a.m. Eastern Time on October 26, the five-year-old company publicly filed its S-1 registration statement with the U.S. Securities and Exchange Commission. In the first half of 2026 it generated revenue of $16.286 billion, up from $1.163 billion in the same period a year earlier, while recording a net loss of $80.748 billion, the vast majority of which—$72.39 billion—consists of non-cash accounting charges stemming from the rise in fair value of the convertible notes and other financing instruments it issued to fund its expansion.

The roadshow and bookbuilding are expected to launch within weeks, with a listing falling as early as November. At that point, this fastest-growing AI company will, carrying a scale of capital consumption Wall Street has never seen, directly face the test of a valuation of roughly $2 trillion.

The core question raised by the filing is not whether Anthropic is growing—it is, and so fast that even veteran tech investors are surprised—but whether the market is willing to pay $2 trillion for that growth: growth that depends on future compute commitments running into the hundreds of billions of dollars, on three hyperscale cloud partners that increasingly view it as a competitor, and on two major customers that together account for about one-fifth of revenue.

The answer to that question will set the benchmark for how every AI company that lists thereafter is priced. And rivals are already racing ahead: according to media reports, OpenAI is seeking to complete a funding round at a valuation of more than $1.4 trillion before Anthropic is priced—whoever prices first sets the standard.

An income statement no one predicted

A net loss of $80.7 billion over six months, on paper, describes a company consuming capital without restraint—but about $72.4 billion of it, roughly 89 cents of every dollar of loss, is an accounting artifact rather than a cash outflow.

The bulk of that $70-odd billion comes from an accounting line item that involves no cash outflow. Anthropic earlier issued a large volume of convertible notes, which can in the future be converted into stock under agreed conditions; the higher the company's valuation, the more valuable they become. Accounting rules require that each time the valuation is marked up, the company record that increase as an expense for the period.

And according to media reports, in this year's mid-year Series H round, Anthropic raised $65 billion in one go at a post-money valuation of about $965 billion. In other words, the more valuable the company becomes, the larger the loss on paper. That $80.7 billion is more like the shadow cast on the statements by a surging valuation, not a flow of real money. In the same period a year earlier, the figure was $12.97 billion.

Breaking down the statement, the real operating picture is milder than the paper loss, but still substantial in scale.

In the first half of 2026, Anthropic spent $8.945 billion on compute and infrastructure, exceeding the $7.326 billion spent in all of 2025 and nearly four times the $2.334 billion spent in the first half of 2025; relative to revenue of $16.286 billion, the compute bill had already consumed about 55% of sales before any spending on research and development, sales, or general and administrative costs.

At the same time, sales, marketing and partnership spending surged from $120 million in the same period a year earlier to $3.168 billion—growth was bought. The operating loss itself has narrowed from $2.135 billion to $693 million; excluding non-cash items on the company's disclosed basis, adjusted operating profit has swung from a loss in the same period a year earlier to a profit of about $1.6 billion—one of the few bright spots in the entire prospectus.

Growth investors are willing to accept losses precisely because of how long revenue can keep growing at a high rate—and that is where things get interesting. According to media reports, the company disclosed to potential investors that third-quarter 2026 revenue grew 45% quarter over quarter to about $16.8 billion, meaning a single quarter's results nearly matched the entire first half; but the 45% quarter-over-quarter growth rate is itself already slowing—second-quarter revenue more than doubled from the first quarter to about $11.6 billion.

Some investors who have reviewed the disclosure documents are instead considering a lower valuation. At the company's investor day, attendees discussed a valuation close to $1.5 trillion, a marked pullback from the roughly $2 trillion previously expected for the November stock sale.

A compute bet in the hundreds of billions

If the income statement is shocking, the commitments on the balance sheet are the real story.

The S-1 discloses that since the end of 2025, leases and commercial agreements newly signed and amended by Anthropic involve about $417 billion in future minimum payment obligations, mainly leases of compute equipment and data centers and purchase commitments to certain cloud service providers, taking effect between 2026 and 2027 with terms ranging from less than one year to 20 years; there are also agreements whose terms have not yet been finalized, expected to involve about $125 billion to $150 billion.

The two together total roughly $540 billion to $570 billion.

According to reports, the market sums up this web of commitments as "as much as $750 billion in compute investment," of which about $450 billion is already contracted and mostly non-cancelable, broken down as roughly $111 billion with Google, about $110 billion with Amazon Web Services, and about $31 billion with Microsoft.

These are not optional purchases. Most of the contracted portion must be paid regardless of usage, which means Anthropic has effectively converted future revenue into fixed obligations before that revenue is secured. The company is betting that demand for Claude—especially Claude Code, the coding assistant that has become a darling of developers—will grow fast enough to feed these GPUs. If demand weakens, the commitments will not shrink; the bills will be paid regardless of revenue, amounting to a senior claim on every dollar of revenue.

To finance this expansion, Anthropic is increasingly relying on debt. According to media reports, in October 2026 a large credit facility led by Blackstone closed, with the company offering lenders a yield of more than 10% on chip-financing debt—a rate that indicates deep-pocketed private capital still has appetite for AI infrastructure exposure, but is demanding a considerable risk premium for it.

The deal follows Google's April announcement of an investment plan of up to $40 billion and Amazon's additional $5 billion commitment. In the S-1, the company describes its infrastructure partnerships as providing it with the compute to keep pace. That sentence conveys both confidence in expansion and a passive predicament: capacity must be purchased in advance of demand.

Relying on partners it views as competitors

The filing is unusually candid here. The risk factors section states that Anthropic depends on a limited number of cloud providers (Amazon, Google, Microsoft) and suppliers such as Broadcom for compute and key components, and that these suppliers have strategic and investment relationships with the company and are also its customers, creating complex conflicts of interest that could affect its ability to obtain compute.

This is the structural contradiction at the heart of the AI buildout boom: the companies selling Anthropic the chips and cloud capacity needed to train and run its models are also racing to develop competing models of their own. Google has Gemini, Amazon has embedded its own models throughout AWS, and Microsoft is deeply tied to OpenAI.

This dependence runs both ways, and it makes Anthropic unlike any recent large IPO candidate. According to third-party estimates, Amazon holds 15% to 20% of the company and Alphabet holds 10% to 15%; at a $2 trillion IPO valuation, Amazon's stake would be worth between $300 billion and $400 billion, and Alphabet's between $200 billion and $300 billion.

According to Amazon's own financial disclosures, Amazon alone recorded $16.8 billion in pre-tax gains from its Anthropic stake in the first quarter of 2026. This windfall means the hyperscalers simultaneously play the roles of landlord, shareholder and competitor. A landlord that also competes with its tenant has little incentive to keep rents low or capacity abundant, and the S-1's own warning about constrained compute access shows Anthropic is well aware of this.

Concentration risk on the revenue side is likewise written into the document. In the first half of 2026, two major customers together accounted for about 20% of revenue, and the S-1 warns that significant reliance on a limited group of customers may persist; many large customers have not signed long-term contracts and can reduce or stop purchasing at any time according to their own needs. In 2025, about 47% of sales were made through the Amazon and Google cloud marketplaces, and in the first half of 2026 that share rose further to 48%, with growth coming through others' storefronts accounting for a rising share.

The channel does accelerate customer onboarding, and it also deepens dependence, as the composition of accounts receivable confirms: receivables generated through CSP marketplace channels rose from 42% of total receivables at the end of 2024 to 60% at the end of 2025.

In the week the document was disclosed, the market was stress-testing the limits of AI leverage, from capex scrutiny across the industry to the private credit market on which Anthropic itself depends. According to market data, as of Thursday, October 22, the Nasdaq Composite fell 2.58% in a single day, its largest one-day drop since June, led by chipmakers Nvidia and AMD, as investors reassessed the prices they are paying for AI assets. The warning sign is that the capital market supplying the blood for Anthropic's hundreds of billions of dollars in compute commitments is also the market that will sell hardest when returns fall short of expectations.

At a $2 trillion valuation against 2025 revenue of $4.585 billion, investors are paying a multiple of about 435 times trailing 12-month revenue. Even annualizing third-quarter revenue of about $16.8 billion—which is only a linear extrapolation, not a forecast—implies revenue of about $67.2 billion and a forward multiple of nearly 30 times. According to media reports, SpaceX, which listed in June 2026, traded at about 26 times revenue. This premium is buying a winner-take-all outcome in frontier AI, with almost no margin for error.

More noteworthy is the second implication of these numbers. The conventional reading is that Anthropic's losses prove AI is a capital black hole; the deeper reading is that these losses are a transfer: Anthropic is converting hundreds of billions of dollars of investor capital into demand for Nvidia chips, Broadcom custom chips, Google TPUs and the broader data center ecosystem.

Hundreds of billions of dollars in compute commitments are not only a risk for Anthropic; they also provide a floor of revenue support for the entire AI infrastructure complex. If Anthropic stumbles, the pain will not stop at its own shareholders—it will transmit to every company that sells it shovels and picks. AI infrastructure has turned a model company into a demand engine for the semiconductor cycle.

Cyclical hype or structural shift

Is this a bubble about to burst, or a genuine paradigm shift? The answer is both, and the distinction matters.

Valuation multiples are cyclical; they are the product of frenzied capital chasing scarce assets, and they will fall back if growth slows or a competitor achieves a decisive leap in capability. The dot-com bubble era produced companies trading at hundreds of times revenue, most of which eventually collapsed, but a handful grew into their multiples—Amazon foremost among them—and then dominated their respective industries for decades.

The structural shift, however, is real and has nothing to do with valuation multiples. Artificial intelligence is not a product cycle; it is a general-purpose technology being embedded into every layer of the economy, and the company controlling the frontier models sits at the chokepoint of this transformation. Anthropic's hundreds of billions of dollars in compute commitments are not isolated irrational exuberance; they are the entry ticket to a market where capacity, not code, becomes the scarce resource. Hyperscalers understand this well, which is why they are simultaneously Anthropic's suppliers, shareholders and competitors. This configuration is unprecedented, and it will not mean-revert.

The strongest argument against buying this IPO is straightforward and deserves full weight: Anthropic is committing to fixed costs that are predicated on revenue continuing to grow at a pace no public company has ever sustained, while its revenue depends on customers who can leave at any time and partners with ample incentive to squeeze it.

If revenue growth slows further from its current pace, if the 45% quarter-over-quarter growth in the third quarter proves to be a peak rather than a brief pause, the company will face a two-front squeeze: non-cancelable bills coming due on one side, and hyperscale cloud partners holding competing models on the other, happy to take the customers Anthropic can no longer serve. In that scenario, the $72.4 billion book adjustment is the least of its worries, because operating losses—cash losses—will become the binding constraint.

And the discussion around a $1.5 trillion valuation is an early signal: confidence is already wavering among those funding this expansion.

If Anthropic's quarter-over-quarter revenue growth falls below 30% for two consecutive quarters, or if it renegotiates or abandons a substantial portion of its non-cancelable compute commitments, the structural growth thesis is broken. Either outcome would show that demand has failed to keep pace with the fixed cost base, at which point the valuation multiple would fall back and converge toward infrastructure-company levels rather than sustaining a platform-winner premium.

What happens next

In the near term, the IPO's pricing will set the tone. A successful listing near $2 trillion would validate the frontier AI premium and pave the way for OpenAI's fundraising; a weak market reception, or a down round toward $1.5 trillion, would force a repricing across the entire private AI ecosystem, and the IPO window for other loss-making unicorns could close with it.

In the medium term, watch three indicators: quarter-over-quarter revenue growth (whether the 45% pace holds or continues to slow), cash burn excluding accounting charges (whether operating losses as a share of revenue narrow), and customer concentration (whether Anthropic can expand its customer base beyond its top two customers). These indicators will tell investors whether the company is maturing or merely scaling up the size of its commitments.

Governance is the fourth thing to watch: the founder holds 50.1% of voting power through Class F common stock held via Founder LLC, and a long-term interest trust has the right to elect and remove a majority of the board. What public shareholders are buying is a growth certificate without majority voting rights.

In the long term, the question is whether Anthropic can convert its model-quality lead into a durable moat before its hyperscale cloud partners close the gap. If it can, today's losses will be seen as the cost of building a great company; if it cannot, these compute commitments will become the most expensive lease in corporate history.

The real takeaway from this filing is that the funding source for the AI race has changed. It is no longer venture capital, but growth-hungry public investors, debt markets demanding double-digit returns, and compute contracts that lock in a decade of spending before the revenue arrives. What Anthropic has filed is not just a listing application but a quote sheet asking the market to pay in advance for the prospect that artificial intelligence becomes core economic infrastructure.