Anthropic Files Prospectus: Net Loss of $80.7 Billion in First Half, Plans to Raise Up to $100 Billion

Six years ago, on October 26, Ant Group submitted its IPO filing, aiming for the world's largest IPO. Six years later, on the same day, another company planning to use "ANT" as its stock ticker also filed a prospectus with Nasdaq.

Anthropic, the AI giant behind Claude. According to market reports, the company plans to raise up to $100 billion through its IPO, with a target valuation of $1.5 trillion to $2 trillion, which could set a new global record for IPO fundraising.

But more noteworthy than the sky-high valuation is a set of sharply contrasting figures in the prospectus.

Revenue surged 30-fold over five quarters, yet the company posted a net loss of $80.7 billion in the first half of the year, and it has future compute procurement and infrastructure arrangements exceeding $600 billion in scale.

Why is an AI company that has just begun to achieve operating profitability, with hundreds of billions of dollars in cash on its books, in such a hurry to pursue the largest IPO in history?

The answer may not only concern Anthropic's valuation, but also hold the key to the business logic of the entire AI industry's next phase.

01 Revenue surged 30-fold—what makes Anthropic worth $2 trillion?

Anthropic's confidence in pursuing a $2 trillion valuation comes most directly from the simultaneous growth of revenue scale, profitability, and enterprise customer demand.

First, its revenue growth rate has far surpassed that of traditional enterprise software companies.

In 2023, Anthropic's full-year revenue was only $41 million, even less than its $69 million in interest income during the same period. But by the second quarter of 2026, its quarterly revenue had reached $11.555 billion, equivalent to nearly 282 times its full-year 2023 revenue.

Growth truly accelerated in 2025. From $376 million in the first quarter of 2025 to $11.555 billion in the second quarter of 2026, quarterly revenue grew about 30-fold in just over a year.

For traditional enterprise software companies, maintaining an annual growth rate above 30% for multiple consecutive years is already quite impressive. Yet Anthropic expanded its quarterly revenue nearly 30-fold in just over a year.

This means that frontier large models are shifting from technology R&D dependent on capital investment to a scaled commercial business capable of generating tens of billions of dollars in quarterly revenue.

Second, even more important than the revenue surge is that Anthropic is beginning to prove that the large-model business can be profitable.

In recent years, the commercialization of large models has consistently faced a challenge: the stronger the model capabilities and the larger the scale of usage, the higher the compute consumption behind it. Revenue growth may not cover cost growth, and there may even be a situation where the more users there are, the greater the losses.

Anthropic was once in the same position. In the first quarter of 2025, every $1 the company spent on compute and infrastructure costs generated only $0.40 in revenue; by the fourth quarter, that ratio was still below $0.70.

But entering 2026, the situation changed markedly.

In the first quarter, revenue corresponding to every $1 of compute and infrastructure costs rose to $1.40, and in the second quarter it further rose to $2.08. In just over a year, the output-to-input ratio per unit of compute increased more than fivefold.

This change is also reflected in the income statement. In the second quarter of 2026, Anthropic recorded an operating profit of $974 million for the first time, with an operating margin of about 8.4%.

From earning only 40 cents of revenue for every $1 of compute cost to earning $2 for every $1 spent, Anthropic is crossing the most critical threshold in the commercialization of large models: revenue growth is beginning to outpace compute cost growth.

Finally, what truly supports Anthropic's long-term valuation is the continuously expanding paid demand from enterprise customers.

Currently, more than half of Fortune 500 companies spend more than $100,000 annually on Anthropic. Across all customers, more than 1,900 enterprises spend over $1 million per year, of which about 250 spend more than $10 million annually.

This shows that Claude is no longer just an AI tool that enterprise employees try out, but is becoming a productivity product that enterprises are willing to spend real money to procure.

More noteworthy is that customers are not only willing to pay, but are also continuously increasing their investment.

Anthropic disclosed a net revenue retention rate of about 900%. Simply put, the same group of enterprise customers is spending about nine times as much on Claude this year as last year.

Behind this is the continuous expansion of AI use cases. Take Claude Code as an example: an enterprise may initially let only some programmers use it, and as actual results are validated, it gradually expands to more R&D teams and business processes, with procurement scale increasing accordingly.

And this demand still has room to grow. According to relevant disclosures, even for enterprises that have already procured Claude on a large scale, related spending on average accounts for less than 0.1% of their total expenditures.

Therefore, Anthropic's growth potential comes not only from continuously acquiring new customers, but also from existing customers continuously expanding the scale of their AI usage.

From the revenue surge to improved compute efficiency to enterprise customers continuously increasing procurement, Anthropic has already delivered a commercialization report card that supports a high valuation.

02 Behind the $80.7 billion huge loss lies an even bigger bill

Compared with the 30-fold revenue surge, what is more noteworthy in Anthropic's prospectus is the huge contrast between the $80.7 billion net loss and the hundreds of billions of dollars in long-term compute commitments.

On the surface, this looks like a company that generated $16.3 billion in revenue in half a year but lost $80.7 billion. Yet a closer breakdown of the financial data reveals that the book loss for the period and the future funding pressure are actually two different matters.

The first layer: most of the $80.7 billion huge loss is not money actually spent in operations.

In the first half of 2026, Anthropic had a net loss of $80.748 billion, of which about $72.388 billion came from fair value changes in financing instruments such as convertible notes, derivative liabilities and SAFEs, accounting for nearly 90% of the net loss.

Put simply, this loss mainly stems from accounting adjustments brought about by the rise in the company's valuation, rather than actual cash payments of an equivalent amount during the same period.

Previously, some investors provided funds to Anthropic through convertible notes and other means, which can be converted into company stock in the future under agreed conditions. As Anthropic's valuation has continued to climb, these conversion rights have become increasingly valuable, and the fair value of the corresponding liabilities has risen accordingly and been recorded in profit or loss for the period.

In February this year, Anthropic's post-money valuation was about $380 billion, and in May it rose further to $965 billion. The sharp rise in valuation also amplified the related book losses.

Thus a counterintuitive result emerged: the higher the company's valuation, the larger the book loss may be.

After excluding specific items such as fair value changes and equity incentives, Anthropic's adjusted net profit for the first half of the year was $2.671 billion.

Of course, a non-cash loss does not equal a fake loss, and the related financing instruments may still bring effects such as equity dilution. But the $80.7 billion net loss clearly cannot be directly equated with cash actually burned.

Likewise, adjusted profitability does not mean the company has achieved cash flow self-sufficiency. For an AI company like Anthropic, the real funding pressure must also be assessed in light of actual cash flow and future compute payment obligations.

The second layer: more alarming than the book loss is the hundreds of billions of dollars in future compute bills.

To secure future compute supply, Anthropic has signed massive long-term agreements with companies including Google, Amazon, Microsoft and Broadcom.

Among them, the agreement with Google involves at least about $114.5 billion, Amazon about $110 billion, Microsoft about $44.9 billion, and a five-year AI accelerator leasing arrangement related to Broadcom about $161.2 billion.

In addition, Anthropic has reached compute procurement and leasing arrangements with a number of infrastructure companies, with some contracts running as long as 20 years.

However, these amounts involve different bases such as minimum payments, lease commitments and optional procurement quotas, and cannot simply be added together, let alone all be regarded as confirmed cash payment obligations.

Moreover, some agreements even adopt a model similar to the "take-or-pay" approach in the energy industry: regardless of how much compute is actually used, the company may still need to pay the agreed minimum fees.

This means Anthropic is using long-term payment commitments to secure future compute supply. Even if AI demand growth slows, or improved model efficiency causes compute demand to fall short of expectations, some contract fees will still need to be paid.

This is also a major difference between Anthropic and traditional software companies: revenue can fluctuate with demand, but a considerable portion of compute costs has already been locked in ahead of time.

03 With hundreds of billions of dollars on the books, why is Anthropic rushing to go public?

Anthropic's choice to pursue the largest IPO in history at this time is driven not only by its ever-rising valuation, but also by the funding needs for continued future expansion.

On the one hand, the scale of financing Anthropic needs is exceeding what the traditional private market can easily absorb.

In February 2026, Anthropic completed a $30 billion Series G round at a post-money valuation of about $380 billion; just three months later, the Series H round was further expanded to $65 billion, with the post-money valuation soaring to $965 billion.

In just three months, the financing scale doubled, and the valuation more than doubled. Now, the company's IPO target valuation has reached the range of $1.5 trillion to $2 trillion.

But the larger the financing scale, the harder it becomes to raise the next round.

A single round of $65 billion is already beyond what ordinary venture capital institutions can easily absorb. More importantly, Anthropic will still need to continue investing huge amounts of capital in compute procurement, chip leasing and data center construction.

On the other hand, an IPO can open up broader and more sustained financing channels for Anthropic.

Going public not only helps the company raise a large amount of capital at once, but also establishes a public market valuation, providing more options for continued financing in the future through additional share issuances, bonds and other means.

For traditional software companies, revenue growth usually does not require a proportional increase in physical infrastructure. But frontier AI companies are different: model training, inference and the expansion of Agent applications all require continuous investment in compute.

Therefore, whoever can obtain long-term funding support will be better able to lock in chip and data center resources in advance, providing support for next-generation model R&D and commercialization expansion.

For Anthropic, the significance of the IPO is no longer just raising $100 billion, but establishing sustained financing capacity for the compute competition in the years ahead.

This is a real pressure for Anthropic.

As of the end of June 2026, Anthropic held more than $130 billion in cash and short-term investments combined.

This also precisely shows that Anthropic is not rushing to go public because it lacks money right now, but rather because the scale of future compute investment is too large and it needs to establish more stable financing channels in advance.

At the same time, AI companies' valuation expectations may also change with the market environment. Going public while investors are still willing to pay high premiums for rapid growth will help Anthropic secure more favorable financing terms.

04 From $1.5 trillion to $2 trillion, what else does Anthropic lack?

Anthropic's revenue growth and profitability have already given it the confidence to aim for a trillion-dollar valuation. But whether it is actually worth $1.5 trillion or $2 trillion, the capital market still disagrees.

The reason is simple: although Anthropic is earning more and more, whether these revenues can continue to grow and how much profit can ultimately be retained remain uncertain.

To clearly see this $500 billion valuation gap, the key still lies in three aspects.

First, cloud vendors contribute nearly half of revenue, but channel dependence may also squeeze Anthropic's profit margins.

In the first half of 2026, sales generated through the AWS, Google Cloud and Microsoft cloud marketplaces accounted for about 48% of Anthropic's total revenue, compared with only about 11% in 2023.

Relying on cloud vendors' customer resources and sales networks, Anthropic can quickly enter the large enterprise market, but it also needs to pay corresponding channel fees. In the first half of 2026 alone, related expenses increased by about $2.7 billion year on year.

This means that Anthropic's revenue growth may not necessarily translate into profit growth in the same proportion. Cloud vendors bring customers and revenue, but they also take a share of the returns.

Therefore, investors cannot simply compare Anthropic's revenue scale with that of other AI companies; they also need to look at channel costs and actual profitability.

Ultimately, revenue growth determines the market's imaginative space, but how much profit can ultimately be retained is what determines whether a high valuation can hold.

Second, enterprise customers are growing rapidly, but the stability of future revenue still needs to be tested.

In the first half of 2026, about 20% of Anthropic's revenue came from two customers, indicating that some large customers still have a significant impact on its revenue performance.

At the same time, many enterprise customers have not signed irrevocable long-term procurement contracts.

This means that even if Anthropic has entered the procurement systems of large enterprises, there is no guarantee that these customers will continue to increase purchases in the coming years.

Once customers switch to other models or begin using lower-cost open-source models, Anthropic's revenue could be affected.

Upstream costs are rigid in the long term, while downstream revenue carries volatility risk. This asymmetry is an important test for its valuation.

Third, the founding team holds more than half of the voting rights, and public investors' actual say may be limited.

Under the proposed listing plan, Anthropic's seven co-founders hold one special Class F share through Founder LLC. Although this share carries no ordinary economic rights, it controls 50.1% of the company's voting rights. In addition, a long-term interest trust also has the power to elect and remove a majority of the board.

Simply put, even if Anthropic brings in a large number of public investors after going public, key decision-making power will still mainly remain in the hands of the founding team and related governance institutions.

Anthropic designed this to protect its long-term AI safety mission and prevent the company from sacrificing technology R&D and safety goals to cater to short-term shareholder interests.

But the problem is that public investors put up money and bear operating risks, yet may not obtain voting rights corresponding to their shareholding ratio.

This may affect investors' valuation judgment of Anthropic. After all, investors care not only about how much money the company can make in the future, but also about whether they can influence major company decisions.

Therefore, although the special governance structure helps Anthropic adhere to its long-term goals, it may also become an obstacle to its pursuit of a $2 trillion high valuation.

Taken together, the valuation controversy surrounding Anthropic actually centers on three questions: how much profit can be retained from revenue, whether customers can continue to pay, and how much governance rights public shareholders can obtain.

Investors optimistic about Anthropic believe that its rapid growth and the enterprise AI market space are enough to support a high valuation; cautious investors will impose higher requirements on channel costs, revenue stability and governance risks.

This is also the real test of Anthropic's IPO: it must not only prove that AI can generate astonishing revenue, but also prove that this business can create returns for shareholders in a long-term and stable manner.

05 Conclusion: The largest IPO in history tests more than just Anthropic

Anthropic's bid for the largest IPO in history shows the outside world the astonishing commercialization potential of frontier large models, and also exposes the huge costs required to support this growth.

For Anthropic, going public is only a new starting point. What it really needs to prove is whether it can transform rapid growth into long-term stable returns while continuing to invest enormous amounts of capital.

And how much valuation the capital market is ultimately willing to pay for it will also provide an important pricing reference for the entire large-model industry.

English text is being prepared.