Anthropic files for US IPO as revenue surges and losses widen

MUMBAI: Anthropic’s numbers are learning to speak fluent growth, but the bottom line is telling a more complicated story. The company behind Claude has filed for a US stock market debut after reporting a sharp jump in revenue, widening losses and hundreds of billions of dollars in long term computing commitments, putting the business of building advanced artificial intelligence under the investor microscope.

Anthropic has filed its registration statement with the US Securities and Exchange Commission (SEC), proposing to list on the Nasdaq Global Select Market under the ticker symbol ‘ANT’. The company had filed confidentially in June, allowing regulators to review its financial disclosures before the public filing. Goldman Sachs and Morgan Stanley are leading the offering alongside 20 other banks. The number of shares to be offered has not yet been disclosed.

The proposed listing arrives at a crucial moment for the AI industry. Businesses are adopting AI tools at a rapid pace, but the cost of developing, training and running increasingly powerful models remains enormous. Anthropic’s filing lays bare the tension: revenue is accelerating, yet the infrastructure needed to sustain that growth is creating financial commitments that could last decades.

For the six months ended June 2026, Anthropic reported revenue of $16.3 billion, up from $1.2 billion in the corresponding period a year earlier. Its net loss widened to $80.7 billion from $13 billion over the same period. The reported loss includes substantial non cash accounting charges linked to the company’s rising valuation, rather than representing an equivalent cash outflow.

That distinction matters. The loss figure does not mean Anthropic spent that amount in cash during the period, but it does underline why investors will need to look beyond headline revenue growth to understand the economics of the business.

The company’s expansion has been driven largely by enterprise customers using Claude for software development, workplace tasks and other business applications. Its net revenue retention rate, a measure of how much revenue a company generates from existing customers over time, stands at approximately 900 per cent, according to the filing.

Anthropic also disclosed that 79 per cent of its revenue came from customers outside its five largest accounts in the third quarter. The figure suggests that its business extends beyond a handful of major clients, although customer concentration and the durability of spending will remain important considerations as it enters public markets.

Consumption based revenue has become a major engine of growth. Revenue from customers paying according to their use of Anthropic’s models increased from $288 million, or 75 per cent of total revenue, in 2024 to $3.8 billion, or 83 per cent, in 2025. Between January and June 2026, this category generated $13.1 billion, accounting for 80 per cent of revenue.

In other words, Claude is not simply attracting users; it is generating a growing business from the volume of work customers ask its models to perform. But usage driven revenue also ties commercial growth to the cost of providing that computing power.

Computing the cost of growth

Anthropic’s infrastructure bill is one of the most consequential parts of its filing. The company spent $8.9 billion on computing power and infrastructure in the first half of 2026, an increase of 283 per cent from a year earlier.

The commitments stretch well beyond the current financial year. Since the end of 2025, Anthropic has entered into additional computing agreements with minimum obligations of approximately $417 billion. These contracts are generally expected to begin in 2026 or 2027 and may run for as long as 20 years.

That is a formidable bet on future demand. Such agreements can secure the capacity required to train and operate advanced models, but they can also leave a company with substantial bills even if customer demand fails to match expectations. Anthropic has warned that many of its contractual payments will remain due regardless of how much computing capacity it ultimately uses.

There is, however, evidence that the economics are changing as the business scales. Computing and infrastructure costs exceeded 10 times revenue in 2023, fell to 653 per cent of revenue in 2024 and 160 per cent in 2025, and stood at 55 per cent in the first half of 2026. In the second quarter, the ratio declined further to 48 per cent of revenue.

The falling ratio suggests that revenue is growing faster than infrastructure expenditure. It does not eliminate the underlying risk, however, particularly given the size and duration of the company’s contractual commitments. Investors will want to know whether improving cost ratios can continue as models become more capable and demand for computing power grows.

Debt adds another layer to the picture. Anthropic reported total indebtedness of $6.2 billion as of 30 June 2026. It also has a revolving credit facility of up to $15 billion, increased from $3.25 billion, and said it could raise additional debt to meet future financing requirements.

For a company whose products are built on computational muscle, access to capital is as important as access to customers. The IPO could strengthen Anthropic’s financial position, but the eventual terms and the market’s response will help determine how comfortably it can fund its expansion.

Safety risks move into the spotlight

The prospectus does not confine itself to financial risks. It also devotes significant attention to the potential dangers associated with advanced AI systems, including behaviour that is unintended, misaligned or autonomous.

Anthropic warns that such systems could attempt to resist shutdown or conceal and manipulate information. Without adequate safeguards and oversight, it cautions, advanced AI could pose catastrophic or existential risks to humanity.

These warnings carry particular weight for a company whose commercial proposition depends on developing increasingly capable models. AI safety is not merely a question of public positioning; it is also a potential source of operational, legal, regulatory and reputational exposure.

Anthropic further cautions that its public positions on AI regulation and safety, along with its engagement with policymakers, could attract heightened scrutiny, political backlash or disproportionate regulatory action. The company’s own stance on how AI should be developed and deployed could therefore become a business risk as well as part of its identity.

That tension is already visible in its dealings with the US government. The Trump administration has targeted Anthropic after it refused to lift certain restrictions on the Pentagon’s use of its technology. The US Department of Defense subsequently designated the company a national security risk and barred it as a supplier.

The dispute illustrates how AI companies can find themselves navigating more than competition for customers. Decisions about how their systems may be used can bring them into conflict with governments and potentially affect their access to contracts, markets and public sector business.

A public listing will bring additional scrutiny to how Anthropic manages these competing pressures. Investors will have to assess not only how quickly its technology is being adopted, but also how its safety commitments and policy positions could affect future growth.

Founders retain a firm grip

Anthropic’s governance structure is another factor prospective shareholders will need to consider. The company is registered as a public benefit corporation, meaning it must weigh its stated objective of developing AI safely alongside its financial obligations to shareholders.

Its seven co founders, led by chief executive Dario Amodei and president Daniela Amodei, are expected to retain just over half of shareholder voting rights after the listing. That would limit the ability of outside investors to outvote the founders on corporate decisions, even as the company opens its ownership to public market participants.

The co founders have pledged to donate 80 per cent of their shares to charity. The filing, however, states that these commitments are not legally binding. The distinction is important for investors assessing how much weight to place on the pledge when evaluating the company’s long term ownership structure.

Anthropic was founded in 2021 by Dario Amodei and a group of researchers who left OpenAI. Its growth has accelerated in 2026, supported by demand for Claude Code and its broader AI models as businesses increasingly use AI for software development and internal operations.

The IPO is expected to be among the most closely watched offerings associated with the AI industry. Reports cited in the supplied information suggest that Anthropic is targeting a potential offering of up to $100 billion at a valuation of $1.5 trillion to $2 trillion. These remain reported targets, however, rather than final terms disclosed in the registration statement. The number of shares, final valuation and timing of the offering have yet to be confirmed.

The proposed listing would bring one of the leading developers of advanced AI models into the public markets ahead of rival OpenAI, which has reportedly pushed its own market debut to 2027. Following SpaceX’s record setting IPO earlier this year, Anthropic’s offering would also test investor appetite for companies whose growth prospects depend on large, long duration infrastructure investments.

For investors, the central question is whether Anthropic can turn its extraordinary demand into durable financial returns without allowing the cost of computing, financing requirements and safety risks to overwhelm the business. Its filing offers a detailed look at both sides of the AI boom: a rapidly expanding enterprise market and the expensive, uncertain commitments required to keep it running.

The final valuation and offering terms will determine how investors price that balance. For now, Anthropic has put its ambitions on the record, but the market will decide what those ambitions are worth.