Anthropic has filed to go public in the US. Its IPO filing offers a closer look at its exponential revenue growth, mounting infrastructure costs, potential losses and concerns over the safety of its AI systems.
It plans to list its Class A common stock on the Nasdaq Global Select Market under the ticker symbol ANT.
The company reported approximately $16.3 billion in revenue for the first half of 2026. It grew 1,368% year-on-year for the three months ending June 30, 2026. Its net loss widened to $80.7 billion, largely due to a $72.4 billion non-operating accounting charge related to the fair value of bonds convertible into its shares.
Its operating loss for the first half of 2026 stood at $693 million. However, Anthropic cautioned that its growth rate could decline and fluctuate as its business matures. In Q2, the startup reported $11.6 billion in revenue and $974 million in operating profit.
That said, Anthropic’s IPO filing does not disclose its third-quarter 2026 revenue or October’s annualised revenue run rate. The omission comes as growth slows, with Reuters reporting that the company’s revenue growth decelerated for the first time in the latest quarter.
Anthropic has raised an aggregate $141.1 billion in capital since inception. As of June 30, 2026, it held $112.4 billion in cash and cash equivalents plus $19.3 billion in short-term investments.
The company noted that its top five customers accounted for about 25% of revenue in the nine months ended September 30. More than 11,900 customers spent at least $100,000 over the trailing twelve months, more than 1,900 spent $1 million or more, and roughly 250 spent $10 million or more.
Arun Chandrasekaran, Distinguished VP Analyst at Gartner, says Anthropic’s financial performance suggests frontier AI can become a high-growth business at a scale never seen before. That said, he notes that the company’s long-term economics remain unproven. “Its enormous infrastructure commitments create substantial financial exposure, making revenue growth per dollar of compute, rather than revenue growth alone, the important measure of long-term sustainability,” he says.
Chandrasekaran adds that Anthropic’s growth shows enterprises are moving beyond experimentation into production deployments, particularly in coding and increasingly agentic workflows. “Its approximately 900% net revenue retention indicates extraordinary expansion among existing customers, although most of this growth is still concentrated in its coding capabilities,” he says.
Heavy Compute Commitments
Anthropic’s compute and infrastructure costs rose from $944 million in Q1 2025 to $5.6 billion in Q2 2026, driven by multi-year cloud computing agreements.
Anthropic’s filing puts its total compute and infrastructure commitments at roughly $637 billion. The company disclosed substantial long-term computing infrastructure obligations. Deals with Google are valued at at least $114.5 billion, Amazon at least $110 billion, and Microsoft at at least $44.9 billion, along with equipment leases with Broadcom worth about $161.2 billion.
Future obligations related to data centres, equipment, and cloud services run into the hundreds of billions of dollars, much of it non-cancelable.
“AI is now a capital race, and whoever can secure the most chips, power and data centres trains the best models. Very few companies on earth can write checks at this scale, and fewer still have the three biggest cloud and chip players lined up as backers. Anthropic has both the growth and the money to keep pressing its lead,” wrote Nicholas Mugali, CEO and Principal of World Trade Securities, in a post on X.
He notes Anthropic has $131.7 billion in cash and investments, alongside another $77 billion in committed funding from Google, Amazon and Broadcom. “That’s $209bn before a single IPO share is even sold,” he quips.
Compute Efficiency
Anthropic is generating more revenue from every dollar it spends on computing infrastructure. It generated $2.08 in revenue for every $1 spent on compute in the second quarter of 2026, up from just $0.40 in the first quarter of 2025, according to its IPO document.
Dharmesh Thakker, General Partner at Battery Ventures, said, the hyperscalers turn megawatts into GPU hours and sell them on long-term contracts. “That locks in revenue, but it also caps it,” he said. “The labs take on the commitments and keep the value they create, turning those hours into high-value tokens.”
Clark Tang, an Analyst at Altimeter Capital, says that the company’s key challenge is whether “revenues generated outpace their fixed compute commitments”. Under-provisioning compute could leave Anthropic struggling to meet demand, while committing too much capacity could put pressure on its finances if revenue growth slows.
However, improving compute efficiency does not guarantee higher profits. Tang cautions that the economics depend on factors such as competitive pricing, customer demand and the computing requirements of more capable models. “There are real risks that the revenues generated may slow down relative to the growth of the fixed cost base,” he wrote on X.
Chandrasekaran says that higher consumption does not necessarily translate into a willingness to pay premium prices, adding that it remains unclear whether customers will continue choosing Claude if competing models offer comparable performance at lower costs.
Using Claude to Build its own AI
Anthropic is also relying on Claude to accelerate its own research and software development. More than 90% of contributions to its codebase in September 2026 were written by Claude, according to the filing. The average daily volume of code contributed per engineer in the three months ended September 30 was more than 45 times the average recorded in 2024, as engineers increasingly directed and reviewed AI-generated work rather than writing code line by line.
The company said the value of its internal Claude usage, calculated at customer list prices, increased more than 250 times between January and September 2026. In April, that monthly value exceeded its monthly personnel expenses, excluding stock-based compensation.
Existential AI risks
In its filing, Anthropic warns that increasingly capable AI systems could behave in unintended or misaligned ways, including taking harmful actions without being directed to do so. The company states that investing in its stock carries “a high degree of risk.”
It warns that advanced AI systems “could pose catastrophic or existential risks to humanity,” potentially enabling new bioweapons, destabilising the geopolitical order, or evolving beyond the ability of “humans to understand and control.”
Chandrasekaran believes Anthropic’s focus on safety could strengthen enterprise trust and set it apart from competitors, but cautions that these warnings also highlight the tension between commercial growth and responsible AI deployment.
Anthropic’s public benefit corporation structure seeks to balance shareholder interests with its mission to develop AI responsibly. Its Long-Term Benefit Trust, comprising five trustees, can elect and remove directors who collectively hold a majority of board seats.
The founders will also retain substantial voting power. A single Class F share held through an entity formed by the co-founders carries 50.1% of the voting power on matters submitted to shareholders generally. Meanwhile, Anthropic has also established a board-level safety and risk committee to oversee risks associated with advanced AI.
