Anthropic, the developer of the Claude family of large language models, filed its S-1 registration statement on Monday, Oct. 26, 2026, setting up a Nasdaq listing next month under the ticker ANT. According to reports, the company plans to raise $100 billion at a valuation of as much as $2 trillion, arriving on the heels of SpaceX’s record-setting debut.
Like frontier rival OpenAI, Anthropic generates revenue through two channels: subscriptions that grant a fixed level of access, and usage-based pricing measured in tokens, the model by which enterprise clients primarily consume the service. The company aims beyond a chatbot, positioning Claude as an everyday component of most knowledge workers’ workflows worldwide, and plans over the longer term to use the models to accelerate drug discovery and scientific research. Claude Code offers the clearest real-world example: a platform that plans, writes, tests and debugs software with a high degree of autonomy, now widely used across Silicon Valley and cited as a key driver of the company’s explosive growth.
For the second quarter ended June 30, Anthropic posted revenue of $11.6 billion, up from $787 million in the year-earlier period and from roughly $4.7 billion in the prior quarter. First-half 2026 revenue approached $16.3 billion, a 12.5-fold increase from the same period of 2025. The bulk of sales comes from consumption-based usage, as companies embed Claude into their own products and workflows, while consumers and businesses may also buy access through subscriptions.
Growth has not yet translated into consistent profitability. First-half 2026 net loss reached $80.7 billion, which includes a $72.4 billion non-operating, non-cash accounting charge; excluding that item, the net loss stood at $8.3 billion. The operating loss narrowed from $2.1 billion in the first half of 2025 to $693 million in the first half of 2026, and the company recorded $974 million of operating profit in the second quarter, an early sign that revenue is beginning to catch up with the cost of developing and running Claude. One profitable quarter, however, does not prove durable earnings, and spending commitments are mounting to support the company’s long-term ambitions.
Running AI requires GPU-equipped data centers, specialized real estate and infrastructure such as liquid cooling, making capital expenditure enormous. The S-1 lists named compute and GPU procurement commitments extending through 2036 totaling $637 billion, not all of them fully binding:
The filing also discloses plans to procure additional TPU compute from Google and others between 2028 and 2031, expected to cost $125 billion to $150 billion; other deals are described with fewer specifics, and actual spending could ultimately run higher.
In the third quarter ended Sept. 30, 79% of Anthropic’s revenue came from customers outside its top five, and over the nine months ended Sept. 30, no single customer accounted for more than 10% of sales. Heavy reliance on a few accounts leaves many AI companies vulnerable if one client falters. The filing lists more than 11,900 customers that each spent over $100,000 with Anthropic in the trailing 12 months as of Sept. 30, pointing to a diversified revenue base.
On the cost of revenue disclosed in the filing, Anthropic’s first-half 2026 gross margin came in at 85%, signaling considerable pricing power and standing alongside Nvidia’s gross margins running in the mid-70s percentile. Investors may question whether reported cost of revenue, drawn mainly from compute and infrastructure, fully captures the expense of keeping the models running. Models require continuous training: first-half research and development spending reached roughly $9.7 billion, covering model development, research and internal workloads. Sales and marketing totaled $3.8 billion over the same period and includes channel-partner fees paid to Amazon, Google and Microsoft; R&D and sales and marketing are excluded from cost of revenue.
The prospectus runs beyond 400 pages, with 94 dedicated to risk factors, an unusually large section for an S-1. Beyond routine business risks such as retaining talent and deploying capital effectively, more than 90 pages lay out deeper hazards tied to the AI business, ranging from potential to destabilize the geopolitical order to technology outstripping human ability to understand and control it. The filing states that failing to put proper model guardrails or safeguards in place could harm the company’s reputation, brand or financial condition, and lists models acting in unintended, misaligned or autonomous ways as a risk.
The debate around the offering ultimately hinges on whether Anthropic holds a durable moat. The view advanced in the discussion: Claude need not lose its technological lead for the business to come under pressure, because competing models only need to be good enough for the work customers actually require. The concern frames AI as moving toward commoditization, with demand potentially growing even as competition forces prices lower, while the company keeps spending heavily on next-generation models.
