Anthropic’s registration statement filed with the U.S. Securities and Exchange Commission on Oct. 26 presents an unusual financial picture ahead of its planned Nasdaq listing under the ticker “ANT”: first-half 2026 revenue reached $16.286 billion, while the company reported a net loss of $80.748 billion largely driven by the accounting treatment of its financing instruments rather than operating costs.
The filing places Anthropic among the largest companies to enter public markets from the artificial intelligence sector, with revenue roughly 14 times higher than in the comparable period a year earlier. Its disclosures also show the scale of infrastructure spending needed to serve enterprise demand for generative AI models, including $8.945 billion in computing and infrastructure costs during the six-month period.
The company had not set an IPO price, share count, market capitalization or expected offering date in the filing. Market discussions cited separately have placed a possible valuation between $1.5 trillion and $2 trillion, though Anthropic did not endorse such figures in its registration statement.
Accounting revaluations drove reported net loss
Anthropic attributed approximately $72.388 billion of its first-half net loss to fair-value changes related to convertible notes, derivative liabilities and simple agreements for future equity, or SAFEs. These instruments can rise sharply in value as a private company’s implied valuation increases, creating a non-cash accounting charge for the issuing company.
That treatment produced a much larger loss on the income statement than Anthropic’s operating performance alone would suggest. The company reported a first-half operating loss of roughly $693 million, compared with an operating loss of about $2.135 billion in the prior-year period.
The filing also indicated that operating cash flow was positive during the first half. Cash flow measures the movement of cash generated and spent in the business, while net income includes non-cash revaluations and other accounting items. For a company funded through complex private-market instruments, the difference can be substantial.
Quarterly results showed that Anthropic moved into operating profitability in the second quarter. It reported operating profit of about $974 million for the three months, offsetting an operating loss in the first quarter but leaving the company with a $693 million operating loss for the half-year.
The result suggests Anthropic’s commercial operations were improving rapidly even as its headline net loss expanded. It also leaves public-market buyers with a more complicated task than simply comparing net income with revenue growth, since a large part of the reported loss depended on the valuation and structure of outstanding financing arrangements.
Usage revenue accounted for four-fifths of sales
Usage-based billing generated about $13.1 billion, or approximately 80% of Anthropic’s first-half revenue, according to the filing. Subscription revenue contributed about $3.2 billion.
That split puts customer use of Anthropic’s models ahead of fixed recurring subscriptions as the company’s largest commercial engine. Usage pricing can produce fast growth when customers expand AI deployments, but it also links revenue more directly to model consumption, customer budgets and computing availability.
Anthropic disclosed a sizable cohort of enterprise customers. As of Sept. 30, more than 11,900 customers had generated at least $100,000 in revenue over the previous 12 months, while more than 1,900 had generated at least $1 million.
Those figures indicate that revenue was not limited to a handful of large technology companies. They also show the increasing concentration of AI spending among businesses deploying models at scale for software development, internal knowledge systems, customer support and other workloads that require sustained access to computing capacity.
A separate media report cited third-quarter revenue of about $16.8 billion, up from $11.6 billion in the second quarter. If the reported $16.8 billion quarterly figure were multiplied by four, it would produce a simple annualized revenue run rate of $67.2 billion. That calculation is illustrative rather than guidance, and Anthropic’s filing did not provide a valuation based on it.
Compute costs remain the central operating constraint
Computing and infrastructure expense equaled about 55% of Anthropic’s first-half revenue. The $8.945 billion cost reflects the expense of training advanced models and running them for customers after release, including access to specialized chips, servers, data centers and related infrastructure.
Anthropic warned that capacity planning can create a timing mismatch between costs and revenue. The company may need to recognize spending on new compute capacity before customers generate enough usage to support it. Demand beyond long-term reserved capacity may also force the company to purchase short-term compute at higher unit prices.
Those risks are especially relevant for a business whose revenue is dominated by usage. A shortage of available capacity could constrain sales, while excessive commitments could weigh on margins if demand fails to arrive as expected.
Separate public reporting has put Anthropic’s multi-year cloud, compute and infrastructure commitments at at least $518 billion, while cautioning that totals differ depending on whether they include minimum contractual payments, optional purchases, capacity reservations or other arrangements. Anthropic’s filing distinguishes between binding payment obligations and arrangements that may not require the company to buy a fixed amount of capacity.
The commitments connect the company’s IPO case to a larger buildout of physical technology infrastructure. The International Energy Agency estimated that global data centers used roughly 415 terawatt-hours of electricity in 2024, or about 1.5% of worldwide electricity consumption. AI-focused capacity adds pressure to secure power, chips and construction resources alongside cloud contracts.
Competition will test whether growth can support commitments
Anthropic identified model performance, pricing, data security and product experience as factors that could influence customer decisions in enterprise AI. Those pressures could affect how much the company can charge for model usage and how efficiently it can convert demand into operating profit.
The filing provides little support for claims that capital will automatically move from cryptocurrencies or decentralized networks into AI-related assets. Anthropic’s disclosures concern corporate infrastructure contracts, enterprise software revenue and its prospective public offering, not token markets or blockchain projects.
For cryptocurrency traders, the more direct relevance lies in the infrastructure competition that Anthropic’s numbers describe. Large AI companies are increasing demand for data-center capacity, power generation and advanced computing hardware, but that does not establish a guaranteed benefit for tokens associated with decentralized computing, energy systems or hardware supply chains.
Anthropic’s public-market debut, if completed, would give public equity markets a clearer financial benchmark for a company spending heavily to turn frontier-model demand into revenue. Its filing shows a business with rapidly expanding sales, improving operating results and infrastructure commitments large enough to make cost discipline and capacity utilization central to its next stage of growth.
Explore how AI reshapes finance and trading in 2026 with Toobit Academy’s web3, AI, and crypto insights.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
