Anthropic's preparations for a potential public listing are providing an unusually detailed look at the economics behind the development of frontier artificial intelligence.
A confidential IPO prospectus reviewed by Reuters shows a company growing at extraordinary speed, with revenue approaching $4.6 billion in 2025, but also reporting an operating loss of more than $8 billion as spending on computing infrastructure surged. The same document describes approximately $518 billion in future commitments for cloud computing, infrastructure and related resources.
The figures illustrate one of the central characteristics of the current AI industry: building increasingly capable models requires enormous amounts of capital, computing capacity and infrastructure long before the resulting technology necessarily generates comparable operating profits.
Anthropic's potential IPO would therefore be significant not only because of the size of the company or the valuation investors may eventually assign to it, but because it could offer public markets one of the clearest views yet into the financial structure of a frontier AI laboratory.
From AI Startup to Global Infrastructure Consumer
Founded in 2021, Anthropic has developed into one of the leading companies building large-scale AI models through its Claude family of systems.
Its growth has been exceptionally rapid. According to Reuters, Anthropic's revenue increased roughly twelvefold in 2025 to nearly $4.6 billion. At the same time, however, its operating expenses approached $13 billion, with more than half of that spending associated with computing and infrastructure.
The numbers demonstrate the unusual economics of frontier AI.
Traditional software businesses can often scale additional users at relatively low marginal cost once a product has been developed. Frontier AI companies operate differently. Training increasingly sophisticated models requires large quantities of specialized computing resources, while serving those models to millions of users also requires continuous access to data-center capacity.
The result is a business in which revenue can grow extremely quickly while infrastructure costs grow alongside it.
Anthropic's financial disclosures provide an unusually clear illustration of this relationship.
The $518 Billion Commitment
The most striking figure in the prospectus is the approximately $518 billion in future commitments related to cloud, computing and infrastructure.
The number should not be interpreted as an immediate cash expenditure or as a single investment program. Rather, it represents contractual and planned commitments extending into the future as Anthropic secures the computing capacity required to operate and expand its AI systems. Reuters reported that the company had entered into major infrastructure arrangements with technology and cloud providers as it seeks to secure the resources necessary for future growth.
Nevertheless, the scale is significant. It shows that frontier AI development is becoming closely connected to long-term infrastructure contracts. The companies building the models increasingly need to secure computing capacity years in advance, while cloud and infrastructure providers must invest billions of dollars to supply that demand.
This creates a circular capital ecosystem:
- AI companies require computing power.
- Cloud providers build data centers and purchase accelerators to provide it.
- Semiconductor companies supply the processors and networking equipment.
- Data-center operators require electricity, cooling systems and physical infrastructure.
- Capital markets finance much of this expansion.
Anthropic sits near the center of this ecosystem as one of the largest consumers of AI computing resources.
A $42 Billion Net Loss With a Different Underlying Picture
Anthropic's headline financial numbers are also unusual.
Reuters reported that the company recorded a $42 billion net loss in 2025, but much of that figure was related to accounting treatment of financial instruments that could potentially convert into shares. Approximately $34 billion of the loss came from this accounting charge, leaving an operating loss of roughly $8 billion.
This distinction matters when examining the company's underlying operations. Anthropic was not simply generating $4.6 billion of revenue and then spending $42 billion of cash on its operations. The reported net loss incorporates accounting effects associated with financing instruments, while the operating loss more directly reflects the economics of running the business.
The operating figures nevertheless remain substantial. An operating loss of more than $8 billion against approximately $4.6 billion of 2025 revenue demonstrates how expensive frontier AI development was during that period.
At the same time, the company has continued to grow rapidly. TechCrunch reported that Anthropic's annualized revenue run rate had reached approximately $65 billion by the end of July 2026, based on reporting from Bloomberg, compared with $47 billion in May and approximately $9 billion at the end of 2025.
These figures are run-rate estimates rather than reported annual revenue, so they should not be confused with audited full-year results. They nevertheless demonstrate the speed at which the business has been expanding.
Amazon and Google Are Part of the Equation
Anthropic's financial structure also illustrates the increasingly complicated relationship between AI laboratories and the technology companies supplying their infrastructure.
Amazon and Google are among Anthropic's major strategic partners and investors, while both companies also operate their own cloud platforms and AI businesses. Amazon has made substantial investments in Anthropic and has positioned AWS as a major infrastructure provider for Claude. Google has also provided Anthropic with significant computing resources and investment.
This relationship creates both cooperation and competition. Cloud providers benefit when AI companies consume large amounts of computing capacity. At the same time, the largest cloud companies are developing their own AI models, accelerators and software platforms.
Anthropic therefore depends on companies that can simultaneously act as suppliers, investors, distribution partners and competitors.
The prospectus highlights this structural feature of the AI industry. Reuters reported that Anthropic pays approximately 16% of its revenue to cloud partners, illustrating how significant infrastructure providers are to the economics of the company. The relationship also extends beyond Amazon and Google. Anthropic has been assembling multiple sources of computing capacity as it attempts to secure enough infrastructure to support future model development and inference.
Why Frontier AI Requires So Much Capital
The economics of Anthropic provide a broader explanation for why AI companies are raising increasingly large amounts of capital.
Training a frontier model can require enormous quantities of GPUs or other accelerators, high-bandwidth memory, networking equipment, storage and electricity. But training is only one part of the equation.
Once a model is deployed, every user interaction consumes computing resources. As AI applications become more sophisticated, models can require significantly more computation to produce an answer, particularly when systems perform multi-step reasoning, use external tools or operate autonomously.
This creates two separate infrastructure requirements:
| Requirement | Description |
|---|---|
| Training capacity | Computing resources required to develop and refine frontier models. |
| Inference capacity | Ongoing computing resources consumed each time a deployed model generates an output. |
The infrastructure must also be continuously upgraded. New generations of AI accelerators can deliver better performance and efficiency, encouraging companies to replace or expand existing computing clusters. Data centers need additional power and cooling, while networking systems must handle enormous volumes of data moving between processors and storage.
Anthropic's commitments therefore reflect a wider transformation in the technology industry. AI companies are increasingly becoming major infrastructure customers rather than conventional software startups.
The IPO Question
Anthropic's decision to prepare for a public listing is closely connected to these capital requirements.
Co-founder Daniela Amodei previously explained to TechCrunch that developing frontier models involves very large upfront costs for training and inference and argued that companies working at the frontier would eventually need access to substantial amounts of capital.
That provides an important context for the IPO process. A public listing could give Anthropic access to a much broader pool of capital than private markets alone. It could also provide public investors with exposure to the financial development of one of the world's leading AI model companies.
However, the prospectus also shows why the economics are difficult. A rapidly growing AI company can generate billions of dollars in revenue while simultaneously committing hundreds of billions of dollars to future infrastructure.
The central financial question is therefore not simply whether AI demand is growing, but whether revenues and operating economics can eventually scale sufficiently to support the enormous infrastructure commitments required to maintain leadership. That question extends well beyond Anthropic.
A Different Model of Technology Investment
The Anthropic case highlights how different the current AI investment cycle is from earlier technology waves.
The early internet created relatively capital-light software businesses after telecommunications infrastructure had already been built. Cloud computing changed the model by moving computing resources into centralized data centers, but the largest infrastructure investments were still concentrated among cloud providers.
Frontier AI combines elements of both. The AI laboratory develops software and models, but its products depend on an enormous physical infrastructure layer underneath them. This makes the economics of the sector closely connected to semiconductors, data centers, electricity, networking and cloud computing.
The result is an ecosystem in which capital flows through multiple layers. Investors finance AI companies; AI companies purchase cloud capacity; cloud companies purchase chips and build data centers; semiconductor companies invest in manufacturing capacity; and utilities and infrastructure operators expand electricity supply.
Anthropic's $518 billion figure illustrates the scale that this ecosystem can reach.
What the Prospectus Could Mean for the AI Industry
Anthropic's potential IPO will be closely watched because it could provide public markets with a new benchmark for valuing frontier AI companies.
The company was privately valued at approximately $965 billion after a $65 billion funding round announced in May 2026, according to TechCrunch. Reports based on the IPO prospectus have suggested that Anthropic could seek a valuation above $2 trillion, although the eventual valuation, offering structure and timing remain subject to the IPO process.
The more important issue for the technology industry, however, is the financial model behind those valuations. Anthropic's numbers show a company capable of extraordinary revenue growth but operating within an infrastructure model that requires enormous amounts of capital.
That tension is likely to remain central to the AI economy. If AI adoption continues expanding, demand for computing, data centers, networking equipment and electricity could remain a major driver of investment across the technology infrastructure sector. If the economics of AI applications develop more slowly than expected, however, the enormous commitments made to infrastructure could become a significant financial consideration for companies throughout the ecosystem.
For now, Anthropic's prospectus offers an unusually detailed snapshot of the scale of the frontier AI race.
The company is no longer simply building an AI model and selling access to software. It is participating in a capital-intensive infrastructure ecosystem that spans cloud computing, semiconductors, data centers and energy.
Its potential IPO could therefore become an important test of how public markets value a new type of technology company: one capable of generating billions of dollars in rapidly growing revenue, while simultaneously requiring extraordinary amounts of capital to build the computing infrastructure needed for its next generation of products.
That is perhaps the most important message contained in Anthropic's financial disclosures. The AI revolution may be delivered through software, but its economics increasingly depend on the physical infrastructure required to make that software possible.
