The rapid rise of OpenAI and Anthropic, reflected in multi-billion dollar revenues and astronomical valuations, is beginning to confront a harsh reality. The further scaling of these artificial intelligence giants now directly depends on three fundamental factors: access to computing power, colossal investments in data centers, and unpredictable regulatory decisions.
An analysis of internal documents and public statements shows that OpenAI's revenue is almost linearly tied to available computing capacity. From 2023 to 2025, the company increased its capacity from 0.2 GW to approximately 1.9 GW, which synchronously raised annual recurring revenue from $2 billion to over $20 billion. The developer of ChatGPT explicitly calls computing its most scarce resource, stating that expanding infrastructure would directly accelerate product adoption and monetization. The company has already shifted from dependence on a single provider to a diversified supplier ecosystem, but net profit figures remain undisclosed.
A similar logic is evident at Anthropic. Raising $65 billion in a Series H round at a $965 billion valuation was record-breaking, but the funds are primarily directed toward expanding computing capacity for Claude. Meanwhile, the startup's annual revenue has already exceeded $47 billion, which only confirms that such growth is impossible without constantly increasing hardware capacity.
The Cost of Infrastructure: Trillions of Dollars and Uncertain Returns
Scaling advanced models has become a capital-intensive challenge for the entire supply chain. According to analysts' estimates, the four largest hyperscalers — Meta, Microsoft, Amazon, and Alphabet — will spend $5.3 trillion on capital expenditures in 2025–2030. OpenAI and Anthropic are not included in this selection, highlighting their dependence on external infrastructure.
The Bank for International Settlements, in its annual report, sounds the alarm: the five largest cloud infrastructure providers will spend over $1 trillion on AI capital expenditures in 2025–2026. These commitments already outpace the companies' profits and free cash flow, forcing them to attract debt financing. If monetization expectations are not met, a sharp reduction in funding could collapse the entire investment cycle.
Regulatory Risk: An Invisible Wall for Models
Beyond the cost of computing, AI giants face sudden restrictions from authorities. In June, Anthropic was forced to halt access to its Fable 5 and Mythos 5 models due to a directive from the U.S. government under export controls. The company noted that the authorities' letter contained no specific details about the security issue, and the complaints were related to a possible bypass of protective mechanisms.
In the same month, the Donald Trump administration asked OpenAI not to release GPT-5.6 immediately for broad access due to security concerns. The company was forced to first provide the model to a limited number of clients. This episode became part of an ongoing conflict between Anthropic and U.S. authorities, especially after the U.S. Army deployed Claude in a military operation, and the head of the Pentagon called the developer a "supply chain risk." The startup's CEO, Dario Amodei, stated that the company would prefer not to cooperate with the Pentagon rather than agree to the use of technology that could undermine democratic values.
IPO on the Horizon, but Questions Remain
Despite all the risks, both companies continue to explore opportunities for going public. OpenAI filed a confidential application on June 8, and Anthropic a week earlier. However, for investors, the key question will be not only the companies' valuation but also the economics of scaling: how much does it cost to sustain user growth, how quickly do data centers pay off, and how stable are the funding channels.
In July, Republican senators introduced a bill to protect U.S. information and communications technology supply chains, which would allow the U.S. Department of Commerce to ban transactions with technologies that pose a threat to critical infrastructure or national security. This only adds uncertainty to an already complex picture.
My analysis: The AI market is entering a phase where the winner is determined not so much by model quality as by the ability to finance and operate giant infrastructure. Regulatory risks are becoming a new "bottleneck" capable of stifling even the most promising player. Investors should closely monitor not only the reports of OpenAI and Anthropic but also the actions of governments, which can change the rules of the game at any moment.