While OpenAI and Anthropic are demonstrating impressive growth rates in revenue and market capitalization, their future is no longer determined solely by the quality of their algorithms. The key factor is becoming access to physical infrastructure: data centers, energy, and chips. An analysis of the current situation shows that scaling AI is turning into a capital-intensive logistical challenge, where regulatory and supply chain risks are becoming as important as innovation.
Computing as the New Currency
Data from recent company reports directly indicates a correlation between computing power and revenue. For OpenAI, for example, the volume of available computing grew from 0.2 GW to 1.9 GW between 2023 and 2025, while annual recurring revenue soared from $2 billion to over $20 billion. The company's CFO directly calls computing the most scarce resource. Anthropic, in turn, raised $65 billion in a Series H round at a valuation of $965 billion, stating that annual revenue exceeded $47 billion. The primary goal of these funds is to expand infrastructure for the Claude model.
The Price of Leadership: Trillion-Dollar Investments
Scaling advanced models requires enormous capital expenditures not only from developers but from the entire chain—from chip manufacturers to energy companies. According to analyst estimates, the four largest hyperscalers (Meta, Microsoft, Amazon, Alphabet) will spend $5.3 trillion on capital expenditures in 2025–2030. The Bank for International Settlements warns that the five largest cloud providers will invest over $1 trillion in AI infrastructure in 2025–2026. Moreover, such commitments often outpace current profits, forcing companies to take on debt financing. This creates a "bubble" risk: if monetization fails to meet expectations, a sharp reduction in funding could collapse the entire investment cycle.
The Regulatory Sword of Damocles
Beyond financial and technological challenges, both companies have faced sudden pressure from U.S. authorities. Anthropic was forced to suspend access to its Fable 5 and Mythos 5 models due to a government directive under export controls. The Donald Trump administration also asked OpenAI not to release GPT-5.6 to the public, citing security concerns. Although the restrictions were later lifted, these episodes demonstrate that regulatory risk has become an integral part of the infrastructure strategy. Moreover, Anthropic's conflict with the Pentagon, which labeled the developer a "supply chain risk," shows how politicized this sector has become.
IPO as a Test of Strength
Both companies have filed confidential IPO applications. However, for investors, the key question will be not so much valuation as the economics of scaling: how sustainable are funding channels, how quickly do data centers pay off, and can regulators at any moment restrict access to the flagship product.
My comment: The AI sector is entering a phase of maturity where innovation gives way to an infrastructure war. The paradox is that the faster revenue and valuations grow, the more vulnerable companies become in the face of regulators and resource scarcity. Investors should closely monitor not only earnings reports but also political statements that could overturn any financial forecasts.