OpenAI and Anthropic are showing rapid revenue growth, but further scaling is increasingly dependent on access to computing power and data center funding. OpenAI reported that the increase in capacity from 0.2 GW to 1.9 GW increased annual revenue from $2 billion to more than $20 billion, making computing the most scarce resource in AI. Anthropic raised $65 billion in the H Series round, estimating $965 billion, to allocate funds for infrastructure expansion for Claude.

Scaling AI models is becoming a capital-intensive task for the entire supply chain, with Goldman Sachs estimating the capital costs of four major hyperscanners at $5.3 trillion for 2025-2030. The Bank for International Settlements has warned that AI infrastructure liabilities are already outpacing companies' profits, which could lead to excessive investments with uncertain returns. If expectations of monetization do not materialize, it could trigger a sharp reduction in funding and hit the entire investment cycle.

In addition to the cost of infrastructure, companies face regulatory risks: Anthropic temporarily restricted access to models due to export controls, and OpenAI received a request from the Trump administration to delay the major release of GPT-5.6. Both companies have filed confidential IPO bids, but for investors, the key issues remain scaling economics, data center profitability, and resilience to regulatory constraints.