Anthropic has disclosed more than $10 billion in additional cloud computing commitments to Microsoft and Google, underscoring the enormous cost of securing enough computing power to develop and run its artificial intelligence models.
The company disclosed the updated figures in an amended IPO filing submitted on November 10. The agreements increase Anthropic’s minimum spending commitments with the two technology giants through 2033 as demand for its Claude AI models continues to grow.
Under the revised agreements, Anthropic has committed to spend at least $49.2 billion on Microsoft’s Azure cloud services, up from $44.9 billion previously disclosed. Its minimum commitment to Google Cloud has risen to approximately $121.6 billion, compared with $114.5 billion in the earlier filing.
Together, the increases amount to roughly $11.4 billion in additional minimum spending commitments.
These are long-term contractual obligations rather than immediate payments. The agreements require Anthropic to pay the committed amounts regardless of actual usage, subject to the contracts’ terms and exceptions.
Anthropic’s computing costs continue to rise
Training advanced AI models and serving millions of user requests require extensive computing infrastructure, including specialised chips, data centres and cloud services.
Anthropic has been expanding its access to this infrastructure as it competes in the AI market. Its cloud commitments give it access to computing capacity while allowing it to distribute workloads across different providers and hardware systems.
However, the spending also creates financial pressure. In its latest filing, Anthropic warned that short-term computing contracts are significantly more expensive than longer-term arrangements and could weigh on operating margins in the near term.
The company reported an operating loss of $567 million in the third quarter, compared with an operating profit of $974 million in the previous quarter. Anthropic attributed the change primarily to spending on more expensive short-term computing capacity.
The company’s infrastructure obligations have also expanded considerably. Its amended filing puts total committed spending on compute and infrastructure at approximately $900.9 billion as of and subsequent to September 30, 2026, although the total covers obligations extending across multiple years and includes different types of agreements.
Anthropic reassesses a separate $34.9B deal
The updated filing also addresses Anthropic’s $34.9 billion agreement with cloud infrastructure provider Lambda.
Anthropic said it has the contractual right to cancel the agreement because Lambda did not establish a data centre by an agreed deadline. However, the company said it does not currently intend to exercise that right.
The disclosure highlights a challenge facing the AI industry: securing computing capacity depends not only on financial commitments but also on providers delivering infrastructure on schedule.
Anthropic’s growing commitments to Microsoft and Google also deepen its reliance on major technology companies that supply cloud infrastructure while investing in the AI developer and distributing its models.
As Anthropic prepares for its planned public listing, investors will be watching whether revenue growth can keep pace with the cost of computing capacity. The company’s ability to manage these obligations will be an important factor in assessing its long-term profitability.
Read also: ShinyHunters stole 335 million Binance user records, demands $900,000