Anthropic is facing growing tension with major cloud providers — Amazon Web Services, Google Cloud, and Microsoft Azure — as the company aims to acquire massive computing infrastructure but increasingly finds itself treated less as a customer and more like a competitor. Internal documents reveal a worsening ability to secure the cloud capacity essential for scaling its AI model Claude.
According to a confidential S-1 filing leaked recently, Anthropic estimates its total cloud spending could exceed $750 billion, including pending deals still under negotiation. Yet even these eye‐popping numbers may fall short as the company struggles to lock in enough support from traditional hyperscale cloud providers.
To compensate, Anthropic is building out its own data centers and teaming up with third-party data center operators. However, the company’s disclosures express concern that these alternative providers lack the reliability of established hyperscalers due to limited operational track records. It’s a trade-off between control and risk.
The heightened urgency comes amidst financial pressures. In the first half of 2026, Anthropic reported a net loss of $80.7 billion while generating $16.3 billion in revenue. Only two customers combined brought in about 20% of that revenue, showing a high customer concentration risk. Meanwhile, revenue for Q3 has already hit $16.8 billion, a 45% increase over the previous quarter, and there are now reports the company is eyeing a public offering as soon as November.
Another major risk surfaced in the filing: liability exposure from AI behavior that could cause harm and potentially exceed Anthropic’s insurance coverage. This adds a layer of uncertainty for investors concerned about how the company will manage downstream consequences of its models.
What’s driving the cloud crunch?
Cloud providers have increasingly prioritized their own AI ambitions, which puts firms like Anthropic in a precarious position. When your suppliers are also building competing models, contracts and capacity allocations can shift fast. Anthropic’s problems highlight how tight the supply chain is for GPU-scale compute — both a scarcity of resources and the shifting incentives among cloud giants.
Investors and analysts have already flagged the magnitude of compute expenses as a key growth concern for AI startups. Anthropic’s $750 billion projection underscores just how central cloud deals remain for business models in generative AI. But reliable access isn’t just about cost — it’s about strategic control, latency, and trust. Whether a third-party data center can meet performance or security expectations becomes a deciding factor.
The gamble before the IPO
With its Thursday financials showing Q3 revenue of $16.8 billion, up significantly from Q2, Anthropic’s trajectory looks bullish — but the cost base is rising even faster. The looming IPO sets the stakes even higher: investors will soon scrutinize how cloud usage, customer concentration, risk liabilities, and infrastructure investments balance out. If cloud partners continue to view Anthropic as a rival instead of a client, that could hamper capacity deals and delay or raise the costs of its public debut.
This turmoil over cloud partnerships reveals a broader tension in the AI industry. Just as compute demand surges, providers are rethinking how to allocate scarce infrastructure, especially when customers are also competitors. The performance, reliability, and cost of cloud resources are not just balance sheet concerns anymore — they’re strategic levers.
