Anthropic, the AI start-up behind the Claude family of large language models, has laid out a financial play that makes the three major cloud providers — Amazon Web Services, Google Cloud and Microsoft Azure — central to its commercial strategy. The company’s forecast envisages at least $80 billion in cloud spending through 2029 to run and serve its models, a figure that underlines how generative AI is not just a software trend but an infrastructure business of unprecedented scale.
Beyond raw compute bills, Anthropic expects a rising share of its commercial revenues to flow to cloud partners. Its projections show revenue-share payments to cloud vendors swelling from roughly $1.3 million in 2024 to about $360 million in 2025, then to $1.9 billion in 2026 and $6.4 billion in 2027. The start‑up is offering cloud providers slices of customer sales when enterprise clients procure Anthropic’s services through those platforms — an arrangement that accelerates cloud operators’ capture of AI value beyond virtual machines and storage.
Those slices are not trivial. In AWS resale deals, Anthropic concedes roughly half of the gross margin on AI resale business to Amazon, and revenue‑share payments are expected to account for about 10 percent of Anthropic’s projected revenues in the coming years. Microsoft has already institutionalised the partnership: Azure sales staff are incentivised to promote Anthropic’s models and count related sales toward performance targets, a step that integrates the start‑up directly into cloud go‑to‑market channels.
Anthropic frames the multicloud approach as a competitive lever. By placing Claude across AWS, Google Cloud and Azure, it believes it can reach enterprise buyers more effectively than rivals that depend on a single cloud route. For the hyperscalers, that distribution is a lever to lock in customers and extract recurring economics — both from the raw compute Anthropic will buy and from the resale or referral commissions tied to enterprise customers.
The financial maths behind these arrangements expose the capital intensity of advancing large models. Anthropic projects that model‑training costs alone could reach as much as $100 billion by 2029, highlighting how next‑generation generative AI requires sustained, large-scale investment in cloud compute and specialized chips. That pressure on operating expenditure helps explain why start‑ups accept steep revenue splits with cloud providers: access to capacity, sales channels and integration with cloud platforms can be worth surrendering margin for, at least in the short to medium term.
For Amazon, Google and Microsoft this is a strategic windfall. Cloud infrastructure is convertible into a recurring revenue stream and, through resale and revenue shares, into a direct cut of expanding AI services revenues. For Anthropic and comparable model builders, it is a reminder that model innovation is only one side of the equation. Profitability will hinge on contract terms, efficiency improvements, and potentially negotiating power that may shift as scale grows or if alternative hardware and cheaper on‑prem options emerge.
The arrangement also carries broader market implications. Heavy dependence on three U.S. cloud providers concentrates economic leverage and could shape product roadmaps, pricing and enterprise procurement behaviour. It raises questions about bargaining power between AI vendors and clouds, the durability of margins for independent model makers, and whether vertically integrated players — cloud providers building their own models — will use preferentialing to tilt the market in their favour.
For investors and policy makers, the headline numbers crystallise a simple truth: the commercial success of generative AI will be decided as much in data centres and sales desks as in research labs. The next few years are likely to see intensified competition among cloud firms to capture AI workloads, more creative revenue‑share and resale models, and significant pressure on start‑ups to either scale quickly or accept long‑term partnerships that cede a portion of their economic upside.