• Amazon will invest up to $25bn in Anthropic, alongside a $100bn long-term AWS cloud commitment.
• Deal reinforces Amazon’s position at the centre of AI infrastructure and cloud computing.
What happened
Amazon will invest up to $25bn in Anthropic, deepening a partnership built around artificial intelligence infrastructure and compute supply. The investment includes $5bn upfront and up to $20bn in additional funding tied to commercial milestones. This adds to Amazon’s earlier $8bn stake.
At the same time, Anthropic has committed to spending more than $100bn over the next ten years on Amazon Web Services (AWS). The agreement locks AWS in as its core cloud provider and secures long-term access to large-scale computing capacity.
The deal also extends to Amazon’s custom AI chips, including Trainium2 and Trainium3. Anthropic expects to reach around one gigawatt of capacity by year-end and scale up to five gigawatts over time.
Amazon continues to invest heavily in AI infrastructure, with around $200bn in planned capital expenditure this year, largely focused on data centres and chips. While its own models, such as Nova, have seen limited traction, AWS remains central to global AI compute supply.
The move follows Amazon’s broader push into the AI ecosystem, including plans to invest up to $50bn in OpenAI.
Why it’s important
The deal highlights a shift in AI competition from models to infrastructure control. Access to compute has become the main constraint on AI development.
Amazon is positioning itself at the centre of this constraint. It supplies cloud infrastructure, designs custom chips and invests directly in leading AI developers. This creates a tightly integrated role across the AI value chain.
The $100bn cloud commitment effectively locks Anthropic into AWS for a decade. In return, Amazon secures predictable, high-volume demand for its infrastructure at a moment of explosive AI growth.
For Anthropic, the agreement reduces uncertainty over compute supply and supports rapid scaling of advanced models, particularly in coding and design applications.
The scale of the deal also signals rising barriers to entry. Frontier AI development now requires long-term infrastructure commitments that only a handful of firms can afford.
More broadly, cloud providers are no longer neutral infrastructure layers. They are becoming strategic gatekeepers of AI progress, shaping who can compete and at what scale.
Also read: Anthropic adds Novartis CEO Vas Narasimhan to board
Also read: Anthropic launches Project Glasswing with Nvidia and Cisco for AI security
