Anthropic is best known for its AI models, but it’s looking to partner with companies at all levels of the AI stack.

The company has announced a strategic agreement with Micron Technology that covers memory and storage architecture design, a supply deal, enterprise AI adoption, and a direct investment in Anthropic’s Series H funding round. It’s a broad agreement, and the shape of it reflects how seriously Anthropic is thinking about the compute constraints that have dogged it through its period of explosive growth.

The technical centerpiece of the deal is a collaboration on how memory and storage subsystems perform across AI workloads. Micron brings high-bandwidth memory (HBM), DRAM, and SSDs to the table — the components that determine how quickly data moves through an AI system during both training and inference. The two companies will analyze how these subsystems interact across Anthropic’s full infrastructure stack, with the goal of improving performance, reducing energy consumption, and improving what the announcement calls “token economics” — the cost per output that ultimately determines how profitably Anthropic can serve Claude at scale.

Tom Brown, Anthropic’s co-founder and chief compute officer, was direct about the motivation: memory and storage are central to how efficiently Claude can be trained and served, and getting those layers right matters as much as the model architecture itself. That’s a meaningful statement from a frontier AI lab, which typically focuses its public commentary on model capabilities rather than the hardware beneath them.

Alongside the technical work, Micron and Anthropic have signed a multi-year supply agreement covering Micron’s data center portfolio. For Anthropic, which saw 80x annualized demand growth in a single quarter and has had to scramble for compute ever since, locking in supply relationships at this level is as much about operational security as it is about performance optimization.

The relationship runs in both directions. Micron has already deployed Claude models internally, using them for coding acceleration and agentic workflows across engineering, manufacturing, and enterprise functions. The company says these deployments are delivering measurable productivity gains on some of its most complex operational challenges.

Micron is also putting money in alongside the collaboration, participating in Anthropic’s Series H round as a strategic investor. Samsung and SK hynix are also part of that round as infrastructure partners — a sign that the leading memory manufacturers see alignment with frontier AI labs as a strategic priority worth backing financially, not just commercially.

For Anthropic, the Micron deal fits a pattern. The company has been building out its infrastructure relationships aggressively, signing agreements with Amazon, Google, Broadcom, and SpaceX for compute capacity. Locking in a memory and storage partner at this level closes another gap in that strategy. The direction of travel is clear: Anthropic wants supply relationships and technical collaboration at every layer of the stack, not just at the model and cloud level.