Anthropic's $19B TeraWulf Lease, Claude Code Gateway and Model Research

Anthropic spends on capacity and access

The largest number of the day belongs to TeraWulf, whose shares surged after it announced a $19 billion lease deal with Anthropic for AI infrastructure. The size of the commitment shows how much Anthropic is willing to spend on infrastructure, and the stock move suggests investors saw it as a major contract for TeraWulf.

On the customer side, Anthropic added an enterprise gateway meant to simplify Claude Code access on AWS and Google Cloud. For engineering organizations that already run on one of those clouds, a gateway may remove some friction from rolling out a coding agent to teams, though the practical gains will depend on how it fits existing access controls and billing.

Rethinking how models use information

Anthropic also published research on a global workspace in language models, a concept concerned with how models process and integrate information. The work aims to improve the coherence and relevance of outputs. It is research rather than a product change, so its direct effect on the models developers use today is unclear.

A more practical approach to handling information appears in the IEEE Spectrum report that small AI models are gaining traction around the world, including in pharmaceuticals. There, small models prune unnecessary context before question answering, which the report says improves performance while cutting operating costs. For teams running retrieval-heavy systems, trimming context with a cheaper model before calling a larger one is a pattern worth testing against their own latency and cost numbers.

AI adoption and headcount move together

A recent study finds that companies that adopt AI heavily also hire more. An association like this does not by itself show that AI adoption causes hiring, since growing firms may also be more likely to buy new tools. Still, it cuts against the assumption that heavy AI use goes hand in hand with shrinking teams.