Anthropic signs $19B data center lease to scale AI compute

Anthropic’s multi-decade commitment highlights the critical shift toward vertical infrastructure control in the AI race. For the streaming sector, this signals a massive expansion of the compute backbone required for real-time personalization, automated video encoding, and agentic ad-decisioning at scale. The move away from general-purpose cloud capacity toward purpose-built, high-density campuses suggests that the next phase of streaming innovation will be gated by access to specialized, water-efficient cooling and dedicated power grids. Watch for Anthropic's potential deployment of custom Samsung-manufactured inference silicon in this facility to further optimize tokens per watt.

The lease follows reports from The Information in July 2026 that Anthropic is in early-stage discussions with Samsung Electronics to manufacture a custom AI accelerator chip using a 2-nanometer process. This potential silicon partnership, alongside the TeraWulf site, reflects a broader strategy by AI labs to insulate themselves from single-supplier shocks and reduce reliance on Nvidia. Per TechCrunch in July 2026, Anthropic intends to maintain a diversified stack including chips from Google, Amazon, and Nvidia, even as it builds out proprietary hardware capabilities.

The shift toward high-density infrastructure is a response to the cooling requirements of modern AI clusters. According to Deloitte in May 2026, next-generation AI racks are expected to reach densities of 370kW, making traditional air cooling obsolete. Vendors are increasingly adopting liquid cooling solutions like the one at the Hawesville site; Nvidia recently introduced its Rubin architecture, which achieves 100% liquid cooling for all components, per company announcements in June 2026.

Energy availability has become the primary bottleneck for scaling these models. Gartner estimated in early 2026 that global data center electricity demand will exceed 1,000 TWh by the end of the year, roughly double the 2023 baseline. This infrastructure crunch is pushing AI developers toward "brownfield" sites—like TeraWulf’s former smelting plant—which provide immediate access to existing high-voltage transmission lines that would otherwise take years to permit and construct.