This deployment demonstrates a new approach to scaling AI inference—combining wafer-scale chips and liquid cooling—to boost performance and energy efficiency, potentially reshaping cost and latency dynamics for AI services. It also signals growing capital investment in bespoke infrastructure to meet the compute demands of advanced AI workloads.
Cerebras opened a purpose-built AI data center in Oklahoma City hosting wafer-scale processors that collectively deliver 44 exaflops of compute, which the company says is the fastest AI infrastructure on Earth. The facility uses single, dinner-plate-sized wafer-scale engines with on-chip memory to eliminate off-chip latency and accelerate inference by orders of magnitude. The build is hardened for tornado resilience, relies on large-scale liquid cooling connected to a 6,000-ton chiller plant, and is powered primarily by natural-gas–generated electricity with battery buffers and 3 MW backup generators. The site design also anticipates expansion of cooling and power capacity as demand grows.
Comments
Want to join the conversation?
Loading comments...