More inference,
radically less energy.

Reserve capacity for Kelvin inference systems.

SHARE INTEREST
GENERATION 01
NORMAL INFERENCE SYSTEMS
COMPUTE-IN-MEMORY
PRODUCT
Inference servers and rack-scale systems.
Built on Thermodynamic Compute-in-Memory.
System Configurations
Server
A complete inference system built on our Thermodynamic Compute-in-Memory architecture.
Rack
Kelvin Servers integrated with rack-scale networking, power, cooling and management.
KELVIN WORKLOADS
Accelerate AI inference.
AI inference
Long-context inference
Disaggregated inference
Mixture-of-experts models
High-concurrency inference
DEPLOYMENT
Inference systems for your workloads.
Kelvin Servers and Racks are complete inference systems for prefill and decode. Rack configurations integrate networking, power, cooling and management, scaling from single- to multi-rack deployments.
Inference deployment
From your model and serving stack to Kelvin inference systems.
PRE-ORDERS
Build your next inference deployment with Kelvin.
Reserve priority capacity in the first available delivery window for your configuration.