Announcing SmartTec: the battery-backed AI cloud

Today we're announcing SmartTec. We've been quiet about this for almost three years. Here's what we're building and why.

The problem

AI workloads need a lot of power. Training a frontier model requires 5-50 MW of continuous power for weeks. Inference at scale requires 1-100 MW sustained, 24/7. The US grid is not built for this. Interconnection queues run 4-7 years. Some markets are functionally closed to new AI tenants.

What we're building

A battery-backed AI cloud. NVIDIA and Cerebras compute behind battery storage we engineer ourselves on z1power LFP — a founder-owned partner company — deployed at our owned site in Mead, Oklahoma. When the grid drops, the batteries take over in under 10 milliseconds by design. Your training job keeps running.

When you can use it

Phase 1 power-on: Q4 2026. We're reserving capacity now for three design-partner teams who want launch pricing and direct engineering access. After Q1 2027 we'll open general reservations.

Where to start

• →See the compute options: /compute
• →See the power story: /power
• →See how orchestration works: /aura
• →Apply for design partner: /deployments
• →Talk to us: /contact