Open Grid · free

Bring your context.
Bring your agent.
Bring your keys.

Your docs. Your agent. Your keys. Grid designs the AI campus. Free.

Never stored · Per-workspace isolation · Zero-retention.

< 30 min

Site + MW → package

53+

Checks per run

First-principles

No hardcoded values

Any accelerator

NVIDIA · AMD · Cerebras · Intel

Zero-retention

Docs never stored

Why now

Power deals close in days.
Your design should too.

  • AI is a race for power — sites close in days.
  • The old way: weeks of feasibility and design.
  • Open Grid: a fundable design in minutes.
Old wayWeeks
Open Grid< 30 min

What Open Grid opens

Three things become yours.

Bring your context

Ground every run on your real docs.

  • Connect Drive & Gmail
  • Read-only grounding
  • Nothing stored

Bring your agent

Drive Grid from your own tooling.

  • Public REST API
  • MCP server (OAuth 2.1)
  • Forward an RFP by email

Bring your keys

Run on your own model provider.

  • Bring-your-own keys
  • Per-workspace isolation
  • Runs bill to you

Built for every AI platform

Silicon-neutral by design.

One engine, every accelerator — each modeled from its published reference. No lock-in.

NVIDIA GB200 / GB300NVIDIA Rubin NVL144NVIDIA HGX B200 / H200AMD Instinct MI300X–MI355XCerebras CS-3Intel Gaudi 3SambaNova SN40

Incumbent depth

NVIDIA — to the GB200 NVL72 rack and Quantum-X800 spine.

Challenger depth

Cerebras CS-3 (~23 kW, near-all-liquid) and AMD Instinct — from each vendor’s specs.

New accelerator? Send the reference parameters — we design around it.

Add your platform

Platform names describe reference-architecture compliance only. Not affiliated with or endorsed by any vendor.

How it works

Connect. Generate. Control.

01

Connect

Link Drive/Gmail or forward an RFP.

02

Generate

Power, cooling, floor plan, budget — 53+ checks.

03

Control

Pull via API/MCP. Run on your keys.

Who it’s for

One engine, every team.

Hyperscaler

A campus design at the speed you sign power.

AI DC developer

Site + MW → a fundable package, fast.

Offtaker

See what the power really costs.

Land developer

Prove your parcel hosts an AI campus.

EPC / vendor

A buildable, code-checked scope.

Investor

Underwrite in hours, not a study.

Built for enterprise

Your data stays yours.

Security

  • Docs never stored
  • No training on your data
  • AES-256 / TLS 1.2+
  • SOC 2 controls

Control

  • Per-workspace key isolation
  • Encrypted secrets vault
  • Runs bill to you
  • Engine never exported

Rigor

  • Deterministic
  • Zero hardcoded values
  • Reference-compliant: NVIDIA, AMD, Cerebras, Intel
  • 53+ gates per run
NEC 110.26NFPA 855NFPA 37NFPA 2001IBC §707ASHRAE 90.1IEEE 1547UL 1008SOC 2

FAQ

Straight answers.

Is my data trained on?

No. Grounding only — never stored, never trained on.

Do I get the engine?

No. Grid stays hosted; you get UI, API, and MCP.

Really free?

Yes, forever. Bring your keys and those runs bill to you.

How is a workspace isolated?

Its own key and credentials, encrypted, never cross-readable.

Build on your own

Connect, plug in your keys, drive from your agent. Free, in minutes.

Start free

Partner with us

Gigawatt scale, or enterprise isolation and SSO? Let’s talk.

Talk to us