← All work

AI Compute Infrastructure · CTPO

Bare Metal

Vertically integrated urban AI compute — sovereign, low-latency bare-metal GPU and inference capacity, running inside the cities that need it.

Role

CTPO

Model

Behind-the-meter power · waste-heat recovery

Anchor

San Francisco → NA & EU

Targets

<2ms latency · 99.99% · ≤1.2 PUE

Bare Metal platform Bare-Metal GPU · In-Metro Inference · Clean Power Bmetal

Compute belongs where the users and the power are. Bare Metal builds modular data centres inside cities and gives the waste heat back to the neighbourhood.

Context

AI teams need low-latency, sovereign capacity — not a region three hops away. Bare Metal operates bare-metal GPU clusters and in-metro inference endpoints in densely-powered urban locations, paired with behind-the-meter renewable generation and waste-heat recovery into surrounding district systems.

What I built

  • The corporate and product platform presenting three integrated capabilities: bare-metal compute, in-city inference serving, and distributed power & thermal engineering.
  • An access-request flow with typed inquiries (compute, partnership, investor, press) behind honeypot and time-trap bot defences.
  • A complete, counsel-ready legal suite — eight policies covering GDPR and CCPA/CPRA — living in the codebase.
  • A restrained monochrome design system with disciplined motion, built to read as operational seriousness rather than hype.

How it's built

A Next.js 14 App-Router site in strict-mode TypeScript and Tailwind, with Framer Motion intersection reveals and Lenis smooth scroll — all respecting prefers-reduced-motion. The contact path is a serverless function that validates and screens submissions (honeypot plus a sub-1.5s time-trap) before anything is accepted. Metadata, Open Graph, an auto-generated sitemap and a cookie-consent gate round it out; deployed on Netlify with security headers at the edge.

Outcome

A credible public face for an infrastructure company — visually restrained, technically buttoned-up, and legally complete before launch.

Next.js 14TypeScript (strict)TailwindFramer MotionLenisServerlessNetlify

Core feature

Rebuilt in Bare Metal's own UI

A full-bleed dark platform spec-band that summarizes BMETAL's urban AI compute targets — in-metro latency, availability, PUE, and redundancy — over the signature atmospheric Mercury-Flow render, giving prospects the whole bare-metal/inference/power promise at a glance.

Platform "At a glance" spec-band (dark Mercury-Flow frame)

[ Platform · At a glance ]

Bare-metal, inference, and power — operated whole.

<2ms
In-metro latency
Target · edge to user
99.99%
Availability target
Platform SLA
≤1.2
PUE, urban envelope
Waste-heat recovery
N+1
Power & cooling redundancy
Behind-the-meter
Reserved bare-metal GPU Low-latency inference Behind-the-meter power SOC 2 in progress
Operating in San Francisco