GPU compute, watched by AI agents.
Per-node thermal & power observability. AI-agent rebalancing. Privately racked NVIDIA RTX Pro 6000 Blackwell and B200 fleets, rented by the hour, to AI startups and model labs.
- RTX Pro 6000 · B200 · B300
- Hourly billing — no annual commitment
- Pre-attested SKU hashes
- On-site swap in < 90 min
Per-GPU hourly pricing
Five providers, three SKUs, the same hour.
The NVIDIA SKUs Kilnwright racks — the RTX Pro 6000 Blackwell workstation card and the B200 and B300 server GPUs — priced per GPU-hour against CoreWeave, Lambda, Crusoe, and Fluidstack. The as-of stamp and per-row citations are visible in the matrix; Kilnwright sits at or under each peer on the RTX Pro 6000 and within a few cents on the Blackwell server SKUs.
Hourly USD per GPU · public-market starting rates, refreshed periodically.
Prices, availability, and configurations vary. Kilnwright's rates are published at /pricing; competitor rates are public-market starting figures.
How we compare
Where Kilnwright sits, by category.
Kilnwright is a managed GPU cloud with per-node observability, AI-agent rebalancing, active response, and transparent cost & SLA reporting — positioned on managed reliability rather than bargain pricing or cheapest capacity. The four categories the brief covers: tenant GPU clouds (CoreWeave, Lambda), spot marketplaces (Vast.ai and peers), traditional colocation, and hyperscalers.
Prices, availability, and configurations vary; competitor figures are public-market starting rates, refreshed periodically.
Spin up a tenant
Spin up a tenant in three steps.
One billing contact, one password, one per-tenant API key. Drop straight into the dashboard with an active session.