Start Building →
Custom quote · 24-hour turnaround

Request a custom quote.

Tell us about your GPU requirements — fleet size, region, commitment term, target price — and we'll prepare a custom quote with verified availability. Typical response time under 24 hours, often same day for clear specs.

No credit card requiredEngineer (not sales) drafts the quoteVerified availability, not estimates

1 Contact information

2 GPU requirements

3 Additional requirements

Quote FAQ

Before you send.

Honest answers to what we hear from buyers in their first quote request. Don't see yours? Put it in the notes field — we'll cover it in the reply.

How is this different from your on-demand pricing?
The price ladder on the homepage is for self-serve, single-GPU, hourly. A custom quote covers volume (typically 8+ GPUs), multi-month commits, specific regions, and InfiniBand clusters — where we can offer significantly better $/hr than the public rate card. For 1–4 GPUs hourly, self-serve is faster and usually cheaper.
How fast can you actually provision a custom quote?
Single-node and small fleets (under 8 nodes): usually 24–72 hours from quote acceptance. Multi-node InfiniBand clusters depend on size and region — 1–2 weeks is typical for 16–64 nodes; 3–6 weeks for 100+ nodes in a specific region. We'll give you a real lead time in the quote, not a marketing number.
Which regions can you actually deliver in?
Active capacity in US East, US Central, US West, EU (Frankfurt, Amsterdam, Paris, Dublin), and UK (London). We can source APAC and LATAM through partners — expect a longer lead time and minor pricing premium. Tier-3 datacenters available on request.
Do you offer reserved / committed pricing?
Yes. Commit terms from 1 month to 36 months. Discounts scale with term length and fleet size. We don't penalize early termination for genuine reasons — talk to us if your needs change.
What networking and storage can you include?
NVLink/NVSwitch within nodes (standard on HGX). Between nodes: NVIDIA Quantum-2 InfiniBand 400 Gb/s or RoCEv2 mandatory for distributed training; standard 1–100 Gbps Ethernet for inference. NVMe local storage, plus shared hot-tier NVMe and S3-compatible object storage are configurable in the quote.
Can I bring my own contract / MSA?
For enterprise commits we're open to negotiated MSAs, DPAs, and security questionnaires. Our standard Terms and DPA cover most cases — we usually settle redlines within a week.
Will I talk to a salesperson?
No. The first reply comes from an engineer who reviewed your spec. A commercial lead joins only if/when contract terms come into play. We don't do high-pressure outreach or follow-up calls you didn't ask for.