About us
Making compute something you buy by the hour
NexGPU aggregates distributed GPU supply worldwide so anyone can rent by the hour, from entry cards to H100s. We do not build datacenters. We turn scattered supply into a marketplace you can order from in one click.
- Countries and regions
- 51
- GPU models in stock
- 68
- Live rentable nodes
- 1,175
- GPUs connected
- 2,498
01 — What we think
Compute should be metered like electricity
For a decade, getting compute has meant one of two things: buy hardware, or sign an annual contract with a cloud vendor. Both require you to forecast demand accurately before you start — and anyone doing research or building a product knows that forecast is usually wrong.
We think the sensible shape is different: take it when you need it, pay nothing when you do not, and let price follow supply and demand rather than one vendor's capacity plan. That is what NexGPU does — aggregate scattered global GPU supply into one market where both price and availability are visible.
The direct consequence is price. The same H100 costs $3.582 per GPU-hour here, 47% below the AWS on-demand rate. Not because our silicon differs, but because the supply route does.
02 — How we operate
Four rules we will not work around
Not values statements — specific product decisions.
Three charges, all stated up front
Compute, storage and network, each with its unit rate and running total on the instance page. The storage charge in particular keeps accruing while an instance is stopped, and we say so on the pricing page, the order form, the instance card and the low-balance email — rather than in clause fourteen.
Rates lock; we never reprice retroactively
The rate at order time holds until that instance is destroyed. A rising market affects new instances only, never a job already running.
Comparisons round against us
When we compare against competitors, our advantage rounds down and our disadvantage rounds up. We currently do not win the H200 tier, and the pricing page marks it "+6%" rather than dropping the row.
No charge for idleness
No minimum spend, no setup fee, no minimum rental period, no port fees. Destroy an instance and billing stops immediately.
03 — Company
Who we are
- Operating entity
- Nebula Global LLC
- Registered address
- 1359 S Drexel Way, Lakewood, CO 80232, US
For contracts, invoicing or a longer-term compute arrangement, reach us through the channels below. Enterprise customers can book a call — we will ask about the shape of your workload before quoting anything.
Have a specific question?
Technical sizing, cost estimates or contract terms — asking us directly is faster than reading pages.
Start building on NexGPU
Enterprise R&D team or solo developer — either way, your first job can be running in minutes.
Sign up to browse live network pricing. No payment method required.