Vera Rubin NVL72 / Vera Rubin NVL72 price
NVIDIA Vera Rubin NVL72 price
There's no list price for Vera Rubin NVL72 at rack scale. These are the things that move the number, and how to get a quote you can sign.
Why we don't post a number
Vera Rubin capacity is sold deal by deal, mostly before it's built. Any figure on a page would be a guess, and wrong for your size and term. We'd rather get you live quotes from operators within 24 hours.
Think in cost per token
The newest rack is rarely the cheapest to rent. The case for it is output: NVIDIA claims one-tenth the cost per million tokens against GB200 NVL72 (NVIDIA). Compare quotes on what your workload produces per dollar and per megawatt.
What moves the price
Term length
Operators finance new builds against your contract. Three to five years prices best; short terms are rare this early in a generation.
Rack count
Scale earns better rates and more attention, but ties you to a specific build and its schedule.
Prepayment and deposits
Money down lowers the operator's cost of capital and usually your rate. Check what's refundable if delivery slips.
Start date
Early slots carry a premium. A start two quarters later can widen your options a lot.
Site and power
Region, power price and cooling design all feed the rate. Sovereign or data-residency needs narrow the field.
What's included
Bare metal vs managed Kubernetes or Slurm, storage, network between racks, support and uptime commitments.
How to compare quotes fairly
- Normalise to the same term, rack count and start date.
- Check the unit: per GPU-hour, per rack-month or per megawatt.
- Price the delay risk. A cheaper rack that arrives two quarters late may cost you more.
Send us the size, term and start date and we'll come back with live options. Request capacity →
Questions
How much does it cost to rent a Vera Rubin NVL72?
Can I buy Vera Rubin NVL72 racks instead?
RackShore runs this site and may be paid a referral fee by the provider you choose. Buyers pay us nothing. Spot a mistake? Tell us.