Skip to content

Cloud GPUs

Production GPUs at one-third the price.

B200 from $3.25/hr. H100, H200, and B300 priced proportionally. Sign up and launch in minutes — no quotas, no waitlist, no procurement queue.

What's true about iFrame GPUs

Three claims. Each linked to its proof.

3× more affordable.

Our published pricing is consistently 3× below AWS, Azure, and GCP list pricing for equivalent SKUs. The savings come from a tighter operating model and research-derived stack efficiency, not from older hardware or thinner SLAs.

Instant access.

Most users go from signup to a running GPU in under five minutes. Reserved capacity is available the same day for typical configurations.

No queues.

No quota tickets. No four-week procurement cycles. The capacity on the pricing table is capacity you can buy right now.

Live pricing

Pay 3× less, today.

Per-second metering, billed hourly. The same NVIDIA hardware your team already uses, with hyperscaler list pricing shown alongside so you can verify against your incumbent.

Refreshed weekly · 2026-04-22
  • B300
    192 GB HBM3e
    from $3.25/hr
    up to $4.60 · $12.00 hyperscaler
    73% vs hyperscalersLaunch →
  • B200
    180 GB HBM3e
    from $3.25/hr
    up to $4.60 · $9.80 hyperscaler
    67% vs hyperscalersLaunch →
  • H200
    141 GB HBM3e
    from $2.95/hr
    up to $4.18 · $8.40 hyperscaler
    65% vs hyperscalersLaunch →
  • H100 80GB
    80 GB HBM3
    from $2.25/hr
    up to $3.18 · $6.85 hyperscaler
    67% vs hyperscalersLaunch →

iFrame range: floor at 5-year reserved + 20% prepay; ceiling at 1-month commit + 10% prepay (1-day terms forced to 100% prepay). Hyperscaler comparison uses each provider's on-demand list pricing for the closest equivalent SKU. Spot, committed-use, and marketplace pricing excluded. Run the calculator.

See full pricing

How fast is instant

Minutes on iFrame. Days to weeks at hyperscalers.#

The same SKU, the same time-to-allocate measurement: signup to first booted instance, no special procurement path.

iFrame

≤ 5 min

Self-serve. B200 / H200 / H100 80GB across all 7 regions.

Hyperscaler

1–4 wks

Quota raise → procurement → capacity reservation → launch. Typical for B200 / H200.

Reserved capacity

Predictable workloads, predictable price.#

Multi-month commitments unlock additional discount, dedicated bare metal, named support, and bundled VPC interconnect setup.

6 months
15–20% off

Reserved capacity, named support

1 year
25–30% off

+ VPC interconnect setup, custom SLA

3 years
30–35% off

+ BAA, DPA, audit log streaming, QBRs

Launch flow

Provisioning in three commands.#

  1. 01

    Sign up.

    Email and a credit card. Confirmation in seconds.

  2. 02

    Pick GPU.

    Console, CLI, or API. SKU, region, instance shape — pre-validated for fit.

  3. 03

    Launch.

    SSH into your instance, attach storage, run your code. No new tooling.

Regions

Seven regions today. Sovereign on contract.#

All SKUs available in all regions unless noted. EU and APAC capacity scales with reservation.

  • us-west
    US West
    San Jose
  • us-east
    US East
    Ashburn
  • us-south
    US South
    Dallas
  • eu-central
    EU Central
    Frankfurt
  • eu-west
    EU West
    Amsterdam
  • apac-east
    APAC East
    Tokyo
  • apac-south
    APAC South
    Singapore

FAQ

Buying questions, answered.#

Two minutes from here to a running B200.

Sign up free or talk to sales for reserved capacity, custom SLAs, and VPC interconnect.