Pricing
Straightforward, by the hour.
Three ways to buy. Indicative rates below — final pricing is confirmed when you reserve, based on accelerator, region, and duration.
On-demand
Billed by the second once your instance is running.
- No commitment, cancel any time
- Stop instances on demand
- Standard support
Reserved · Monthly
Reserved blocks of one month or more.
- Capacity reserved in your chosen region
- Priority in queues for newer accelerators
- Business-hours support
Committed · Annual
12-month commitments on larger pods.
- Custom region and accelerator mix
- Dedicated point of contact
- Custom networking and storage
Indicative rate card
Reserved and annual tiers receive a discount applied at reservation. All rates are indicative and may vary based on accelerator, region, and partner availability.
Rates per GPU per hour, in USD. Launch pricing — introductory rates apply during the first capacity wave and will increase as inventory fills. Storage, egress, and network rates depend on the selected region and configuration.
Spend controls
Every account ships with a spend cap.
Launch pricing means we run thin margins on purpose. To keep things sustainable for everyone, new accounts have built-in limits that unlock automatically as you build payment history with us.
New account cap
$250 / 24 hrs
First seven days. Covers about 5 H100 GPU-days. Lifts automatically after your first successful invoice.
Verified cap
$5,000 / month
After ID verification and one paid invoice. Most teams live here.
Custom limits
Unlimited
For reserved monthly or annual commits. Caps removed on contract signature.
Payment methods
Pay with card (Visa, Mastercard, Amex) or crypto (USDC, USDT, SOL, ETH, BTC). A pre-authorization hold is placed at instance launch and released against your usage.
Billing rules
One-hour minimum per instance. Idle instances auto-stop at the cap. Crypto deposits credit a prepaid balance with the same burn-down limits as card holds.