// Pricing

Transparent per-GPU-hour rates.

No dark patterns. Committed use saves up to 40%.

// Compute
GPUOn-Demand / hrReserved 1-moReserved 12-mo
H100$2.99$2.29$1.89
H200$3.49$2.79$2.29
GB200 NVL72ContactContact
B200$5.49$4.39$3.59
L40S$1.29$0.99$0.79
A100$1.79$1.39$1.09
// Cost estimator

Estimate your monthly spend.

Illustrative only. Actual pricing is confirmed on quote.

// New
Try the interactive Bundle Builder
Pick a workload — we recommend GPU + storage + networking and price the full package.
Build a bundle
GPU
Commitment
GPUs: 8
1128256512
Hours per month: 720
1168 (1w)720 (1mo)
Estimated monthly
$17,222
H100 · 8 GPUs · 720 hr · On-Demand
Effective rate$2.99 /GPU-hr
Get a formal quote
// Storage
ItemUnitPriceNotes
Object storageGB-month$0.018S3-compatible, in-region reads free
Block storage NVMeGB-month$0.09Attached to compute nodes
Parallel FS (Lustre)GB-month$0.13For HPC and training pipelines
SnapshotsGB-month$0.05Incremental, encrypted
// Networking
ItemUnitPriceNotes
In-region trafficGBFreeCompute ↔ storage ↔ endpoints
Egress to internetGB$0.03First 100 TB / month included on Enterprise
Private interconnectport-hour$0.7510G / 100G to your DC or hyperscaler
Public IPv4hour$0.005IPv6 is free
// Services & Support
ItemUnitPriceNotes
Cloud MigrationengagementFixed-fee, scoped4-8 weeks, joint team
ML Engineering block160 hrs$28,000Renewable monthly
Priority Supportmonth$2,00024/7 chat + call, 1h P1
Enterprise SupportannualCustomNamed TAM, 15-min P1, QBRs
Training Workshopcohort$12,0002-5 days, up to 20 attendees
// Market prices

How we compare.

Reference on-demand $/GPU-hr against public hyperscaler list prices. Hyperscaler rates as observed; contract & spot pricing varies.

GPUGensparklyAWSGCPAzureSavings
H100 80GB$2.99$12.29$11.06$9.80-73%
H200 141GB$3.49$10.60$10.20$9.90-66%
B200 192GB$5.49WaitlistWaitlistWaitlist
L40S 48GB$1.29$3.64$3.32$3.20-62%
A100 80GB$1.79$4.10$3.67$3.40-52%

// Hyperscaler figures are indicative list prices for equivalent single-GPU on-demand instances. Reserved and committed-use pricing narrows the gap.

// FAQ

Pricing questions.

Everything you need to know about reserved capacity, SLAs, and billing.

// Plans
On-Demand
Per GPU / hour

Spin up instantly. No commitment.

  • Sub-minute provisioning
  • H100 · H200 · L40S · A100
  • Per-second billing
  • Standard support
Get started
Reserved
1 to 12 months

Committed capacity at a lower rate.

  • Up to 40% savings
  • Guaranteed availability
  • Dedicated clusters
  • Priority support
Get started
Enterprise
Custom

Bare-metal, private cloud, or custom fabric.

  • Custom hardware & regions
  • Dedicated account team
  • Compliance & DPA
  • SLA & credits
Talk to sales