Cloud GPU Egress Costs Compared
Egress is the cost of transferring data out of a cloud provider — downloading model checkpoints, shipping training artifacts to another region, or serving inference results. For GPU workloads that routinely move large datasets and model weights, egress fees can add hundreds of dollars to your monthly bill. Some providers include free egress, while hyperscalers typically charge per GB.
$0.00
$0.000
Yes
Reserved instances billed continuously. On-demand billed per hour. Disk and egress included.
$0.00
$0.000
Yes
Billed while running. Bandwidth varies by host (typically 1–10 Gbps). No persistent disk.
$0.00
$0.000
No
Billed per hour. Marketplace — verify host details before long jobs.
$0.00
—
No
Newer provider. Check docs for storage and billing details.
$0.00
—
No
Newer provider. Check docs for storage and billing details.
$0.00
—
No
Newer provider. Check docs for storage and billing details.
$0.00
—
No
NVIDIA-partner cloud. Verify storage billing details before long-running jobs.
$0.00
$0.000
No
Iceland data center, renewable energy. Storage and egress included in instance price.
$0.00
—
No
Runs on stranded/flared gas energy. Competitive pricing on H100 clusters.
$0.00
$0.000
No
Bare metal GPU servers. Storage and bandwidth included. Hourly or monthly billing.
$0.00
—
No
API-first GPU cloud. Check current pricing at fluidstack.io.
$0.00
$0.000
No
All-inclusive pricing. Storage and bandwidth included. EU data centers.
$0.00
$0.050
No
ML-focused. Free egress. Spot and on-demand pricing available.
$0.00
—
No
India-based. Pricing in INR/USD. Check current rates on website.
$0.00
$0.050
No
Pause instance to stop GPU billing. Storage persists and billed separately.
$0.00
—
No
Aggregator — provisions GPUs from multiple underlying providers.
$0.00
$0.000
No
Distributed compute on consumer GPUs. Very cheap but variable performance.
$0.00
$0.050
No
Edge cloud. Free egress. GPU VMs with L40S and A100.
$0.00
—
No
GPU cloud focused on H100 clusters. Contact for pricing on large deployments.
$0.00
—
No
European sovereign cloud. GDPR compliant.
$0.00
—
No
ML platform with GPU studios. Pay-as-you-go GPU compute.
$0.00
$0.000
No
Per-second billing. Pay only when your code is running. No idle charges.
$0.00
$0.000
No
Per-second GPU billing. Pay only for actual compute time.
$0.00
$0.000
No
API-based pricing. Pay per token for inference, per GPU-hour for training.
$0.00
$0.000
No
API inference pricing. Pay per token or per request.
$0.00
$0.000
No
Bare metal GPU servers. Hourly or monthly billing.
$0.00
—
No
Dedicated GPU servers. Hourly and monthly pricing.
$0.00
—
No
Budget GPU cloud. Check pricing on website.
$0.00
—
No
GPU VMs with optional Windows. Check pricing on website.
$0.00
—
No
Per-hour billing. Self-serve sign-up.
$0.00
—
No
Per-second billing. EU-sovereign infrastructure.
$0.00
—
No
Per-millisecond billing. No idle costs.
$0.00
—
No
Per-hour billing. MI300X and H100 clusters available.
$0.00
—
No
Per-hour billing. Enterprise contracts available.
$0.00
—
No
Per-second billing. No minimum commitment.
$0.00
—
Yes
Primarily monthly billing. Hourly available on some plans.
$0.00
—
No
Per-hour billing. 100% renewable energy.
$0.00
—
No
Per-hour billing. French-sovereign infrastructure.
$0.00
—
No
Per-second billing. No minimum commitment.
$9.22
$0.025
No
Very cheap egress ($0.009/GB). Stop instance to pause GPU billing.
$10.24
$0.040
No
EU-focused. Stop instance to pause billing.
$10.24
$0.060
No
Power off instance to stop GPU billing. Cheap egress.
$10.24
$0.100
No
GPU Droplets. Power off to stop billing. Volume storage billed separately.
$10.24
$0.040
No
EU-based. Stop instance to pause GPU billing. Block volumes billed separately (~€0.04/GB/mo).
$10.24
$0.070
No
Gradient notebooks and VMs. Storage billed separately. Stop VM to pause GPU billing.
$10.24
$0.100
Yes
GPU Linodes billed per hour. Transfer included in plan.
$11.26
$0.040
No
Major EU cloud. Competitive egress pricing. Stop instance to pause GPU billing.
$20.48
$0.150
No
Per-second billing. Machines can be suspended between jobs.
$51.20
$0.070
No
Newer provider. Verify billing semantics before long jobs.
$51.20
$0.100
No
Per-hour billing. Kubernetes-native GPU instances.
$81.92
$0.040
No
Stop instance to pause GPU billing. Persistent disk billed separately.
$81.92
$0.050
No
Pay-as-you-go and subscription pricing. Strong in APAC regions.
$81.92
$0.040
No
Pay-as-you-go pricing. Strong in APAC regions.
$89.09
$0.050
No
Deallocate VM to stop GPU billing. Disk storage billed separately.
$92.16
$0.080
No
Stop instance to pause GPU cost. EBS volumes billed separately (~$0.08/GB/mo).
$92.16
$0.100
No
Enterprise cloud. GPU VMs with V100 and A100. Contact sales for custom configurations.
$102.40
$0.070
No
GPU not billed when stopped. Network Volume billed separately. Spot pods can be interrupted.
$102.40
$0.040
No
NVIDIA-backed, InfiniBand fabric, enterprise-grade. Contact sales for committed pricing.