GPU Rental Prices
SearchCtrl K
Fee terms checked 16 October to 3 November 2026, each row dated · prices checked 21 November 2026

GPU cloud fees: storage, egress and job cost

Compare data transfer, storage and billing rules alongside the GPU rate. The table records provider documentation; "not published" means exactly that, which is itself worth knowing.

GPU job cost calculator

Compare compute, storage and egress for one job. Unknown fees stay unknown. Rates below the form are editable only in your browser.

Worked example: one hour on H100 SXM with 100 GB stored for a month and 100 GB sent out, priced at the cheapest listed configuration that publishes its GPU count. Type over any figure to price your own job.

Starting volumes are examples. Rental rates checked 21 November 2026; fee terms checked 16 October 2026. Storage conversion uses 720 hours per month.

Massed Compute
$3.14

Known subtotal; other fees excluded.

Compute: $3.14 · Storage: unknown · Egress: $0.00 · Other fees: unknown. Hatching marks an unknown cost, not a zero or a measured amount.

$3.14 known subtotal. Compute $3.14; storage unknown; egress $0.00. Missing: storage.

Only unambiguous free-transfer or single-rate fee cells are prefilled. Conditional storage classes, free allowances and transfer destinations need a quote; enter applicable rates to complete the estimate. Minimum deposits, taxes, public IPs, licensing, support and regional surcharges remain outside this calculation.

Billable runtime = runtime rounded up to the billing increment, and never less than the provider's minimum where its terms state one plainly (60 seconds at AWS, for example). Compute = billable hours × node GPU count × per-GPU rate. Storage = GB × monthly rate × retained hours ÷ 720. Egress = outgoing GB × rate.

GPU provider fee comparison

ProviderEgressIngressStorageBilled when stoppedBilling incrementMin durationMin deposit
HetznerChecked 3 November 2026free (unlimited traffic on the standard 1 Gbit/s connection)freenot publishednot publishedmonthly, or hourly; a one-time setup fee either waynot publishednot published
Oracle CloudChecked 3 November 2026first 10 TB a month free, then $0.0085/GB from North America and Europefree$0.0255/GB-month block volumeyes on most GPU shapes, at the full running rate; A10 virtual machines pause billingper-second1 minute on virtual machines, 1 hour on bare metalnot published
RunPodChecked 16 October 2026freefree$0.10/GB/month for container or volume disk on a running Pod; $0.20/GB/month for volume disk on a stopped Pod; network volumes $0.07/GB/month under 1TB, $0.05/GB/month over 1TBstorage only: volume disk at $0.20/GB/month and network volumes at their usual rate; container disk is erased, not chargedper-secondnot publishedno fixed deposit; balance must cover at least one hour of the chosen Pod to deploy
LambdaChecked 16 October 2026freefreenot publishedstorage onlyper-minute2 weeks for 1-Click Clustersno deposit; $10 card pre-authorization, refunded in a few days
HyperstackChecked 16 October 2026freefree$0.07/GB/month shared storagefull rate while stopped; when hibernated, only its root disk (about $0.07/GB/month) and any kept public IPper-minutenot publishedno fixed deposit; prepaid balance must cover existing resources plus the new VM for at least one hour
CoreWeaveChecked 16 October 2026freefree$0.070/GB/month distributed file storage; $0.06/GB/month hot object storagenot publishednot publishednone for On-Demand and Spot; Reserved and Flex require a term commitmentnot published
NebiusChecked 16 October 2026free for VM network traffic; $0.015/GiB Standard Object Storage egressfree$0.0530/GiB/month block volumes without replicationstorage onlyper-secondnot published$25 minimum first payment
Verda (DataCrunch)Checked 16 October 2026not publishednot published$0.20/GiB/month for NVMe block storage and the NVMe shared filesystemyesper 10 minutes, prepaid; the unused part is refunded on terminationnone for pay-as-you-gonone stated for card top-ups; 1,000 EUR or USD minimum for invoice top-ups
Voltage ParkChecked 16 October 2026freefreenot publishedfull running rate; a deallocated virtual instance pays for storage onlyper-hourno minimum terms for on-demandnot published
CrusoeChecked 16 October 2026freefree$0.08/GiB/month persistent disk; $0.07/GiB/month shared diskstorage onlyper-secondno minimum commitment on demandnot published
JarvislabsChecked 16 October 2026free in-regionfree into Object Storage; not published for instances$0.10/GB/monthstorage onlyper-minuteno minimum commitment$5 per its FAQ; its home page says $10
OVHcloudChecked 16 October 2026free, except S3 object storage egress and Asia-Pacific regions beyond 1 TB/month per projectfreeabout $0.048/GB/month for Classic block storage, about $0.096/GB/month for High Speedfull rate when turned off or paused; a shelved instance pays only for its snapshotper-second1 minute per resource$30 credit charge for first Public Cloud project
SpheronChecked 16 October 2026included in the hourly rateincluded in the hourly ratenot publishedstorage only (retained disk, plus public IP where provider charges)per-minute20 minutes (60 minutes on a 2x AMD MI300X)$15
Microsoft AzureChecked 16 October 2026first 100 GB/month free, then $0.087/GB falling to $0.05/GB by volume (from North America or Europe)freeabout $0.08/GiB/month for Premium SSD v2 in East US; Premium SSD is priced per disk sizestorage only when Stopped (Deallocated)per-minutenot publishednot published
AWSChecked 16 October 2026$0.09/GB first 10TB after free tierfree$0.08/GB-month gp3 example regionstorage onlyper-second60 secondsnone (no upfront payment for On-Demand)
Together AIChecked 16 October 2026freefree$0.16/GiB/month for the shared filesystemstorage onlyper-hournone on demand; reserved clusters from 3 days$15 minimum first credit purchase ($5 after that)
falChecked 16 October 2026not publishednot publishednot publishednoper-secondnot publishednot published
ModalChecked 16 October 2026$0.04/GiB over the plan's monthly allowance (1 TiB Starter, 10 TiB Team, 100 TiB Enterprise)not published$0.09/GiB/month for Volumes, the first 1 TiB a month freenoper-secondno minimum usage-time incrementsnone; pay as you go with a payment method on file
VultrChecked 16 October 2026$0.01/GB over included bandwidthfree$0.018/GB/month Standard Object Storage over included 1000GByesper-hour1 hournot published
Massed ComputeChecked 16 October 2026freefreenot publishedyesper-hournot publishednot published
Google CloudChecked 17 October 2026$0.12/GiB to North America for 1 to 1,024 GiB a month (Premium Tier; rates vary by destination and volume)free$0.000136986 per GiB-hour for Balanced Persistent Disk (us-central1) (about $0.10 per GiB-month at 730 hours)attached disks and external IP addresses onlyper-second1 minutenot published
DigitalOceanChecked 17 October 2026$0.01/GiB beyond each plan's included transferfree$0.10/GiB/month for Volumes Block Storagefull running rate; a powered-off GPU Droplet is billed until destroyedper-second60 seconds or $0.01, whichever is higher (the GPU Droplets pricing page says a 5-minute minimum)card charged at the end of the billing cycle or when usage passes a threshold
TensorDockChecked 17 October 2026freefree$0.00005/GB/hour in TensorDock's example; hosts set the rate (about $0.037 per GB-month at 730 hours)storage only, at a lower rate, once the GPUs are stopped and unreservedper-secondno minimum$5.00 minimum deposit
Thunder ComputeChecked 17 October 2026freenot published$0.219/GB/month for instance disk beyond the first 100 GB per GPU; snapshots $0.05/GB/monthno stop option; delete the instance and keep a snapshot, billed as snapshot storageper-minutenot publishedauto-pay with a saved card, or prepaid credit
Latitude.shChecked 17 October 2026$0.01/GB beyond the 20 TB included per serverfree$0.08/GB/month for Block Storagefull running rate for virtual machines; shutting one down does not pause billinghourly, paying only for the time usedno long-term commitmenta $5 card hold, refunded after 7 days
ScalewayChecked 17 October 2026freefree€0.000130 per GB-hour for Block Storage 5K (Paris, before VAT) (about €0.095 per GB-month at 730 hours)storage volumes and IP addresses only when powered off; standby is billed as runningper-minutenot publishednot published
Hot AisleChecked 17 October 2026not publishednot publishednot publishednot publishedper-minuteone GPU for one minuteprepaid credits by card (Stripe) or USDT and USDC
Prime IntellectChecked 17 October 2026not publishednot published$0.00015/GB/hour in Prime Intellect's network disk example; the rate varies by provider (about $0.11 per GB-month at 730 hours)pause only on some providers (Runpod), at a lower cost than running; disks are billed until terminatedper-minutenot publishedprepaid credits, deducted every minute while a Pod runs
SaladCloudChecked 17 October 2026not publishednot publishednot publishedbilled only while running; billing stops when the instance stops, is reallocated or is preemptedper-secondnot publishedprepaid credits, added $5 to $10,000 at a time
ReplicateChecked 17 October 2026not publishednot publishednot publisheddeployments are charged while instances are online, including setup and idle timepriced per secondnot publishedcredit bought upfront
Fireworks AIChecked 17 October 2026not publishednot publishednot publishedscale-to-zero; billed by GPU-second only while replicas are activeper-secondnot publishedprepaid credits
BasetenChecked 17 October 2026not publishednot publishednot publishedno GPU charge while scaled to zero (min_replica 0)per-minutenot publishedinvoiced when usage passes $50 or at the end of the month
GMI CloudChecked 17 October 2026not publishednot publishednot publishednot publishedpriced per GPU-hournot publishedcredit balance, reloaded from a saved card near a threshold
CerebriumChecked 17 October 2026not publishednot published$0.05/GB/month; the first 100 GB freecompute charged per second of use; persistent storage billed per GB per monthper-secondnot publishednot published
KoyebChecked 17 October 2026$0.02/GB in the EU and US beyond the included bandwidth (pricing page; Koyeb's FAQs give other figures)not publishednot publishedpaused Services are not chargedper-secondnot publisheda $29 card hold, cancelled at once
DaytonaChecked 17 October 2026not publishednot published$0.000108 per GiB-hour after the first 5 GiB (about $0.079 per GiB-month at 730 hours)GPU sandboxes are deleted when stoppedper-secondnot publishedspot GPU sandboxes need Tier 2 (a linked card and a $25 top-up)
LiumChecked 17 October 2026volume egress billed at cost per GBnot published$0.00005 per GB-hour for volumes (about $0.036 per GB-month)only running pods are chargedper-secondno minimum$10 minimum top-up by card or stablecoin
Impossible CloudChecked 17 October 2026not publishednot published€7.99 per TB/month for its separate S3 object storagenot publishedpriced per GPU card-houra minimum contract per listing (7 months on the one read)not published
SynteqChecked 17 October 2026bandwidth included on every plannot published$4/TB/month for block storagenot publishedmonthly, prepaid1 month; cancelling gives no partial refundfirst month paid up front
HyperAIChecked 17 October 2026not publishednot publishedfirst 20 GB free; extra storage from $2 a month for 20 GB to $230.40 a month for 5 TBcompute billed only while a container is runningnot publishednot publisheda balance deposit before buying compute; containers stop when it runs out
RuncrateChecked 31 October 2026freenot published$0.03/GB/monthnot publishedper-secondno minimum$20 minimum top-up; a launch needs $10 or an hour of the machine, whichever is more
GPUwerkChecked 31 October 2026freefreenot published75% of the running rate until terminatedper-minuteno minimum$10 minimum top-up; the first $10 top-up becomes $20 in credit

Verified cells link to their source. Each row or cell shows the day it was read. Green egress cells = flat free egress (13 of 42 providers); conditional or in-region-only free is not highlighted. Corrections: [email protected]. Licensed under CC BY 4.0 with attribution.

Data egress costs: what leaving the cloud costs

Data egress cost is the fee for moving data out of a cloud: downloading trained model weights, exporting datasets, or streaming results to another provider. Ingress (uploading in) is usually free; egress is where the surprises live. Of the 42 GPU clouds in the matrix, 13 publish flat free egress, 16 charge for it or attach conditions (per-GB rates, included-bandwidth caps, or free only in-region), and 13 do not publish a rate at all. "Not published" is worth treating as a question to ask before you commit, not as free.

ProviderData egress costData ingress cost
Hetznerfree (unlimited traffic on the standard 1 Gbit/s connection)free
Oracle Cloudfirst 10 TB a month free, then $0.0085/GB from North America and Europefree
RunPodfreefree
Lambdafreefree
Hyperstackfreefree
CoreWeavefreefree
Nebiusfree for VM network traffic; $0.015/GiB Standard Object Storage egressfree
Verda (DataCrunch)not publishednot published
Voltage Parkfreefree
Crusoefreefree
Jarvislabsfree in-regionfree into Object Storage; not published for instances
OVHcloudfree, except S3 object storage egress and Asia-Pacific regions beyond 1 TB/month per projectfree
Spheronincluded in the hourly rateincluded in the hourly rate
Microsoft Azurefirst 100 GB/month free, then $0.087/GB falling to $0.05/GB by volume (from North America or Europe)free
AWS$0.09/GB first 10TB after free tierfree
Together AIfreefree
falnot publishednot published
Modal$0.04/GiB over the plan's monthly allowance (1 TiB Starter, 10 TiB Team, 100 TiB Enterprise)not published
Vultr$0.01/GB over included bandwidthfree
Massed Computefreefree
Google Cloud$0.12/GiB to North America for 1 to 1,024 GiB a month (Premium Tier; rates vary by destination and volume)free
DigitalOcean$0.01/GiB beyond each plan's included transferfree
TensorDockfreefree
Thunder Computefreenot published
Latitude.sh$0.01/GB beyond the 20 TB included per serverfree
Scalewayfreefree
Hot Aislenot publishednot published
Prime Intellectnot publishednot published
SaladCloudnot publishednot published
Replicatenot publishednot published
Fireworks AInot publishednot published
Basetennot publishednot published
GMI Cloudnot publishednot published
Cerebriumnot publishednot published
Koyeb$0.02/GB in the EU and US beyond the included bandwidth (pricing page; Koyeb's FAQs give other figures)not published
Daytonanot publishednot published
Liumvolume egress billed at cost per GBnot published
Impossible Cloudnot publishednot published
Synteqbandwidth included on every plannot published
HyperAInot publishednot published
Runcratefreenot published
GPUwerkfreefree

Same verified cells as the full matrix above, sliced to transfer fees; each verified value links to the provider page it was read from. Why it matters: a big checkpoint or dataset leaving a per-GB-egress cloud can cost more than the compute that produced it, and the gap between a free-egress neocloud and a hyperscaler's tiered egress compounds with every export. If your workflow moves large artifacts between clouds regularly, weight the egress column as heavily as the $/hr sticker on the live price table.

Why this matters more than the $/hr

Three patterns to watch. First, egress: moving a trained model or dataset out of a cloud with per-GB egress can cost more than the training run. Second, storage while stopped: some providers keep billing for attached storage when the instance is off. Third, billing increments: per-second billing makes bursty work dramatically cheaper than per-hour billing at the same $/hr rate. The calculator and the live table cover the compute side; this page covers the rest.

Compare your offers

Saved prices keep their original dates. Confirm the configuration, terms and availability at the source.

Select up to three offers to compare their rates and terms.

Type a GPU, a provider, a model or a question.