Simple, transparent pricing
Every rate on this page is the rate you pay — metered per second, invoiced monthly through Stripe, identical in all four regions. New accounts start with $10 of free credit.
On-demand
Launch any SKU the minute you sign up. Metered by the second from $0.49/hr — no quotas, no capacity requests, and when you stop an instance, the meter stops with it.
- SSH root access
- Prebaked ML images
- Persistent NVMe volumes
- Live spend metering
One meter, four products
- GPU Cloud
- Bare Metal & VMs
- Storage
- Serverless AI
from $0.49 /GPU-hr
H100 SXM at $2.49/hr · billed per second
Reserved
Hold capacity for 1–12 months at up to 35% below on-demand, or reserve dedicated HGX pods with a private InfiniBand fabric.
Custom quote within one business day
Talk to our teamTrusted by ML teams shipping in production
Every SKU, one page
Twenty products across GPU, CPU, bare metal, and storage — the same published rate in every region, with no negotiated-deal asterisks.
Explore the GPU fleet in depth| GPU | VRAM | vCPU / RAM | Storage | Network | Price / hr Reserved / mo | Deploy |
|---|---|---|---|---|---|---|
|
B200 SXM Newest silicon1× NVIDIA B200 |
180 GB | 28 vCPU · 360 GB | 3 TB NVMe | 100 Gbps Ethernet | $5.49/hr $2,599/mo −35% vs on-demand | Deploy |
|
H200 SXM 1× NVIDIA H200 |
141 GB | 24 vCPU · 256 GB | 2 TB NVMe | 100 Gbps Ethernet | $3.69/hr $1,749/mo −35% vs on-demand | Deploy |
|
H100 SXM Most popular1× NVIDIA H100 |
80 GB | 26 vCPU · 224 GB | 1.5 TB NVMe | 100 Gbps Ethernet | $2.49/hr $1,179/mo −35% vs on-demand | Deploy |
|
H100 SXM ×8 Node Best for training8× NVIDIA H100 |
640 GB | 208 vCPU · 1,843 GB | 22 TB NVMe | 3.2 Tbps InfiniBand + NVLink | $18.32/hr $2.29/GPU-hr effective $8,689/mo −35% vs on-demand | Deploy |
|
H100 PCIe 1× NVIDIA H100 |
80 GB | 26 vCPU · 200 GB | 1 TB NVMe | 25 Gbps Ethernet | $2.09/hr $989/mo −35% vs on-demand | Deploy |
|
A100 80GB SXM 1× NVIDIA A100 |
80 GB | 24 vCPU · 192 GB | 1 TB NVMe | 25 Gbps Ethernet | $1.29/hr $609/mo −35% vs on-demand | Deploy |
|
RTX 5090 1× NVIDIA RTX 5090 |
32 GB | 16 vCPU · 64 GB | 500 GB NVMe | 10 Gbps Ethernet | $0.79/hr $379/mo −34% vs on-demand | Deploy |
|
L40S 1× NVIDIA L40S |
48 GB | 16 vCPU · 96 GB | 750 GB NVMe | 25 Gbps Ethernet | $0.69/hr $329/mo −35% vs on-demand | Deploy |
|
RTX 4090 Starter1× NVIDIA RTX 4090 |
24 GB | 12 vCPU · 64 GB | 400 GB NVMe | 10 Gbps Ethernet | $0.49/hr $229/mo −36% vs on-demand | Deploy |
Multi-GPU nodes ship with NVLink and a dedicated InfiniBand fabric. H100 SXM ×8 Node works out to $2.29 per GPU-hour on demand.
| Instance | vCPU | RAM | NVMe storage | Transfer | Price / hr Reserved / mo | Deploy |
|---|---|---|---|---|---|---|
|
VC-S2 Shared Shared vCPU |
2 vCPU | 4 GB | 80 GB NVMe | 4 TB transfer | $0.02/hr $14/mo | Deploy |
|
VC-S4 Shared Shared vCPU |
4 vCPU | 8 GB | 160 GB NVMe | 5 TB transfer | $0.04/hr $28/mo | Deploy |
|
VC-D4 Dedicated Dedicated vCPU |
4 vCPU | 16 GB | 200 GB NVMe | 6 TB transfer | $0.10/hr $69/mo | Deploy |
|
VC-D8 Dedicated Dedicated vCPU |
8 vCPU | 32 GB | 400 GB NVMe | 8 TB transfer | $0.20/hr $135/mo | Deploy |
Virtual CPU instances for the services around your GPUs — data prep, control planes, APIs. Monthly caps mean an always-on box never exceeds its reserved rate.
| Server | Threads | RAM | Storage | Network | Price / hr Reserved / mo | Deploy |
|---|---|---|---|---|---|---|
|
Metal 4E Starter AMD EPYC 4344P (8C/16T Zen 4) |
16 | 64 GB ECC | 2× 960 GB NVMe RAID1 | 10 Gbps · 20 TB egress | $0.37/hr $229/mo −15% vs on-demand | Deploy |
|
Metal X16 Intel Xeon Gold 6526Y (16C/32T Emerald Rapids) |
32 | 128 GB ECC | 2× 1.92 TB NVMe | 10 Gbps · 20 TB egress | $0.62/hr $379/mo −16% vs on-demand | Deploy |
|
Metal E32 Most popularAMD EPYC 9354P (32C/64T Genoa) |
64 | 256 GB ECC | 2× 1.92 TB NVMe | 25 Gbps · 20 TB egress | $1.05/hr $649/mo −15% vs on-demand | Deploy |
|
Metal E64 AMD EPYC 9554P (64C/128T Genoa) |
128 | 768 GB ECC | 4× 3.84 TB NVMe | 2× 25 Gbps · 20 TB egress | $2.99/hr $1,899/mo −13% vs on-demand | Deploy |
Single-tenant hardware delivered in hours, not weeks. Every server includes 20 TB of egress and lives on the same private network as your GPU instances.
| Product | Capacity | Network | Price | Get started |
|---|---|---|---|---|
|
NVMe Block Volumes Triple-replicated NVMe block storage |
10 GiB – 16 TiB per volume | Attach to any instance | $0.08/GB · mo | Start free |
|
Object Storage S3 API with 11×9s durability |
S3-compatible, unlimited buckets | Free ingress | $8/TB · mo | Start free |
|
Snapshots & Backups Scheduled or on-demand snapshots of VMs and volumes |
Incremental, compressed | Cross-region copy | $0.04/GB · mo | Start free |
Storage is provisioned capacity, billed monthly. Object storage egress is free up to 3× stored volume; ingress is always free.
On-demand usage is metered per second and invoiced monthly through Stripe. Reserved rates run on 1–12 month terms — switch a running instance between modes without redeploying.
The list price is the whole price
No line items creeping onto the invoice. The platform around your compute — network, data protection, and humans — ships at $0.
Free VPC networking
Isolated VPCs, private IPs, and security groups in every region — at $0, from the first instance.
20 TB egress per server
Pooled across your whole fleet each month. Ingress is free, always, everywhere.
Snapshot tooling
Scheduled snapshots with cross-region restore, built into every VM and volume — you pay only for the bytes stored.
24/7 engineer support
Infrastructure engineers on call around the clock — median first response under 15 minutes.
Up to 68% less than hyperscaler list
An H100 SXM runs $2.49/hr on Reviosa. The same 80 GB silicon lists in the $6–8/hr class at the big three — before egress fees and hourly rounding.
- Day-one pricing — the listed rate needs no 1-year commit or negotiated deal to unlock.
- 20 TB egress included — hyperscalers meter every gigabyte leaving the building.
- Per-second metering — a 47-second smoke test bills 47 seconds, not a full hour.
1× H100 80 GB SXM, on-demand
$/hrReviosa Cloud
$2.49
Hyperscaler A, list
$6.98
Hyperscaler B, list
$7.86
Public on-demand list prices for comparable 1× H100 80 GB instances, US regions, July 2026. Reviosa rate rendered live from the catalog above. Effective savings vary by region and workload.
Frequently asked questions
Anything else? Our support engineers answer pricing questions too — no sales gate in front of them.
support@reviosa.comPer second, from the moment an instance is SSH-ready until you stop it. An H100 SXM at $2.49/hr meters $0.0007 each second — nothing is rounded up to the hour. Stopped instances accrue $0.00 for compute; attached volumes and reserved IPs continue at their storage rates. Everything rolls into one monthly invoice, itemized per instance.
Ingress is free, always. Every server includes 20 TB of egress per month, pooled across your account, and object storage adds free egress up to 3× your stored volume. If you routinely push more than that, talk to us about a flat-rate commit — there is no surprise tiered pricing waiting at the end of the month.
Reserved capacity runs on 1, 3, 6, or 12-month terms, billed monthly, with discounts that reach about 35% below on-demand at 12 months — $1,179/mo for an H100 SXM, roughly $1.62/hr effective. You can switch a running instance from on-demand to reserved without redeploying, and mixed fleets are normal: reserve the baseline, burst on-demand.
All payments are processed by Stripe: major credit and debit cards for self-serve accounts, plus ACH debit and wire transfer for invoiced enterprise accounts. Your $10 free credit is consumed before your card is ever charged, and you can set a hard monthly spend cap from the billing console.
Everything on this page is on-demand capacity in Reviosa-owned data centers — no waitlists, no preemption. An interruptible tier for RTX and L40S SKUs is in closed beta; for sustained workloads, reserved terms are the better long-run price and your jobs never get reclaimed mid-epoch.
24/7 coverage by infrastructure engineers — there is no tier-one script to escalate past. Production-down issues get a first response within 15 minutes; reserved and enterprise accounts add a shared Slack channel. The platform carries a 99.9% uptime SLA with automatic service credits when we miss it.
Start training in minutes
Create an account, add a card, and launch your first GPU instance. Per-second billing means you only pay for what you use.
No minimum commitment · Cancel anytime · $10 free credit for new accounts