OVHcloud GPU servers: A10, L4, L40S, H100 prices per hour
We may earn commissions when you shop through the links below. Learn more
French public cloud with per-second GPU instances, managed AI Deploy and a serverless LLM API.
- From
- ≈ $0.798 (converted from the EUR list price, €0.70)€0.70
- GPU families
- 5
- Regions
- 8
- Billing
- Per second
In short
OVHcloud is a French cloud provider (founded 1999, head office in Roubaix). Its Public Cloud sells NVIDIA GPU VMs (H100 PCIe, L40S, L4, A10, A100 80GB, 8x H200 NVL and older V100/V100S), mostly from Gravelines in France. Instances are billed per second with outbound traffic included.
Read full summaryShow less
Managed AI tools (AI Notebooks, AI Training, AI Deploy with per-minute billing and scale-to-zero) and AI Endpoints, an OpenAI-compatible pay-per-token API hosted in France, sit alongside them. Prices shown are the EU (EUR) list converted at the ECB rate of 25 Sep 2026.
Pricing
Billing: per second, per minute, monthly, usage-based, serverless. Lowest pay-as-you-go GPU price: ≈ $0.798 (converted from the EUR list price, €0.70)€0.70 per GPU-hour (V100 16GB, on-demand). Lowest plan price: ≈ $5.12 (converted from the EUR list price, €4.49)€4.49 per month (VPS).
GPU offers
| Price type and configuration | Region | Billing | Availability | Price per GPU-hour | Checked · source |
|---|---|---|---|---|---|
| NVIDIA H100 PCIe 80GB | |||||
| On-demand | EU | Per second | ≈ $3.19 (converted from the EUR list price, €2.80)€2.80 | ||
| Serverless, flex | EU | Per minute | ≈ $3.53 (converted from the EUR list price, €3.10)€3.10 | ||
| NVIDIA L40S 48GB | |||||
| On-demand | EU | Per second | ≈ $1.60 (converted from the EUR list price, €1.40)€1.40 | ||
| Serverless, flex | EU | Per minute | ≈ $1.77 (converted from the EUR list price, €1.55)€1.55 | ||
| NVIDIA A10 24GB | |||||
| On-demand | EU | Per second | ≈ $0.867 (converted from the EUR list price, €0.76)€0.76 | ||
| Serverless, flex | EU | Per minute | ≈ $1.03 (converted from the EUR list price, €0.90)€0.90 | ||
| NVIDIA L4 24GB | |||||
| On-demand | EU | Per second | ≈ $0.855 (converted from the EUR list price, €0.75)€0.75 | ||
| Serverless, flex | EU | Per minute | ≈ $0.946 (converted from the EUR list price, €0.83)€0.83 | ||
Show all 9 offers (from ≈ $0.798 (converted from the EUR list price, €0.70)€0.70 per GPU-hour)Show fewer offers
| Price type and configuration | Region | Billing | Availability | Price per GPU-hour | Checked · source |
|---|---|---|---|---|---|
| NVIDIA V100 16GB | |||||
| On-demand | Varies | Per second | ≈ $0.798 (converted from the EUR list price, €0.70)€0.70 | ||
NVIDIA H100 PCIe 80GB
- On-demand≈ $3.19 (converted from the EUR list price, €2.80)€2.80
- Serverless, flex≈ $3.53 (converted from the EUR list price, €3.10)€3.10
NVIDIA L40S 48GB
- On-demand≈ $1.60 (converted from the EUR list price, €1.40)€1.40
- Serverless, flex≈ $1.77 (converted from the EUR list price, €1.55)€1.55
NVIDIA A10 24GB
- On-demand≈ $0.867 (converted from the EUR list price, €0.76)€0.76
- Serverless, flex≈ $1.03 (converted from the EUR list price, €0.90)€0.90
NVIDIA L4 24GB
- On-demand≈ $0.855 (converted from the EUR list price, €0.75)€0.75
- Serverless, flex≈ $0.946 (converted from the EUR list price, €0.83)€0.83
Show all 9 offers (from ≈ $0.798 (converted from the EUR list price, €0.70)€0.70 per GPU-hour)Show fewer offers
NVIDIA V100 16GB
- On-demand≈ $0.798 (converted from the EUR list price, €0.70)€0.70
Research notes for these offers
Compare these prices with other providers’: H100 prices comparedL40S prices compared.
Plans
VPS
- VPS-1 (VPS 2027 range)≈ $5.12 (converted from the EUR list price, €4.49)€4.49
- VPS-2 (VPS 2027 range)≈ $9.68 (converted from the EUR list price, €8.49)€8.49
- VPS-3 (VPS 2027 range)≈ $13.96 (converted from the EUR list price, €12.24)€12.24
- VPS-4 (VPS 2027 range)≈ $26.79 (converted from the EUR list price, €23.49)€23.49
Serverless endpoint
- AI Endpoints (serverless LLM API, pay per token)Not verified
Show all 13 plansShow fewer plans
GPU instance
- t1-le-45 (1x Tesla V100 16GB, Limited Edition)≈ $574.71 (converted from the EUR list price, €504.00)€504.00
- l4-90 (1x L4 24GB)≈ $615.76 (converted from the EUR list price, €540.00)€540.00
- a10-45 (1x A10 24GB)≈ $632.64 (converted from the EUR list price, €554.80)€554.80
- t2-le-45 (1x Tesla V100S 32GB, Limited Edition)≈ $656.81 (converted from the EUR list price, €576.00)€576.00
- l40s-90 (1x L40S 48GB)≈ $1,149.42 (converted from the EUR list price, €1,008.00)€1,008.00
- a100-180 (1x A100 80GB, form factor not published)≈ $1,254.33 (converted from the EUR list price, €1,100.00)€1,100.00
- h100-380 (1x H100 PCIe 80GB)≈ $2,212.18 (converted from the EUR list price, €1,940.00)€1,940.00
- h200-1920 (8x H200 NVL 141GB)≈ $34,961.60 (converted from the EUR list price, €30,660.00)€30,660.00
Research notes for these plans
Storage and data transfer
- Block Storage Classic
- ≈ $0.049 (converted from the EUR list price, €0.043)€0.043
- Block Storage High Speed Gen2
- ≈ $0.099 (converted from the EUR list price, €0.087)€0.087
- File Storage Standard
- ≈ $0.161 (converted from the EUR list price, €0.142)€0.142
- Object Storage Standard (Multi-Zone), first 50 TiB
- $0.015≈ €0.013 (estimated conversion from $0.015)
- Object Storage Standard (One Zone)
- $0.0075≈ €0.0067 (estimated conversion from $0.0075)
- Local NVMe included with GPU flavors
- $0.00≈ €0.00 (estimated conversion from $0.00)
- Egress (data out)
- $0.00≈ €0.00 (estimated conversion from $0.00)
Before you compare
- Suited to
- GPU compute that must stay in the EU (France), with published per-second list prices
- Teams that want VMs, managed GPU containers (AI Deploy) and a serverless LLM API from one European provider
- Main limitation
- On the EU price list, modern GPUs (H100, A100, L40S, L4, A10) are sold only in Gravelines (GRA9/GRA11) and H200 only in Paris 3-AZ; Frankfurt, London and Warsaw list only older V100S, and the separate US entity lists only L40S, L4 and V100S (Virginia and Oregon)
- Price basis
- GPU prices per GPU-hour, billed per second; egress included; 6 storage options listed
Who it fits
Suited to
- GPU compute that must stay in the EU (France), with published per-second list prices
- Teams that want VMs, managed GPU containers (AI Deploy) and a serverless LLM API from one European provider
- Inference on L4, L40S or H100 PCIe without egress charges
- Self-hosting AI agents (OpenClaw, Hermes Agent, n8n) on low-cost VPS with an EU-hosted model API
Good fit for
- Private LLM hostingEU-hosted GPUs in France with dedicated passthrough GPUs, customer-managed/BYOK volume encryption, a published DPA, and AI Endpoints stating that prompts are not stored. Whether GPU instances fall inside the SecNumCloud-qualified scope was not verified.
- TrainingSingle-node training on up to 4x H100 PCIe, 4x A100 80GB or 8x H200 NVL, plus per-minute AI Training jobs. No multi-node GPU cluster or InfiniBand offering is listed for Public Cloud GPU; the GPU page quotes up to 25 Gbit/s private networking.
- Fine-tuningPer-second billed H100 PCIe (about $3.19/h) and L40S (about $1.60/h) instances plus AI Notebooks/Training with framework images; OVHcloud publishes fine-tuning tutorials on AI Notebooks.
- Batch inferencePer-second instances with included egress, per-minute AI Training jobs, and an AI Endpoints batch mode for supported models.
Show 4 moreShow fewer
- Real-time inferenceAI Deploy offers autoscaling (including scale to zero, with 30 s+ cold starts) on L4/L40S/H100, and AI Endpoints serves an OpenAI-compatible API limited to 400 requests/min per model per project. Modern GPUs are available only in Gravelines.
- AI agentsOfficial guides for OpenClaw and Hermes Agent on VPS wired to AI Endpoints, a pre-installed n8n VPS image, and VPS plans from EUR 4.49/month without commitment.
- RAG applicationsAI Endpoints provides embedding models (bge-m3, Qwen3-Embedding-8B) and LLMs, with official RAG tutorials using LangChain and LangChain4j; S3-compatible Object Storage has free egress.
- RenderingL40S, L4 and Quadro RTX 5000 flavors are positioned for 3D and video. Windows pricing is listed only for L4, V100/V100S and RTX 5000 flavors (the GPU deployment guide still names only T1/T2 for Windows images), and the RTX 5000 flavors are tagged 'coming soon' in the order catalog and are not in the GPU catalog.
Look elsewhere for
No workload is rated a partial fit or not suitable.
Not sure OVHcloud fits what you want to build?Help me choose
Strengths and limitations
Strengths
- Transparent public price list and order-catalog API; GPU instances metered per second after a 1-minute minimum
- Outbound public traffic included for instances outside Asia-Pacific; Object Storage egress free
- Company-wide compliance portfolio per OVHcloud (ISO 27001/27017/27018/27701, SOC 1/2/3 Type 2; HDS, C5 and SecNumCloud for specific services, scope per product not published) and a published DPA
- BYOK and customer-managed keys via OKMS for encrypted Block Storage
Show 2 moreShow fewer
- Managed AI stack hosted in France: AI Notebooks, AI Training, AI Deploy (autoscaling to zero) and OpenAI-compatible AI Endpoints
- 99.99% monthly availability SLA stated for GPU instances
Limitations
- On the EU price list, modern GPUs (H100, A100, L40S, L4, A10) are sold only in Gravelines (GRA9/GRA11) and H200 only in Paris 3-AZ; Frankfurt, London and Warsaw list only older V100S, and the separate US entity lists only L40S, L4 and V100S (Virginia and Oregon)
- No spot or preemptible pricing, and Savings Plans exclude GPU instances; monthly billing saves about 1-5% on most GPU flavors (about 45% on A100)
- At most 4 GPUs per instance (8x H200 NVL); H100 is the PCIe variant and no multi-node GPU cluster is listed on Public Cloud
- Separate EUR, Worldwide-USD and US price lists; the USD list is not a straight FX conversion of the EUR list
Show 2 moreShow fewer
- Stopped or paused instances are billed until deleted; a public IPv4 address is included by default (an additional Public or Floating IP is billed)
- A10 and Quadro RTX 5000 flavors are tagged 'coming soon' in the order catalog, and V100/V100S 'LE' flavors are a limited-edition, while-stocks-last offer
Infrastructure
Regions
8 regions across 6 countries; 4 are recorded as EU regions.
- Europe (5)
- Frankfurt (Limburg) · Germany · DE1 (EU); Gravelines · France (EU); London (Erith) · United Kingdom · UK1; Paris (3-AZ) · France · EU-WEST-PAR (EU); Warsaw · Poland · WAW1 (EU)
- North America (3)
- Hillsboro (labelled 'Seattle' on the regions page) · United States; Montreal area (Beauharnois, BHS) · Canada · BHS5; Vint Hill (labelled 'Washington DC' on the regions page) · United States
Region notes
Frameworks
How OVHcloud works with common AI serving, agent and training frameworks. Support levels, strongest first: an official template, documentation, a community guide, or manual setup on a general-purpose machine.
Hosting
- n8n
- Official template
- PyTorch
- Official template
- Hermes Agent hosting providers
- Documented setup
- Ollama hosting providers
- Documented setup
- OpenClaw hosting providers
- Documented setup
- vLLM hosting providers
- Documented setup
- Dify
- Manual setup
- Flowise
- Manual setup
- LangGraph
- Manual setup
- SGLang
- Manual setup
- Text Generation Inference
- Manual setup
Operations
Managed service: yes. Each feature below states what the provider documents, with the source it was checked against. “Not verified” means no reliable source was found.
Features
Compute and access
- Custom containers (Docker)
- Yes
- Root access
- Yes
- SSH access
- Yes
- API access
- Yes
- Jupyter notebooks
- Yes
- Templates
- Partial
- Multi-GPU instances
- Yes
- Bare metal
- Yes
- Kubernetes
- Yes
Show all 21 featuresShow fewer features
Management
- Managed service
- Yes
- Serverless
- Yes
Storage and data
- Persistent storage
- Yes
- Snapshots
- Yes
- Managed backups
- Yes
Billing
- Per-second billing
- Yes
- Spot pricing
- No
- Reserved pricing
- Partial
Regions and compliance
- EU region
- Yes
- SOC 2 report
- Yes
- GDPR data processing agreement
- Yes
- Customer-managed keys (BYOK)
- Yes
SLA and support
- Uptime SLA
- 99.99% uptime
- Support channels
- 4 listed
Sources and verification
How we verify
Prices, regions, storage, features, framework support, SLA and support channels link to the page they were read from, and every price shows when it was checked. Fit labels, strengths and limitations summarise the same research record. The OVHcloud record was last checked on 26 Sep 2026. A check date older than 7 days is marked “Stale” with an amber dot, and one older than 30 days “Needs refresh”.
- Last verified
- 26 Sep 2026, 19:09 UTC
- Method
- Source-by-source review by AI agents, independently re-checked
- Confidence
- High
- Checked by
- AI research agent; every value re-checked against its source by a separate AI agent
Verification notes
On 26 September 2026 we checked every price in this listing against OVHcloud's order catalog and live price pages, converting EU list prices at the ECB rate of 25 September 2026 (1.1403). OVHcloud's Worldwide USD list is not a straight conversion of the EU list; its figures are in the offer notes. Not verified: the A100 form factor, the H200 variant, whether A10 and RTX 5000 can be ordered yet, and live GPU stock. Later corrections are listed on the methodology page.
Commercial relationship
AIHostCompare has no active affiliate link with OVHcloud; the Visit provider link goes to its official website. Commissions never change ordering, alternatives, fit labels or conclusions. Read the affiliate disclosure.
Pricing
Show all 41 sourcesShow fewer sources
Documentation
Terms & policies
Status
Secondary sources
Alternatives to OVHcloud
| Fact | This providerOVHcloud | DigitalOcean profile | Modal profile | Nebius profile |
|---|---|---|---|---|
| From | ≈ $0.798 (converted from the EUR list price, €0.70)€0.70 | $1.57≈ €1.40 (estimated conversion from $1.57) | $0.59≈ €0.527 (estimated conversion from $0.59) | $1.55≈ €1.38 (estimated conversion from $1.55) |
| GPU families | H100 · L40S · A10 · +2 more | MI300X · H200 · H100 · +2 more | B200 · H200 · RTX PRO 6000 · +6 more | B200 · H200 · H100 · +1 more |
| EU regions | 4 in the EU | 2 in the EU | None in the EU | 4 in the EU |
| Billing increment | Per second | Per second | Per second | Per second |
| Egress | No egress feesSource | $0.0093≈ €0.0083 (estimated conversion from $0.0093) Source | $0.037≈ €0.033 (estimated conversion from $0.037) Source | No egress feesSource |
- DigitalOcean profile$1.57≈ €1.40 (estimated conversion from $1.57)
- Modal profile$0.59≈ €0.527 (estimated conversion from $0.59)
- Nebius profile$1.55≈ €1.38 (estimated conversion from $1.55)
Related comparisons and guides
- Compare AI hosting and GPU cloud providers
- GPU Cost Calculator
- VRAM / Model Fit
- Agent Hosting Selector
- How to choose a GPU cloud provider
- How much VRAM do you need to run an LLM?