OVHcloud AI Endpoints: EU-hosted open models and GPU capacity

Contents

1. What OVHcloud offers on sovereignty
2. Strengths and limits
3. Pricing
4. Questions to ask OVHcloud
5. Sources

Type	European cloud provider
Pricing	Published
Website	ovhcloud.com

OVHcloud's AI Endpoints is a serverless API to more than 20 open-weight models, hosted and run in the EU. OVHcloud states that customer data is never used to train its models and that only what billing requires is kept⁠1AI EndpointsOVHcloud.

Best for: Developers who want open-weight models as an EU-hosted API, with GPU instances available when they need to run models themselves.

Not the right fit for: Companies without developers, or those that need a business application rather than infrastructure.

What OVHcloud offers on sovereignty

AI Endpoints is billed by input and output tokens, with batch mode at half the base rate, and OVHcloud notes that the same models can be deployed on the customer's own infrastructure or another cloud. It lists the service as ISO 27001, SOC 2 Type II and HDS certified⁠1AI EndpointsOVHcloud.

For companies that want to run models themselves, OVHcloud rents GPU instances with one to four GPUs: from $1.00 an hour for an L4 or A10, $1.80 for an L40S and $2.99 for an H100⁠2Cloud GPU instancesOVHcloud. The GPU page does not state where instance data is stored; its sovereignty references concern other OVHcloud products.

Strengths and limits

Strengths

• States that data is never used for training and that it keeps only what billing requires.
• The same models can be deployed on the customer’s own infrastructure or another cloud.
• A path from API to rented GPU instances, with published hourly prices.

Limits

• The GPU instance page does not state where those instances’ data is stored.
• Per-token prices sit on a separate page from the product description.
• Infrastructure and model access only.

Choose another option if: You want a business system that uses AI, not infrastructure to build one.

Pricing

Published: GPU instances from $0.60 an hour (Quadro RTX 5000) to $2.99 (H100) and $1.80 (L40S); AI Endpoints billed per token, batch at 50% of the base rate.

Questions to ask OVHcloud

• Which regions are GPU instances available in, and can we restrict them to the EU?
• What are the per-token prices for the models we plan to use?
• Which certifications cover GPU instances as distinct from AI Endpoints?
• Is a dedicated deployment of an AI Endpoints model available?

See how OVHcloud compares with the other options in the main guide.

Sources

1. AI Endpoints. OVHcloud. Accessed 8 Nov 2026.
2. Cloud GPU instances. OVHcloud. Accessed 8 Nov 2026.

Further reading

• Buyer's guide · Updated quarterlySovereign AI for business: the options compared for 2026For a company, sovereign AI is less about nations and data centres than about three questions: who owns the system, where the data lives and who could be compelled to hand it over. Seven options, compared on the evidence.
• Guide · CostWhat self-hosting AI costs a company in 2026Control has a price, but it is often lower than buyers expect, and the most expensive option is rarely the most sovereign. Worked examples from published European prices.
• Guide · DefinitionsSovereign vs private vs local AI: what each term meansFour terms that vendors use interchangeably describe four different guarantees. Knowing which one a company needs saves money and prevents false comfort.