Summary
- OVHcloud has unveiled 55 cloud and AI products, comprising 38 cloud infrastructure and related services and 17 AI products.
- Its catalogue lists 21 generally available products, seven beta, 20 alpha and seven forthcoming.
- Nvidia B300 infrastructure is listed as available, while Rubin NVL8 systems remain forthcoming and the company’s larger GPU figures describe potential capacity.
OVHcloud has expanded its European cloud infrastructure portfolio with Nvidia Blackwell GPU computing, storage and network services, as part of 55 product launches presented at its AI Summit in Paris on 19 November.
The French provider has introduced products covering compute, data storage, security, networking and private cloud infrastructure, alongside additional AI services. The programme includes 38 cloud infrastructure and related products and 17 specifically associated with AI computing and applications.
Availability varies substantially. OVHcloud’s official product catalogue lists 21 offerings in general availability, seven in beta, 20 in alpha and seven described as coming soon. The 55 launches therefore represent a combination of production services, early development releases and future products rather than 55 fully available commercial offerings.
The infrastructure expansion includes Nvidia B300 GPUs, listed as available through Bare Metal and Public Cloud services. A future Nvidia Rubin NVL8 configuration is also included, although its catalogue status remains forthcoming.
OVHcloud says its existing data centre capacity could accommodate up to 60,000 B300 GPUs or 35,000 Rubin R200 GPUs. Reuters reported that the company has approximately 110MW of available capacity, but neither figure establishes that the corresponding accelerators have been installed and commissioned.
The company is expanding its European-hosted offering as businesses and public authorities assess the location, governance and technical dependencies of their computing infrastructure. Product availability and the physical resources supporting those products will determine how quickly the portfolio can be adopted.
Computing capacity depends on network and storage
Large GPU deployments require substantial supporting infrastructure. Accelerators must exchange information with processors, memory, storage systems and other GPUs, while the facility supplies electricity and removes the heat generated by the equipment.
OVHcloud’s new B300 options extend its available computing hardware, but the practical performance of a cluster depends on its configuration. The capacity of individual processors does not establish the throughput available to distributed applications if network or storage resources become limiting factors.
Network connectivity influences how workloads communicate within a data centre and across cloud environments. Applications distributed over several systems may require private connectivity, controlled routing and the ability to balance traffic between different computing resources.
OVHcloud has outlined an expansion of its virtual private cloud architecture, including network peering, application load balancing and private DNS. However, its product list identifies the VPC, Application Load Balancer and Private DNS offerings as forthcoming, rather than generally available.
Some related capabilities have reached production. OVHcloud Connect Edge and private access to the managed Kubernetes service interface are listed as available, while a new network traffic and status visualisation tool is also in general availability.
Other functions remain under development. The provider’s BGP service is in beta, while site-to-site VPN, Global Load Balancer and web application firewall offerings are in alpha. The proposed expansion of its global points of presence is also marked as coming soon.
These distinctions affect how customers can design infrastructure. A service requiring a particular routing or security feature cannot necessarily be moved into production merely because that feature appears in a supplier’s announced roadmap.
Storage has received a similar staged expansion. OVHcloud has made an updated object storage metadata system generally available, while private S3-compatible object storage endpoints remain in beta and its Backup Center service is in alpha.
Metadata management influences how stored objects are located and accessed. Improvements can support workloads making large numbers of storage requests, although the resulting application performance still depends on storage devices, network connections and software configuration.
Private storage endpoints provide a means of controlling connectivity to data services. They can reduce reliance on publicly exposed network paths, but security also depends on authentication, access policies, encryption and the configuration of the surrounding infrastructure.
European hosting requires operational assurance
OVHcloud is also expanding security and governance options. Customer-controlled encryption keys for storage volumes are listed as available, while cloud hardware security modules remain in beta and confidential computing through OCCAM is in alpha.
The different development stages create practical procurement questions. Customers assessing regulated workloads need to know which services have completed release testing, what contractual commitments are available and whether the necessary controls operate together in a supported production environment.
European hosting can help organisations address data residency requirements, but the geographical position of a server does not independently establish sovereignty or regulatory compliance. Administrative access, subcontractors, encryption arrangements and dependencies on external services also influence the assessment.
The company has introduced additional private cloud and on-premises products, including a generally available SecNumCloud-qualified cloud platform. Other private cloud products remain at alpha or beta stage, limiting the configurations customers can deploy immediately.
New accelerator architectures also create physical infrastructure requirements. Higher equipment power density can increase the demands on electrical distribution and cooling, while associated network upgrades may require changes to switching equipment and optical connections.
Facility capacity figures cannot be equated directly with GPU availability. Electrical headroom must be matched with suitable space, cooling arrangements, equipment deliveries, networking and commissioning before a computing system becomes usable.
The company’s stated potential to accommodate tens of thousands of new GPUs therefore provides an indication of its infrastructure ambitions, not an audited count of operational processors or a timetable for completing every installation.
OVHcloud’s latest portfolio broadens the computing and supporting services offered through its European infrastructure. Subsequent product releases, verified hardware availability and the progression of alpha and beta services into general availability will establish how much of the new portfolio customers can operate in production.

