From a growing network of AI factories to Starling developing a banking model on sovereign UK cloud infrastructure, there’s plenty to be excited about in European AI right now. Behind these developments is a growing focus on the computing infrastructure needed to support them.
Since NVIDIA’s GTC Paris last year, we’ve seen plans for new infrastructure move into operation, giving businesses more opportunities to put AI to work. Jensen Huang has also outlined plans earlier this week for a further $2 billion in European AI deals, building on $8 billion over the past year.
For us here at Vespertec, that momentum is encouraging. More regional computing capacity gives organisations greater choice, while creating opportunities for the engineering teams bringing AI infrastructure into operation.
With GTC Berlin approaching, it’s a good time to look at how far Europe has come since Paris, and what it means for the organisations building and using AI infrastructure across Europe.
From announcements to working infrastructure
At GTC Paris, NVIDIA set out plans to expand Europe’s AI capacity through large-scale GPU deployments, with projects serving both AI developers and industrial users.
French AI unicorn Mistral outlined plans for a cloud platform with 18,000 Blackwell and Blackwell Ultra GPUs in its initial deployment. Mistral now reports that its AI Cloud is serving production workloads and has onboarded external customers.
Deutsche Telekom’s Industrial AI Cloud is also operational, bringing nearly 10,000 NVIDIA Blackwell GPUs into service through DGX B200 systems and RTX PRO Server GPUs.
These developments give European organisations additional options for accessing AI computing capacity and locating their workloads closer to home. The next consideration is which services suit their technical requirements, budgets and operating needs.
What this means for Europe
Manufacturing, engineering and scientific research are long-established European strengths. Accessible AI infrastructure offers new ways to build on them, from running engineering simulations to factory digital twins and training research models.
Regional capacity also gives businesses more options when deciding where to run workloads and how to manage sensitive data. Location alone does not resolve every question about sovereignty, security or governance, but it is an important part of the decision.
UK Bank Starling’s recently announced project illustrates how those choices are entering deployment decisions. The bank is developing a transactional foundation model with NVIDIA on Civo’s UK cloud infrastructure, and linked the choice to its ambition to diversify providers and reduce concentration risk.
Organisations need a practical route from an idea to a service they can afford, operate and expand. Europe’s growing capacity broadens those possibilities and choosing the right deployment approach will help turn them into useful services.
Getting the infrastructure right
For organisations planning their next AI deployment, that means deciding which workloads to run through a cloud provider, which to host on their own infrastructure, and how those environments will work together.
At Vespertec, we start with the workload and bring architecture, procurement and budgets into the same conversation.
For teams building or expanding their own infrastructure, GPUs are one part of a complete system. Networking, storage, power and cooling all need to support the intended workload. Committing to GPUs before aligning those elements can tie up budget in equipment that cannot deliver its full potential.
Those decisions also affect future upgrades – what components can be retained, where capacity can be added, and what will need replacing. Planning deployment and future upgrades together will help organisations make the most of their investment while preventing avoidable redesigns.
Looking towards Berlin
We’ll be following GTC Berlin from our AI Lab here in Stockport, watching for details of the next investments and how organisations will access the European AI capacity.
Europe has considerable expertise to build on, and more of its AI infrastructure is entering operation. The opportunity now is to connect that capacity with well-defined workloads and practical deployment plans. For us at Vespertec, that means helping teams choose and build infrastructure that fits the work they want to do.