Speaking with TFiR, Dirk Alshuth of emma explained how GPU scarcity drives infrastructure overprovisioning and how emma’s professional services approach addresses the mismatch between customer requests and actual AI workload requirements.

What Is emma’s AI Infrastructure Play?

emma expanded beyond traditional CPU and VM provisioning to offer GPU access for AI workloads. The platform sources accelerators from emma’s Luxembourg data center, hyperscaler GPU offerings, and specialized AI cloud providers, with plans to add inference-specific workflows.

Dirk Alshuth: “In emma, we have traditionally provided the services around CPUs and VMs. Now we are also integrating the possibility to provide organizations with the compute power for AI—being able to source and access GPUs from different providers, including our own in Luxembourg at our data center, provisioning them, managing them, optimizing them. And we’re also bringing workflows for inference in future.”

Broader Context

Alshuth identified GPU scarcity as a driver of infrastructure misallocation. When accelerator capacity is limited, enterprises request larger GPUs than their workloads require to secure any access at all. emma’s professional services team intervenes before provisioning to match requests against actual requirements.

Dirk Alshuth: “Accessibility is the most important word. When you have access to something, you can commoditize it. One of the things we see from the many conversations that we have—because scarcity makes people want to talk to you, want to see what you can help them with—is that the requests we get very often do not match what people really need. It’s larger accelerators requested, but actually the need is for smaller accelerators. So we help with our professional services to balance that and see what we have in our data center to provide them with, or where we can provide access to other cloud providers.”

emma offers two engagement models. Self-service deployment is available for teams that already understand their infrastructure requirements and can provision GPUs directly through the platform. For complex or unfamiliar use cases, emma’s professional services team engages with customers to assess workload characteristics, determine appropriate accelerator capacity, and recommend deployment strategies.

Dirk Alshuth: “There are two ways of working with emma. One is self-service. If you know what you’re doing, if you know what you need, if you know what you want, sign up, you get your self-service, you can deploy. But then the use cases where our professional services team is engaging with the customer and making sure to understand what exactly do they need, why they’re asking for specific infrastructure, how would they manage and operate that. We go as far as helping them determine where they need to deploy and how to deploy, or what kind of processes and workflows they should deploy to make it run smoothly and cost efficiently.”

Alshuth outlined typical use cases for emma’s platform. One common pattern involves enterprises with large hyperscaler footprints seeking leverage by diversifying to additional cloud providers. emma’s platform teams enable rapid deployment on new providers without requiring deep provider-specific expertise. Another key differentiator is cloud cost management across on-premises infrastructure, hyperscalers, and European sovereign cloud providers under unified FinOps visibility.

Dirk Alshuth: “We focus on cloud cost management, but the capability of bringing your on-prem, your clouds, your different clouds, even the European sovereign clouds, into that cloud cost management thinking. You will see what you have on-prem, you will see your hyperscalers—which is today nothing very special anymore—but the fact that you can also bring, for European companies very important, the European cloud providers under this cost management scheme is a differentiator.”

emma has approximately 90 employees, with the majority serving engineering and professional services functions. The company maintains a global alliance with PwC, which uses emma as the cloud operations platform for enterprise transformation projects. Alshuth emphasized that emma’s engagement model prioritizes advising customers rather than simply provisioning infrastructure—helping organizations determine optimal deployment strategies and operational workflows alongside resource provisioning.

Watch the full TFiR interview with Dirk Alshuth here.