
The next tranche of compute is already on the management board agenda: clusters, the site, and whether that capacity should run under a sovereign arrangement. Who will set the scale and the location and make the case for both before the management board? That is what you are appointing when you fill a Head of AI Infrastructure seat.
Anyone who brings models into operation on an existing environment is working on machines that are already in place. That job can serve teams and keep pipelines moving. It leaves open which accelerator volume the company will buy, lease or reserve, and where that equipment will stand. If you treat that operational work as the investment case, the proposal has no owner. The platform side of this split is covered in Head of AI Platform search.
Power, space and cooling belong in the same proposal. Utilisation can be reported from a running operation. That report does not tell you whether the next spend should go into owned space, a leased bay or a public cloud. Someone has to put that choice forward and stand behind the approval. Scheduling work onto equipment you already have is operations. Clearing space and power for what you do not yet have is the investment.
In a mid-sized company the same investment case can appear as rented volume or as owned space. The job is still to make the case for scale and location before the management board. A sovereign requirement is a siting choice, not a legal obligation. A public region does not make that choice by itself.
Set the target architecture beside this hire and the gap is plain. The architecture function sets the model families and the integration. That choice still leaves power on the slab unreserved and budget uncommitted. Moving applications into a cloud programme changes how operations run. That move leaves unsettled how much dedicated AI compute you will hold, or whether it must remain at a given data centre.
Product engineering and the internal stack can sit with the CTO. That still leaves open who will carry the next accelerator tranche and its location into the board meeting. If you only extend the CTO remit with this word, the capital proposal often has no voice.
The management board has to state how much compute this hire will put forward, where the equipment will stand and how a sovereign requirement enters the proposal. The seat reports to the management board, because that is where the budget is approved.
What the management board should hear is which compute this person previously attached to a location, and how that capacity stayed available once the budget had been approved. After approval, the same person still carries the gap between the decision and usable power on the slab. That gap belongs with this appointment.
Commission AI executive search for this appointment in Germany.




