NVIDIA AI ENTERPRISE · vGPU · LICENSING

NVIDIA AI Enterprise
and vGPU licences

The licences are not counted the way most people assume: per GPU in one case, per concurrently active user in the other. Getting that wrong is expensive in both directions.

Request a quote Talk to an expert

AVAILABLE TO ORDER

What we license

These are the licences that decide whether the card you bought can do the job you bought it for. We put them in the quote with the hardware, not after it.

Licence	Counted by	Needed for	Note
NVIDIA AI Enterprise	Per GPU	The supported AI software stack and compute vGPU, on NVIDIA-Certified servers	Five years included with each H200 NVL; bought per GPU for L40S, L4 and RTX PRO.
RTX vWS (Virtual Workstation)	Per concurrently active user	Full workstation graphics inside a VM	Not per card: this is where budgets get misread.
Virtual PC / Virtual Apps	Per concurrently active user	Desktop and application streaming	vApps streams one display at 1280×1024.
MIG	Not licensed on bare metal	Hardware partitioning of the card	Free on bare metal; MIG-backed vGPU in VMs needs the vGPU licence. MIG instances on different GPUs cannot talk peer to peer, over PCIe or NVLink, and NCCL does not work with MIG.

All licences above are available to order. Lead time is confirmed with every quote.

WHAT COMES WITH IT

Not just a box on a pallet

EU contract and invoice

One counterparty in Vienna. No cross-border purchase to arrange, no import to clear on your side.

Licences sized with the hardware

Where a card needs a licence to do the job you bought it for, it is in the quote, not discovered three months later.

Engineering, if you want it

Our partner Vixen.UNO designs, deploys and supports it. Under the same contract, or not at all. Your call.

Warranty and support

Manufacturer warranty handled through us, with support alongside the team who runs the system.

HOW TO CHOOSE

Which one is right for you

Without a licence, the card works for twenty minutes

Then a graphics vGPU is degraded, and degraded further if it still has no licence after 24 hours, as NVIDIA’s current client licensing guide (29 September 2026) describes it; a compute (C-series) vGPU drops to idle-level compute. For a compute vGPU, NVIDIA states that full capability returns as soon as a valid licence is acquired. This is documented behaviour, not a fault, and it is the single most common surprise we get called about.

Read the full analysis →

MIG or vGPU: they slice different things

MIG cuts the card in hardware: each instance gets its own compute blocks and its own memory bandwidth. vGPU slices time and needs a licence: per concurrent user for graphics, per GPU for compute. RTX PRO 6000 Blackwell splits into up to four 24 GB instances; the L40S has no hardware partitioning at all.

Read the full analysis →

How many seats you actually need

NVIDIA’s sizing guide gives two to four heavy users per RTX PRO 6000 Server Edition and six to eight light users per RTX PRO 4500 Server Edition. It also notes the best-effort scheduler often allows two- to threefold oversubscription when users are not all active at once.

Read the full analysis →

HONESTLY

When we are not the right fit

✕You want licences without the hardware and without a conversation. We can do it, but we will ask what it is running on first. Half the licence mistakes we see start there.

✕You need a licence type we have not listed. Ask, but if it sits outside what we can properly supply, we will say so rather than take the order.

FAQ

Licensing, in short

Is AI Enterprise per GPU or per server?

Per GPU, and that includes compute (C-series) vGPUs, which NVIDIA licenses only through AI Enterprise. The graphics vGPU licences, RTX vWS, vPC and vApps, are the ones counted per concurrently active user; that difference is the most common budgeting error we see.

Do I need a licence for MIG?

Not on bare metal: MIG is hardware partitioning and needs no licence there. Handed to VMs as MIG-backed vGPU, the slices need the vGPU licence (AI Enterprise per GPU for compute, vWS per user for graphics). And MIG instances on different GPUs cannot talk peer to peer, over PCIe or NVLink, and NCCL does not work with MIG.

What happens when a licence expires?

A VM that boots without a licence runs twenty minutes at full capability. A compute (C-series) vGPU then drops to idle-level compute; a graphics vGPU is degraded, and degraded further if it still has no licence after 24 hours, as NVIDIA’s current client licensing guide describes it. A licensed VM that loses its licence server keeps full speed for up to one day by default for graphics vGPU, and for seven days for compute (C-series) vGPU on the cloud licence service, then degrades; on an on-premise DLS the compute period is configurable. For a compute vGPU, NVIDIA states that full capability returns as soon as a valid licence is acquired.

Which hypervisors support MIG-backed vGPU?

vSphere, Red Hat KVM and Ubuntu KVM, for graphics profiles from release 19.0; on vSphere, time-sliced vGPUs inside a MIG instance need VMware Cloud Foundation 9.1 with release 19.4 or 20.0 onwards. On Windows Server Hyper-V those profiles are not available. On Proxmox, MIG instances are created by hand from the CLI.

Can I buy licences and hardware on one contract?

Yes, that is the point. One EU counterparty, one invoice, and the licence term visible before you order rather than after.

RELATED DEEP DIVE