Pricing

One fixed monthly fee. Stable performance from your AI factory.

Start on infrastructure from GEKI in a European data center: tokens, models and AI agents at the click of a button. One monthly fee. GEKI runs operations.

✗ No per-user fees ✗ No per-token markup ✓ One fixed monthly fee

The monthly fee covers inference, model hosting, agents, monitoring and support - no user licences, no token markup. The bill doesn't grow with every user or request. Optional: sell unused capacity to reduce the fee.

Fully managed

GEKI AI factory

Start with no investment - GEKI provides hardware & data center.

Fully managed, from €1,990 / month 1× RTX 6000 Blackwell (Mini), all included - power, EU hosting, GEKI operations & the full AI stack. Larger packages below.

GEKI procures, hosts and operates the GPUs and the full AI stack - you just use tokens and agents, without buying or running hardware.

✓ Full AI stack: operations, models, APIs, agents
✓ GPUs + hosting in the EU by GEKI
✓ Procurement & capacity planning
✓ No hardware purchase, no ops team

AI factories

Currently available AI factories

Dedicated GPU capacity in a European data center, fully managed. Two hardware lines, three sizes each - a pro-rata 8 CPU cores and 96 GB RAM per GPU.

NVIDIA RTX 6000 Blackwell

Package	Hardware	VRAM	Price / month
Mini	1× RTX 6000 Blackwell	96 GB	€1,990
Standard	2× RTX 6000 Blackwell	192 GB	€3,980
Large	8× RTX 6000 Blackwell	768 GB	€15,920

AMD Instinct MI350P

Package	Hardware	VRAM	Price / month
Mini	1× AMD Instinct MI350P	144 GB	€2,200
Standard	2× AMD Instinct MI350P	288 GB	€4,400
Large	8× AMD Instinct MI350P	1,152 GB	€17,600

All prices excl. VAT. 36-month minimum term. A pro-rata 8 CPU cores and 96 GB RAM per GPU. Optional: peak capacity & token resale.

Token economics

Recommended hardware and effective token prices.*

For each model the right hardware, the expected token volume per month and the effective price per 1M tokens - incl. GEKI operations. Guide figures at high utilisation; actual output and cost depend on workload, context and utilisation.

Model	Recommended hardware	Tokens / month	Effective price / 1M tokens
Gemma 4	1× RTX PRO 6000	approx. 20B	approx. €0.10
Qwen 3.6	1× RTX PRO 6000	approx. 15B	approx. €0.13
DeepSeek V4 Flash	2× AMD MI350P	on request	on request
GLM 5.2	8× H200	approx. 9B	approx. €0.70

* Guide figures on recommended hardware at high utilisation - actual token output and cost depend on workload and setup.

Why token cost looks like this

GEKI bills for the infrastructure used, not for token markups.

It gets cheaper when setup and workload fit: the right GPU, the right model, routing by cost and quality - and no frontier model where an open model is enough.

Optional capacity services

Optional with every AI factory.

Selling unused capacity as tokens

When your dedicated GPU capacity isn't fully used, GEKI can optionally sell unused capacity as tokens on the spot market. Your own workloads always take priority; spot-market traffic only runs on free capacity and is displaced as soon as your demand rises. The proceeds from sold capacity are shared between GEKI and the customer.

Peak capacity management

If your workloads temporarily exceed the dedicated capacity, GEKI can optionally route peaks to vetted European token providers - configurable by policy. Peak capacity is bought at market price.