Office with power for AI25 kW per half-floor, own meter
Power is becoming the bottleneck: AI servers, GPU workstations and test benches need capacity and cooling that ordinary offices don't have. At the PROJEKTCAMPUS near Stuttgart it is already in place – ready to use, with no grid connection process.
What the power is good for
- Your own AI server instead of the cloud: run language models and analytics in your own server room – your data stays in-house, which matters for TISAX and data protection
- GPU workstations for development: training, simulation and rendering right at the desk
- Test benches and ECUs: HiL set-ups, flashing and software testing – with a vehicle hall next door if needed
- Grow without moving: besides the server room, further rooms can host racks, and units can be combined
- Proven in practice: one tenant already runs an AI office on campus with around 100 kW of continuous power – across several meters
For reference: a GPU server with 8 graphics cards draws around 10 kW, a powerful GPU workstation 1–3 kW.
What companies will need more and more
The price per token is falling, but AI agents, longer contexts and automation make usage grow faster. That changes what an office has to deliver:
- Own AI hardware instead of a permanent subscription: above a certain volume your own server is cheaper than the API – it needs power, cooling and a secure room
- Data stays in-house: development data, source code and prototypes stay in your own alarm-protected server room instead of third-party clouds
- Small teams with a lot of compute: 5 people and a GPU server instead of 30 people in an open-plan office
- Quiet for voice AI: enclosed rooms for voice input and video calls
- Fast connectivity and short distances: fibre from 4 carriers, test benches and vehicle halls next door
Edge data centre & colocation
Many small units with their own power instead of one big data centre: the half-floors with 25 kW each and their own meter can be used individually or combined as an edge site – right in the engineering cluster of the Stuttgart region, with fibre from 4 carriers.
- Edge data centre: compute where the data is created – for AI inference, simulation, vehicle and software testing with low latency
- Colocation & private cloud: your own racks in closed, access-controlled and alarm-protected rooms – located in Germany, TISAX-compliant environment
- Distributed grid connection: almost 2 MW across 9 independent connection points, 3 more in reserve
- Grow step by step: from one 25 kW server room to several half-floors up to the underground car park with 250 kW
New: underground space with 250 kW
Our new underground car park offers around 1,000 m² with its own 250 kW connection – ideal for equipment that needs space, power and stable temperatures.
- 1,000 m² of space with its own 250 kW connection
- 3 m clear height for racks, cooling and cable trays
- High floor load capacity for server racks, battery storage and test benches
- Very good PUE achievable: being underground keeps temperatures low and stable all year round
Exact load and cooling figures on request.
The half-floor
- 200 m² for 5–20 people: 3 private offices, 2 team offices, meeting room, kitchenette and your own server room
- Separate and secure: own access-controlled, alarm-protected entrance per unit
- Ready from day one: fully furnished, every office individually air-conditioned
- Flexible: from €300 net per desk, cancellable monthly
FAQ
Aren’t AI tokens getting cheaper all the time?
The price per token is falling, but usage grows faster: AI agents, longer contexts and automation multiply the tokens per task. That is why the total bill still grows at many companies.
When do engineering teams reach €5,000 a month in AI costs?
Sooner than expected: anyone working daily with coding agents via the API uses roughly €300 to €1,000 per person per month – a team of 10 to 20 engineers quickly exceeds €5,000. Automated workflows cost even more, such as overnight analysis of test-bench and measurement data or requirements documents. Rough guide values, as of 2026.
Does your own AI server replace coding agents?
Usually not entirely: for code, many teams use the strongest cloud models. Your own GPU server mainly takes over bulk work (log and measurement data, documents, translations, test cases), round-the-clock jobs and everything that must not leave the company under customer contracts or TISAX. For such data, your own server room is often the precondition for using AI at all.
Does your own AI server pay off compared to the cloud?
Above a certain usage, yes: if your monthly API costs stay above depreciation, power, operation and space, your own server is cheaper – and the data stays in-house. Our AI cost calculator shows the threshold with your figures.
Is the campus suitable for an edge data centre or colocation?
Yes. The half-floors with 25 kW each, their own meter and server room can be used individually or combined as an edge site or for colocation. For larger loads there is the new underground space of around 1,000 m² with a 250 kW connection and 3 m clear height.
How much power is available per unit?
Each half-floor (200 m²) has 25 kW for power and cooling on its own meter. You pay for what you actually use.
What can I run on 25 kW?
For example a GPU server with 8 graphics cards (around 10 kW) or several GPU workstations – plus the cooling they need. For more capacity, units can be combined; just ask us.
How is the campus connected to the grid?
The campus has almost 2 MW of connected load, spread across 9 independent connection points, each below 300 kW. 3 further connection points are held in reserve.
Is there fibre?
Yes, fibre from 4 independent carriers plus Cat7 cabling in every office.
How long do I have to rent?
At least one month, then cancellable monthly. The offices are fully furnished and ready to move in.