CrowdCompute/ EU
Independent minds. Shared infrastructure.EUROPE · CLUSTER 001

300 people.
8 H200s.
One European
AI cluster.

Pool our demand to run frontier open-weight AI on dedicated European compute.

Instead of each of us buying expensive GPUs or sending our code and data to proprietary AI providers, we collectively fund dedicated AI infrastructure in Europe.

Reserve my seat How it works

Target price: ~€59/monthNo payment today. No binding commitment.

Founding Cluster EU-01From collective interest to dedicated compute.
2 / 300 interested1% OF TARGET

Only “Yes, at ~€59/month” responses count. Interest is not a purchase.

OPEN-WEIGHT MODELSEUROPEAN COMPUTEPRIVACY BY DESIGNCOMMUNITY CHOICE
01 / THE IDEA

Compute is expensive.
It doesn’t have to be lonely.

One person renting an 8×H200 server makes little economic sense. Hundreds of people sharing one does.

Don’t buy another subscription to a black box. Help build an alternative.

CrowdCompute brings people with the same need together: capable, open-weight AI on infrastructure we can understand.

The proposal is simple. Pool a monthly budget, fund a dedicated cluster in Europe, and share its capacity. Members fund access to infrastructure; they do not own the physical GPUs.

See the numbers
DEVELOPERSBUILDERSTEAMS
CrowdCompute
Dedicated 8×H200 clusterHOSTED IN EUROPE
Open-weight frontier model

AI is becoming more capable — and more opaque.

We increasingly rely on systems whose model version, system instructions, alignment choices and retention policies we cannot see.

CrowdCompute is an experiment in the opposite direction.

Less black box. More user control.

02 / FIRST CLUSTER

Meet the starting point.

Serious hardware. A shared purpose.

PROPOSED CONFIGURATION
EU–01

Built around the compute.
Not locked to a model.

FOUNDING CLUSTER / EUROPE
ACCELERATORS
8 × NVIDIA H200
MEMORY
~1.1 TB HBM
INTERCONNECT
NVLink
LOCATION
European datacenter
OPERATION
24/7 operation planned
MODEL
Community selected

The compute stays.
The model can change.

Potential candidates include GLM, Kimi, Qwen and other frontier open-weight models. Fund the infrastructure, then choose the model that serves the community best.

Candidates are subject to licensing, memory requirements and deployment testing. No specific model is guaranteed.

03 / OUR PRINCIPLES

Less black box.
More common ground.

The commitments we intend to build the service around.

PRIVATE

Your prompts.
Your business.

No prompt or response retention by default in the planned inference service. Privacy belongs in the architecture.

EUROPEAN

Closer to home.
By design.

Hosted in the EU, with an EU-based infrastructure provider as the target.

TRANSPARENT

Know what
you’re funding.

Infrastructure cost, cluster utilization and available capacity will be visible to members.

04 / OPEN ECONOMICS

A simple equation.
A different kind of scale.

The initial hypothesis. Every euro has a job.

300MEMBERS
€59PER MONTH / TARGET
€17,700SHARED MONTHLY BUDGET

That budget would fund the infrastructure and the work needed to keep it running.

  • GPU infrastructure
  • Storage
  • Networking
  • Monitoring
  • Payment processing
  • Operational reserve
WHY €59?

It is a target based on the economics we believe committed 8×H200 capacity can reach in Europe. No infrastructure contract has been signed yet. The cluster will only launch once the final economics are sustainable.

THE FINE PRINT, UP FRONT.

This is the target economic model, not a final price. The final launch price depends on the infrastructure contract we negotiate.

More members should create better economics,
not just higher margins.

05 / SHARED CAPACITY

Shared doesn’t mean unlimited.
It means thoughtfully managed.

300 members.
Not 300 simultaneous workloads.

The first cluster will initially target approximately 300 members. Not everyone generates tokens at the same time. The goal is fair access to a shared pool, with limits informed by real benchmarks.

No throughput, latency or availability guarantee is being offered at this stage.

  • 01Continuous batching
  • 02Prefix caching
  • 03Fair scheduling
  • 04Usage limits against abuse
  • 05Queues for exceptionally large workloads
A NOTE ON LARGE CONTEXT

Some models may support contexts such as 1M tokens. These would be fair-use or queued workloads, not the default experience.

06 / ON THE ROADMAP

Your workflow.
A different compute layer.

Intended access options. Not available in this MVP.

PLANNED

OpenAI-compatible API

Familiar interfaces for the applications and agents you’re already building.

# Future access
client.chat.completions.create(
  model="community-selected"
)
PLANNED

Built for your tools.

Intended integrations, subject to model and client compatibility.

  • Claude Code
  • OpenCode
  • Aider
  • Continue
  • Cline
  • OpenHands
  • LangChain
  • Agents

Prefer a conversation?

A web chat is also on the roadmap.

07 / COMMUNITY CHOICE

One cluster.
Your collective call.

Which model would you prefer? Share your vote when you reserve a seat. The community helps shape what comes next.

Make your voice count

Votes include all responses, including “maybe” and “no”. They guide the decision; they don’t guarantee a deployment.

Which model would you prefer?

LIVE VOTE
GLM50%
Kimi0%
Qwen0%
Community decides50%
Other0%

2 responses · Anonymous aggregate totals · Percentages rounded

EXPRESSION OF INTEREST

Help launch
EU-01.

Great infrastructure starts
with people who want it.

We will only contract the first cluster if enough people want it. Tell us what you’d build, and help shape what comes next.

~€59/month targetNo payment todayNo binding commitment
INDEPENDENT MINDS. EUROPEAN COMPUTE.

Reserve my seat

EU-01 / 001

A minute of your time. A signal for something bigger.

Would you join at approximately €59/month?
Please don’t include sensitive information.0/500

This is an expression of interest only. You are not purchasing anything and no payment is collected. Privacy policy.

Complete the form and security check to register your interest.

Only “Yes” responses count toward the 300-member target.