Intelligence, on contract.

Lume secures, deploys and operates dedicated AI capacity across Europe, reserved before deployment and brought online fast, on terms built around your workload.

Request capacity
Discover the platform

Key Metrics

20 MW

EU AI Act full enforcement

H100 / B200

GPU clusters

Italian law

At every layer

LIAN Group

Track record

Key Metrics

40 MW

Contracted capacity across 3 sites

GB300 NVL72 / B300

Cluster generations

Q1 2027

First capacity live

0

Broker layers
(direct capacity from operator)

Compute stopped being a software problem.

Power, sites, cooling and networking now decide who gets to build. Chips are not the bottleneck. Infrastructure is. Lume owns what it builds, secures its own power access across Europe, and contracts capacity directly to customers, closing a gap no single operator covers today.

/ 001

Power first

Every site begins with grid allocation, not GPU availability. Power is secured before a site enters development

/ 002

Speed

Sites and power are contracted before deployment begins, so capacity ships in months, not years. Coordinated end to end, not sold in parts.

/ 003

Certainty, by contract

Deliverables and timing are defined contractually, backed by decades of on-the-ground energy and data center operating experience. You get a transaction, not a spec sheet.

One platform. Every layer connected.

Power, site, compute, network, operations, and contract are designed as one coherent system, not assembled piece by piece, ensuring consistency and accountability across every layer.

01
·
POWER

The advantage

Power sourcing, grid access, energisation strategy and physical resilience: the constraint everything else is designed around.
02
·
DATA CENTRE

The vessel

Tier III, purpose-built facilities designed for high-density compute, advanced cooling and staged expansion.
03
·
COMPUTE

The engine

Dedicated GB300 NVL72 and B300 clusters, configured to the workload, not a generic shared pool.
04
·
NETWORK

The connective issue

High-performance, low-latency fabric, purpose-built for distributed training and inference at scale.
05
·
OPERATIONS

The assurance

Monitoring, maintenance, security and workload support: the layer that turns hardware into a service.
06
·
IPPA

The commitment

The agreement that turns your need into deployed capacity, secured and reserved before deployment.
THE CONTRACT

Predictable capacity, by design.

A long-term Intelligence Purchase Agreement reserves defined AI capacity for you, with performance, governance and change-control terms agreed upfront, and structured flexibility to expand, refresh or renew as your needs evolve. The same logic a Power Purchase Agreement (PPA) applies to energy: lock in supply before you need it, not after.

TERM · LONG-TERM BY DESIGN, FLEXIBLE AS YOU GROW
EXPAND
REFRESH
RENEW

Capacity

A defined allocation, reserved rather than a best-effort share of a shared pool.

Term

A long-term, multi-year commitment: a planning horizon long enough to justify the infrastructure behind it.

Performance

Agreed availability and service parameters, not an assumption of uptime.

Flexibility

Structured rights to expand, reconfigure, refresh or renew, built into the term, not renegotiated from scratch.

Three sites contracted.

Total: 40 MW

Every site's capacity is underwritten by a signed Intelligence Purchase Agreement (IPPA): a long-term contract, priced and secured before deployment, the same logic energy buyers have used for decades to lock in power ahead of need. That capacity is yours: reserved to your timeline, with construction and delivery risk carried by us, not you.
SITE 001 / STATUS: CONTRACTED

Available Q1 2027

10 MW   GB300 NVL72

SITE 002 / STATUS: CONTRACTED

Available Q1 2027

5 MW    B300

SITE 003 / STATUS: CONTRACTED

Available Q3 2027

25 MW   GB300 NVL72

WHO IS IT FOR

Built for teams where compute is strategic.

AI-native companies

Training and inference capacity for teams scaling production workloads, without operating infrastructure themselves.

Strategic enterprises

Dedicated infrastructure for large-scale proprietary AI programmes.

Sovereign & regulated institutions

Controlled European infrastructure where location and oversight are contractual requirements.

The intelligence economy is becoming physical.

Demand

Compounding: moving from experiments into production-scale infrastructure.

Power

The real constraint now, not chip supply.

Lead time

Power, sites, hardware and commissioning can’t be procured instantly.

Decision

Capacity is now board-level, not a procurement line.

Secure tomorrow's capacity,
today.

Proven Experience

Grid connections, construction, live-site operations: capabilities proven across a global portfolio of data centre and power infrastructure.
Discuss Availability
lume-compute.com/
LIAN Group
Luxembourg