Revolutionising

Breaking industry-wide bottlenecks with the world’s first turnkey AI data center platform.

We are dedicated to addressing AI’s most pressing challenge —
energy efficiency.

%

better FLOP per utility picoJoule vs Nordic (free cooling) data centers

Up to

%

Up to 30% better TCO than direct-to-chip data center platforms

How? The world’s most energy-efficient AI data center platform, delivered as a service. Our liquid immersion-cooled Sustainable AI Factory platform is already scaled, built for both retrofit legacy data centers and greenfield builds.

Proven and verified globally with design, build, and operate services that deliver AI FLOPS with best FLOP per picoJoule ratio in the world.

PLUG & PLAY →
1.02pPUE

An example configuration of a 2,048 GPU AI data center that uses 37% less energy per FLOP than free cooling, with 100% of the performance output - independently verified. See how it comes together.

Overview
Technical
NETWORK SPINE
NETWORK SPINE

The immersion data halls connect back to central air-cooled spine & leaf networking racks, with space for storage & infrastructure nodes. In a standard AI factory deployment, 2 x 128 node (256 node total) HyperCubes (for a total of 2,048 H100/H200/MI325X accelerators) share this common area.

World-first 400-800GBPS Firmus immersion-warranted OSFP & QSFP transceivers deliver full-rail cluster designs with ease.

NETWORK SPINE
NETWORK SPINE
Data Center Heat Rejection
Data Center Heat Rejection
HyperCube Data Hall
HyperCube Data Hall
CDU & Electrical Room
CDU & Electrical Room
PLUG & PLAY
Tech →
Our Technologies

We're the world leader in immersion cooled data centers.

37% more FLOPS per unit of energy vs free cooling. 30% less cost than DTC to retrofit any data hall. >50% more cost effective to build standalone than air-cooled data centers.

Our platform is verifiably the most cost effective way to build and operate hyperscale AI data centers.

Using advanced immersion cooling technology, we've mastered service delivery to guarantee high uptime for any type of CPU or GPU cluster, unlocking cost and environmental benefits of our platform for everyone to use.

We're the world leader in immersion cooled data centers.
Use →
services / how to use our platform

Flexible ways to access and benefit from our technology, today.

  • Retrofitting
  • New builds
  • Managed Colocation
  • GPU Cloud

Turn a legacy data centre into a liquid-cooled hyperscale AI factory. Our single-phase immersion platform retrofits to air-cooled data centres. Infrastructure is combined with a flexible service delivery model that allows providers to scale rapidly.

Firmus greenfield Sustainable AI Factories are future-proof, energy-dense, use half the energy, and cost half as much to build against legacy AI data centers. Our Global Engineering and Construction (GECO) organization can work with you to build ground-up AI data centers.

For retrofits or new builds, all Firmus Sustainable AI Factories come with managed colocation services that deliver frictionless access to our platform for operators and users. Use our pre-validated hardware or bring your own - we unlock access to breakthrough TCOs for any platform.

From our own data centers in Singapore & Australia, we're delivering an NVIDIA certified cloud service, offering cloud access to large clusters of H100 GPUs with fast networking and low cost. Use as public or private cloud, with enterprize-grade SLAs. Visit SMC to learn more.

Retrofitting

Turn a legacy data centre into a liquid-cooled hyperscale AI factory. Our single-phase immersion platform retrofits to air-cooled data centres. Infrastructure is combined with a flexible service delivery model that allows providers to scale rapidly.

New builds

Firmus greenfield Sustainable AI Factories are future-proof, energy-dense, use half the energy, and cost half as much to build against legacy AI data centers. Our Global Engineering and Construction (GECO) organization can work with you to build ground-up AI data centers.

Managed Colocation

For retrofits or new builds, all Firmus Sustainable AI Factories come with managed colocation services that deliver frictionless access to our platform for operators and users. Use our pre-validated hardware or bring your own - we unlock access to breakthrough TCOs for any platform.

GPU Cloud

From our own data centers in Singapore & Australia, we're delivering an NVIDIA certified cloud service, offering cloud access to large clusters of H100 GPUs with fast networking and low cost. Use as public or private cloud, with enterprize-grade SLAs. Visit SMC to learn more.

Proven →
Verified

MLPerf certified as the most energy efficient cloud infrastructure to train AI.

Ecosystem partners