CONTAINER DATACENTER

Your AI datacenter, delivered in 30 days.

Factory-built. Truck-delivered. Plug-and-run. From empty land to IT-ready in 2 weeks on site.

01 / DELIVERY
30 days
500 kW
02 / DELIVERY
90 days
2 MW
03 / DELIVERY
180 days
10 MW
WHY SPEED

Built in a factory, not on a job site.

A traditional datacenter takes 18 to 24 months to build. Our containerized solution lands at your site in a third of that time, at half the CapEx, with an order of magnitude lower failure rate.

Metric Traditional TGs4
Build time 18–24 months 3–6 months
On-site labor 100% of work on site ~30% of work on site
CapEx ($/W, excl. IT) $9–12 $3–5
Install quality Site conditions vary Factory-controlled, 10× lower failure rate
Horizontal scaling Redesign each time Add modules linearly
4–8×
Faster to first power-on
−50%
CapEx reduction (excl. IT)
THE REAL ECONOMICS

The cost of waiting is bigger than the cost of building.

$7.9M
gross margin lost per 12-month delay on a 4 MW AI cluster

Real economics aren't about CapEx. They're about cash flow timing.

At $0.15/kWh power and a 2.5× IT-rental markup, every month of construction delay costs a 4 MW operator roughly $660K in margin.

The 30% you save on CapEx by building traditionally is dwarfed by what you lose to time. Speed isn't a feature — it's the entire business case.

CapEx saving
(traditional build)
~$2M
Margin lost
(12-month delay)
~$7.9M

Speed isn't a feature. It's the entire business case.

THE QUALITY ARGUMENT

More than 50% of early-life infrastructure failures trace directly to field workmanship.

70–90%
HVAC defects
of HVAC systems contain at least one installation or maintenance defect — driving energy use up to 30% above design.
U.S. DOE Review
~57%
Power faults
of industrial electrical faults are loose connections, moisture ingress, and insulation defects — all site workmanship.
IEEE Gold Book
70–80%
MV cable faults
of medium-voltage cable failures are concentrated in field-fabricated terminations and splices, not in the cable itself.
Industry Field Data

Factory prefabrication moves cable terminations, busway joints, and HVAC commissioning from a −20°C outdoor job site into a controlled environment.

The result: quality consistency lifts by an order of magnitude. Early-life failure rates drop. PUE targets get hit on day one.

Read the full technical memo →
PRODUCT ARCHITECTURE

Modular atomic units. Combine to fit.

One design language, multiple capacities. Pick a power tier, pick a compute tier, ship. Every SKU is a multiple or fraction of the 2 MW atomic unit — defined by the physical limit of a single 40HQ container.

POWER MODULES

Transformer + UPS + battery + ATS in one container. Connects to the grid on one side, to compute on the other.

DP-500
500 kW
20HQ
Compact power core 630 kVA dry-type transformer + 500 kW modular UPS + 250 kWh LFP battery
DP-1000
1 MW
40HQ
Mid-tier power core 1250 kVA transformer + 1 MW UPS + 500 kWh LFP battery
DP-2000
2 MW
40HQ
Atomic unit 2500 kVA transformer + 2 MW UPS + 1 MWh LFP — pairs 1:1 with CC-B300-2MW
GENERATOR MODULES

Diesel genset + day tank + intake/exhaust silencing. Physically separated from the power core for fire isolation, vibration containment, and easier insurance approval.

GS-500
500 kW
20HQ
MTU 8V1600 compact 500 L day tank · 8 hr autonomy on built-in fuel
GS-1000
1 MW
40HQ
Cummins C1100D5 800 L day tank · external fuel-tank compatible
GS-2000
2 MW
40HQ
MTU 12V4000 1500 L day tank · 80 hr autonomy with FuelTank-20K
COMPUTE MODULES

Four density tiers, four GPU procurement models. Pick by how you buy your servers, not by performance.

CC-AIR-240
240 kW
40HQ · AIR-COOLED
Versatile baseline 6 racks @ 40 kW · Edge / storage / management nodes / H100, A100
CC-LIQUID-900
900 kW
40HQ · LIQUID-COOLED
Self-built GPU servers 10 racks @ 90 kW · H200 / B200 self-assembled
CC-NVL72-900
924 kW
40HQ · NVIDIA NVL72
NVIDIA reference design 7 NVL72 racks · 504 GB200/B300 GPUs · in-rack CDU
CC-B300-2MW
1.84 MW
40HQ · OCP ORV3
Hyperscale density 8 racks @ 230 kW · 1,152 B300 GPUs · 21" OCP
▸ EDGE SPECIAL · ALL-IN-ONE
AIO-500
500 kW
SINGLE 40HQ · ~15 t

One container. Power core, generator, and battery, all inside.

The only three-in-one SKU in the lineup. Built for edge sites where a second container isn't an option — remote islands, mountain passes, temporary deployments, or customers who simply want the fewest possible boxes on the ground. Above 500 kW, the 40HQ payload limit (26 t) makes three-in-one physically impossible — DP + GS must split.

+ BESS storage · MV switchgear · Network container · External fuel-tank · Liquid-cooling skid · PV + MPPT

Like IKEA furniture: standard parts, flexible combinations, globally manufactured, locally serviced.

ON-SITE DEPLOYMENT

From empty land to IT-ready in 6 days.

Standardized industrial quick-connectors compress on-site assembly from months to days. Crane in. Plug in. Power on.

DAY 1
Crane positioning ~8 HOURS
DAY 2
Power connections ~2 HOURS · POWERLOCK + MV
DAY 3
Cooling water connections ~3 HOURS · STÄUBLI QRC
DAY 4
Network connections ~3 HOURS · LC FIBER + BMS
DAY 5
System commissioning ~8 HOURS
DAY 6
IT load on FIRST WORKLOADS
POWERLOCK 660A
5 min
Mechanical low-voltage connection. Industry standard for power and broadcast.
STÄUBLI QRC
10 min
Per pair. The de-facto standard for datacenter liquid cooling.
LC DUPLEX FIBER
5 min
Per pair. Standard data uplink and management network.
6 days vs. 6–12 months.
~30× FASTER ON-SITE ASSEMBLY
BUILT FOR

Four customers. One product line.

From a single edge node to a Hyperscale cluster, the same modular SKUs cover the full range. Pick the configuration that matches your scale.

▸ 01 / AI STARTUPS
AI startups racing the GPU clock
500 kW – 2 MW
"Our GPUs already shipped. The datacenter is still being built. We're burning $50K a day in idle hardware."
Recommended: Compact / Standard
▸ 02 / ENTERPRISE
Enterprise private AI nodes
200 kW – 1 MW
"Tight budget, edge deployment, no datacenter on site. We need turnkey, not a construction project."
Recommended: Edge Mini / Compact
▸ 03 / TRAINING CLUSTERS
Large AI training clusters
4 MW – 10 MW
"High density, scalable, modular. We add capacity by adding containers, not by rebuilding."
Recommended: Mega / Hyperscale
▸ 04 / MINING → AI
Crypto mines transitioning to AI
2 MW – 20 MW
"We have the power infrastructure already. We just need to upgrade the compute to liquid-cooled AI."
Recommended: Standard / Mega + BESS

Ready to skip the 18-month build?

Engineering response within 24 hours · Quotes in 5 days