BOREALIS

Borealis products

Three ways to buy Borealis, all on the same closed-loop liquid cooling, engineered around NVIDIA-class compute. Pick the one that matches your site: no building at all, one room, or a hall you already run.

Borealis HALO

A complete 2 MW AI data center that arrives on wheels. Compute, cooling, power, and controls ship factory-integrated across three to four road-legal truckloads, roll onto a prepared pad, and reach first power in 10 to 15 working days. The defining architecture is the Condenser Envelope: the enclosure itself is the heat exchanger. The long walls are full-length coil banks behind louvered intakes, the roof carries a low-speed fan array, and air moves in through the walls and out through the roof with no recirculation. Heat rejection is 100 percent dry in every mode: zero process water, no evaporative assist, no Legionella program. Design PUE is 1.10 annualized, and the unit holds 65 dBA at 10 meters, with a 55 dBA attenuated option. Day one it carries 14 NVL72-class racks, 1,008 GPUs, liquid to the chip with about 85 percent of the heat captured at the cold plates. It is fully reversible: no permanent structure, and the site is restored in days. Capacity scales by replication, 2 MW blocks growing to clusters of 20 MW and beyond on a shared pad and medium-voltage service.

Critical IT load
2.0 MW
Transport
3 to 4 road-legal truckloads
Pad to power
10 to 15 working days
Water use
Zero, 100 percent dry rejection in every mode
Design PUE
1.10 annualized
Day-one payload
14 NVL72-class racks, 1,008 GPUs
Acoustics
65 dBA at 10 m, 55 dBA attenuated option
Silicon roadmap
800 VDC provisioned, two-phase ready
Scaling
2 MW blocks to 20 MW+ clusters
Best for
Stranded power, water-constrained sites, sovereign and edge

Vega AIO

A self-contained AI data center in a single in-line unit. Three rack bays built for NVIDIA-class compute, rear-door heat exchangers, a CDU, and an onboard chiller, carrying 500 kW. About five times the cooling density of a CRAC in the same footprint. Only the condenser line leaves the unit. Roll it in, power it on, running in as little as a week on site. Best for edge sites, labs, and a single high-density room.

Cooling capacity
500 kW (one unit)
Compute bays
3 rack bays, hosts up to 216 GPUs
Density vs a CRAC
About 5x in the same footprint
External plant
None, condenser line only
Time to running
As little as a week on site
Best for
Edge, labs, single high-density rooms

Borealis Retrofit

Borealis drops its complete liquid cooling plant inside a building you already own and power, delivered in wings. Each wing pairs an 829 kW CDU bank, an in-hall chiller, and a rooftop condenser with bays for six NVL72-class racks, rear-door heat exchangers on every one, about eight times the cooling density of a CRAC. The ladder runs 12, 24, 48, 96 racks as 2, 4, 8, 16 wings, up to 12.7 MW and 6,912 GPUs in a single hall. The first wing is live in 14 days and a full hall in 8 to 10 weeks, versus roughly 270 days for a conventional multi-vendor build. You bring the building, the power, and the compute. One vendor owns everything that cools it.

Cooling per wing
829 kW (six NVL72-class racks)
Density vs a CRAC
About 8x
Hall ladder
12 / 24 / 48 / 96 racks = 2 / 4 / 8 / 16 wings
Full hall
Up to 96 racks, 12.7 MW, 6,912 GPUs
First wing live
14 days
Full hall
8 to 10 weeks
Best for
Halls you already own and power

The compute it is built around

Borealis systems are engineered and validated around NVIDIA reference compute: GB300 NVL72 racks (72 GPUs plus 36 Grace CPUs, 132 kW per rack), DGX B300, and DGX racks. The HALO ships 800 VDC provisioned and two-phase ready, so the envelope outlives at least two GPU generations. Borealis sells the system that cools and carries the compute; the compute is yours and slides straight into the bays.

Who it is for

See it, then size it.

Borealis is turnkey liquid-cooled AI data centers from AGT-USA: best-in-class cooling engineered around NVIDIA-class compute, one vendor.

Explore the interactive site → Get a quote →