01 · GPUs
NVIDIA
- GB300 NVL72
- GB200 NVL72
- HGX B300 / B200
- HGX H200 / H100
- RTX PRO 6000 Blackwell
CANUS Your AI factory.Where the power already is.
For AI clouds · AI Factory as a Service
You bring the GPUs. We find the power, own the pod and run it.
450 kW → 1.2 MW
Per Pod, Two Hulls
22.5MW+
Per Campus, Run as One Fleet
120DAYS
Target from Order to Powered Racks
-12–21%
Under Sydney Metro, All-in
0LITRES
Evaporative Cooling Water
Pre-plumbed
Pre-plumbed forevery frontier chip.
Every rack in the lineup plugs into the same pod. Change the silicon; keep the pod. You bring it; we never touch it.
02 · GPUs
AMD
- Instinct MI355X
- Instinct MI350X
- Instinct MI325X
- Instinct MI300X
- ROCm software stack
03 · Servers
Dell
- PowerEdge XE9712
- PowerEdge XE9680L
- PowerEdge XE9680
- PowerEdge XE9785L
- PowerEdge XE7745
04 · Servers
HPE- GB200 NVL72 by HPE
- ProLiant Compute XD685
- Cray XD670
- ProLiant DL380a Gen12
- Cray EX
05 · Servers
Supermicro
- GB200 NVL72 SuperCluster
- 4U liquid-cooled HGX B200
- HGX H200 8-GPU
- Liquid-cooled MI355X
- DLC-2 cooling
06 · Servers
Lenovo
- ThinkSystem SC777 V4
- ThinkSystem SR780a V3
- ThinkSystem SR680a V3
- ThinkSystem SR685a V3
- Neptune liquid cooling
07 · Servers
Gigabyte- GIGAPOD clusters
- G4L3 liquid-cooled HGX
- G893 HGX B200
- G593 HGX H200
- XL44 RTX PRO 6000
08 · Networking
Broadcom
- Tomahawk 6
- Tomahawk 5
- Jericho3-AI
- Thor 2 400G NIC
- PCIe Gen5 switches
09 · Networking
Arista- 7800R4 800G spine
- 7060X6 800G leaf
- 7700R4 Etherlink
- Arista EOS
Third-party brands and models identify compatible hardware, interfaces and specification standards only. No affiliation, endorsement or partnership is implied.
The reverse AI factory
Don’t bring power to the factory.
Bring the factory to the power.
You bring the GPUs. We find the power, own the liquid-cooled pod and run it, and lease you the capacity on one invoice. Target: powered racks 120 days from order.
01 · Power
We find the power.
Solar-and-battery farms and industrial premises that already have power and fibre. No connection queue on the critical path.
02 · Build
We own the pod.
A factory-built, liquid-cooled pod in two hulls, 450 kW to 1.2 MW, craned onto a ready pad. 120-day target deployment from order.
03 · Run
We lease the capacity.
One AI cloud per pod brings its own GPUs. Rent per IT-kW per month, power at cost. We never touch your software.
PortaDC
A whole AI data centrein one pod.
Racks, liquid cooling, power and controls in one factory-built hull, craned onto a ready pad and switched on. Nothing to build on site.
450 kW IT · Liquid-cooled · 120-day target
01 · Hardware
The pod
Two hulls, one product. PortaDC20 carries 450 kW in 20 ft; PortaDC40 carries 900 kW to 1.2 MW in 40 ft. Built and tested in the factory, rated for a 45 °C summer, and craned onto the pad in an afternoon.
02 · Energy
The power
Deployed where a connection already exists: solar-and-battery farms and industrial premises. Behind-the-meter renewable power, passed through at cost under transparent site schedules. No hall, no queue, no evaporative water.
03 · Economics
The invoice
One line for rent per IT-kW per month, one for power at cost. About US$213 to 238 all-in against Sydney metro at about US$270, 12 to 21% under depending on draw. Every watt in and every unit of heat out is metered and signed into the record.
The fleet · Two hulls · 450 kW to 22.5 MW+
Megawattsin months.
450 kW · 4.5 MW · 22.5 MW+
A metro hall is years in a queue. A PortaDC lands on a connection that already exists. Two hulls, one product: PortaDC20 at 450 kW, PortaDC40 at 900 kW to 1.2 MW, every pod arriving complete. Add a pod, not a building, on the same pod every site. The second megawatt is faster than the first.
From one pod to the fleet
01The power
Start where the power is.
Connected sites. No grid queue.
PortaDC lands at solar and battery sites and industrial premises that already have power and fibre. No transmission queue on the critical path, so your GPUs start earning on a 120-day target from order.
01The power
Start where the power is.
Connected sites. No grid queue.
PortaDC lands at solar and battery sites and industrial premises that already have power and fibre. No transmission queue on the critical path, so your GPUs start earning on a 120-day target from order.
- 01The power
- 02The pod
- 03The cooling
- 04The compute
- 05The operations
- 06The fleet
Inside the pod
Compute and cooling,built as one.
Every PortaDC is a complete 450 kW to 1.2 MW AI facility in one shell. Racks, liquid cooling, 415 V power and controls are designed, built and tested together before the pod leaves the factory.
450 kW ITLiquid-cooled to the chipClosed loopFactory-tested and recorded
You bring the GPUs. We bring the powered pod.
01 / 05 · GPU racks
PROPRIETARY TECH · ONE UNIFIED PRODUCT · 03 FILED · 02 PREPARED
Built like a tank.
Runs like a power station.
Bring the AI factory to the power. Engineer the compute to fit.
Facility, cooling, boundary control and scheduling. One unified system that turns a connected site into dependable AI capacity.
01 // FRONTIERGATE™01 // FRONTIERGATE™ · INSIDE PORTADC
The Boundary Controller
Holds the host's power boundary so your firm floor never moves. Firm to the tenant, flexible to the site.
Explore system
02 // SOLAR-SPONGE™02 // SOLAR-SPONGE™
Solar-Synced Scheduling & Thermal Buffer
Shapes the pod's flexible band around the solar day and the host's needs. Lower-cost daytime power, with your orchestration left to you.
Explore system
03 // LARUS AI™03 // LARUS AI™
Firm + Flex Qualification & Operating Record
One hardware-signed record of power in and heat out, from factory test through every hour of operation. Tenant and host read the same numbers.
Explore system
04 // IR-STEALTH™ ARMOR04 // IR-STEALTH™ ARMOR
Ceramic Heat Shield & Zero-Water Cooling Spine
Infrared-reflective ceramic armor built for a 45°C design ambient. Closed-loop liquid cooling with no evaporative water after fill.
Explore system
PortaDC™ · Configure a pod
Order the data centrelike you order the GPUs.
Pick a hull, pick a tier, tell us the workload. Every pod is the same 450 kW building block, one in PortaDC20, two in PortaDC40 at 900 kW to 1.2 MW, so the lease is three numbers and the pod is already built. Powered racks on a 120-day target.
Target rent per IT-kW-month · power passed through at cost · you bring the GPUs
Pod
Hull
24/7 firm production inference
Firm floor plus cushion. Flexible band: zero, never trimmed. About US$213 to 238 all-in, 12 to 21% under Sydney metro.
01 // PortaDC Pulse™
PortaDC20 · 20 ft
Spec // PortaDC20 · Pulse™
- IT power
- 450 kW · firm floor + cushion · band 0
- Racks
- 3 – 5 liquid at 80 – 130 kW
- GPUs (you bring)
- 4 HGX H200 / B200 or GB200 racks + optical and NVMe cache rack
- Cooling
- Direct-to-chip liquid, rear-door hybrid on PortaDC40 · supply at or below 35 °C
- PUE
- 1.10 – 1.21 Design PUE
- Target rent · all-in, US$
- US$165–172 (A$235–245) · ~US$213 to 238 all-in · 12 to 21% under
- Hookup
- One 415 V bay · existing switchboard headroom
- On the pad
- Truck delivery · two-module lift and lock on a ready pad
- Powered racks
- 120-day target deployment from order
Why AI clouds choose PortaDC™
Same GPUs.Different outcome.
Flip the switch. Six readings move from a Sydney or Melbourne hall to a PortaDC pod behind the meter. You own the GPUs either way.
01 · Time to powered racks
Years
Metro grid queue
Prime GPU earning life, not queue time
02 · Liquid cooling
5–15 kW / rack
Legacy air halls
Dense racks without a retrofit
03 · Host flexibility
Blunt load shedding
Or peak penalties
Firm to the tenant, flexible to the host
04 · Power
Metro tariff
Retail grid mix
Power at cost, one line on the invoice
05 · Footprint & control
Shared cages
Multi-tenant metro halls
Your racks, your software, one invoice
06 · All-in monthly cost
~US$270 / IT-kW / mo
Sydney metro, all-in
12 to 21% under Sydney metro
Indicative figures, confirmed in each lease. Power figures are planning blends, not a host tariff. Sydney metro all-in reference about US$270 per IT-kW-month.
Next site // Central-West NSW
Your GPUs have a home.Reserve your first pod.
A sub-5 MW solar and battery farm in Central-West NSW with its own switchyard and an 11 kV connection. Room for one to four pods. Capacity goes to the buyer who commits first.
- Site
- Sub-5 MW solar and battery farm, Central-West NSW
- Connection
- Existing 11 kV / 415 V, ties into the station auxiliary board. No new grid works on the critical path.
- Capacity
- 0.45 to 1.8 MW IT load. One to four pods at 450 kW IT, about 545 kW input each. Surge and Pulse tiers.
- Power
- Daytime solar behind the meter, battery-firmed, passed through at cost under the site schedule. No evaporative water after fill.
- Status
- Connection stage. OEM thermal spec issued, engineering scoping under way.
- First tranche
- 1.8 MW, NSW, H1 2027 target.
- Timeline
- 120-day target deployment from order, on a 16-week schedule from site release.
Subject to connection offer and site agreement.
Ten host sites in screening across NSW.
Reserve capacity or host a pod
Deploy a pod.
Three ways in. Pick your track; a named person replies within one business day. Enquiries do not quote prices or reserve inventory.