Curated AI infrastructure · Quote on request
A short, opinionated catalogue of real, current components for running inference on hardware you own: power, thermal and compute, plus a matched build kit. Every item links to the manufacturer's spec sheet.
Four tiers
Most inference problems are power or heat problems first. The catalogue is organised the same way.
ATX 3.1 power supplies and rack UPS units sized for GPU transients.
Coolers and fans that keep high-TDP silicon out of thermal throttling.
Workstation and data-centre NVIDIA GPUs with ECC memory for inference.
The Green Racks Edge-Build Kit: a matched inference build with a colocation path.
Sovereign Bundle
A matched inference build specified by Green Racks: workstation GPU, ATX 3.1 power, air cooling and a rack UPS, with an optional path into colocation in Osborne Park, WA.
From the catalogue
One card from each tier. Swap any of them in your quote.
Single-slot Blackwell GPU with 24 GB of ECC memory. The most power-efficient way to serve 7–14B models.
1600 W, 80 PLUS Titanium, fully modular ATX power supply with native 12V-2x6 output. Suited to dual-GPU workstation inference builds.
Dual-tower air CPU cooler with two 140 mm fans for high-power desktop and workstation CPUs.
How quoting works
Add items to your quote list. It's saved in your browser; nothing is sent yet.
Tell us where it will run (office or colocation) and your timeframe.
Pricing, lead times and compatibility notes. A quote request is not an order.
Office or rack
Every build can go in your office or into a Green Racks rack in Osborne Park, WA, with 2N UPS (150kVA), dual A/B feeds per rack, N+1 cooling and NOVEC fire suppression.
Why this catalogue
Colocation offer
First month of colocation free on a 12-month term*
* Terms apply. Applies only to new colocation services on a 12-month term, subject to a written quote and Green Racks' terms. May change or be withdrawn. See colocation pricing.
FAQ
Distributor terms are still being finalised, so we quote each request individually with current pricing and lead times rather than publish numbers that may be out of date.
It means our team chose the item based on the manufacturer's published specifications for running AI inference on hardware you own. It is not a claim that Green Racks has tested the item.
No. A quote request is not an order and no payment is taken. We reply with pricing, lead times and any compatibility notes.
Yes. Builds can go in your office or in a Green Racks rack via Aussie Racks, with 2N UPS (150kVA), dual A/B feeds per rack, N+1 cooling and NOVEC fire suppression. First month of colocation free on a 12-month term*.
Not sure what you need?
The PSU sizing guide and the GPU comparison matrix on AI Hardware Lab explain the trade-offs.