Taalai's AI Hardware Co.
Spec sheet · SKU WKS-DGXSTATION-GB300-748G · verified July 2026
DGX Station (GB300)
Desk-side supercomputer In stock

DGX Station (GB300)

A single GB300 Grace Blackwell Ultra Superchip in a desk-side chassis — one coherent 748 GB memory pool shared between CPU and GPU, not multiple separate cards. Runs on standard office power.

SKU WKS-DGXSTATION-GB300-748G
PriceiChecked against NVIDIA's own product listings and major retailers (B&H, Micro Center, NVIDIA's marketplace) as of July 2026 — real published prices, not estimates. $99,000
Request a quote Back to catalog

Specifications

GPU memory
748 GB HBM3eiAn upgraded version of HBM3 with more capacity and bandwidth per GPU. Used in newer datacenter GPUs like the H200 and the Blackwell Ultra chips in the GB300.
Power draw
1.6 kW
Tensor / AI compute
~20 petaflops FP4 (sparse)
Rack units
N/A — desk-side tower, not rack-mounted
Superchip1x GB300 Grace Blackwell Ultra
Compute~20 petaflops FP4 (sparse)
Memory architectureUnified CPU+GPU via NVLink-C2C — not multiple separate GPU pools
Power requirementStandard 120V/20A office circuit — no special power needed

Which models this can run

Real memory math against the 3 models in the model advisor — parameters × 1 GB, plus 20% working room.

In simple terms Imagine an AI model's brain is made of billions of tiny sticky notes, and each one needs about a gigabyte of shelf space. Every single sticky note has to fit on the computer's memory shelf at the same time, with a little extra room to spare — if the shelf's too small, they just don't fit, and the computer can't run that model at all, no matter how good the rest of it is.
✓
Qwen3.5-397B 476.4 GB required · this unit has 748 GB
Runs on a single unit.
✗
DeepSeek-V3.2 805.2 GB required · this unit has 748 GB
Does not have enough memory, and this product does not cluster.
✗
Kimi K2.6 1,200.0 GB required · this unit has 748 GB
Does not have enough memory, and this product does not cluster.

Power and cooling

Electrical requirement: Draws 1,600 W — fits within a single standard 120V/20A circuit (1,920 W safe continuous per NEC's 80% rule). No special electrical work needed.
About 1.3x the continuous draw of an average home (1200 W). Run flat-out for a month, that's roughly 1,152 kWh — about $150/month at a $0.13/kWh commercial rate (US average, EIA — your actual rate depends on region and utility contract).
In simple terms Watts are like how hungry a machine is for electricity — a bigger number means it "eats" more electricity, kind of like a bigger animal eating more food. This one draws about as much electricity as 1.3 whole houses running non-stop, which is why bigger machines like this need special rooms built to feed and cool them.
Power cost, relative to the rest of the catalog $150/mo

Cooling: Air-cooled, designed for office environments

What this is good for

  • Running models up to roughly 620 GB after working-room overhead on a single machine
  • Teams that want server-class memory without datacenter power or cooling
  • Development and inference workloads that do not need multi-node scale

What this is NOT good for

  • Workloads that need more than one Superchip's worth of memory — this does not cluster with other units
  • High-concurrency production serving at scale — see the server tiers below for that
  • Frontier-scale training runs — the rack tier exists for that

Included components

Complete desk-side system: Grace CPU + Blackwell Ultra GPU Superchip, unified memory, chassis, and PSU. No monitor, keyboard, or networking switch included.

Lead time and warranty

Delivery window estimate4–12 week lead time depending on OEM partner, as of July 2026.
WarrantyManufacturer warranty applies (terms vary by OEM partner).
Request a quote for this system
Get matched to a build