NVIDIA RTX PRO 6000 Blackwell Server Edition
Eight Blackwell GPUs with 96 GB each, PCIe, air cooled. Inference, graphics and smaller models.
- 8 GPUs per node
- 768 GB GPU memory
- Air
- From Q1 2027 Santiago, on contract · US on request
Configurations
Pick the footprint that fits. Every one is sold on the same contract.
- 8 GPUsOne node
- The smallest unit. Fits any hall, in Santiago or in the United States.
- 16 to 248 GPUsTwo to thirty-one nodes
- One Ethernet fabric.
- 256 GPUs and moreThirty-two nodes and up
- Rows of standard air-cooled racks.
Specifications
- GPUs
- 8 RTX PRO 6000 per node
- GPU memory
- 768 GB, 96 GB GDDR7 per GPU
- NVLink
- PCIe Gen 5, no NVLink
- CPUs
- 2 x86 CPUs
- System memory
- 128 GB per GPU or more
- Networking
- BlueField-3 SuperNICs, 200 to 400 Gb/s per GPU
- Power
- 5.4 to 6.7 kW per node; 400 to 600 W per GPU
- Cooling
- Air
- Fabric at Fulmis
- InfiniBand or Spectrum-X Ethernet, sized to the workload
- Software
- Slurm and managed Kubernetes
- Location
- A data centre in the Santiago Metropolitan Region, or a colocation facility in the United States
- Available from
- Q1 2027 in Santiago, on contract; in the United States the start is set in the term sheet
Performance
Per node, as NVIDIA publishes it.
- FP4 Tensor, per GPU
- 4 PFLOPS
- FP8 Tensor, per GPU
- 2 PFLOPS
- FP16 / BF16 Tensor, per GPU
- 1 PFLOPS
- TF32 Tensor, per GPU
- 234 TFLOPS
- FP32, per GPU
- 120 TFLOPS
- GPU memory
- 768 GB GDDR7, 1.6 TB/s per GPU
- Max GPU power
- 400 to 600 W, configurable
- Node power
- 5.4 to 6.7 kW
NVIDIA's published figures. Tensor Core numbers carry sparsity unless the row says dense; H100, H200 and RTX PRO figures are per GPU because NVIDIA publishes no node total. Power figures are the ones we reserve and protect to, with NVIDIA's own where it publishes one.
How we deploy it
RTX PRO 6000 Blackwell Server Edition is the eight-GPU node for inference, rendering and models that fit a single GPU's 96 GB. At under 7 kW a node it fits any hall, in Santiago or in the United States.
A node is the smallest unit.
Fulmis places the purchase order, or takes delivery of the fleet you already own. Either way we place it in Santiago or in the United States, commission it against the vendor baseline and operate it for the term. You reserve the capacity, or your own hardware; the workload and the data stay yours.
Terms
- Term
- Three to five years
- Price
- Fixed for the term, in the term sheet
- Commitment
- Take-or-pay
- Prepayment
- None to 25% at signature
- Smallest block
- One node, 8 GPUs
- Place
- Santiago, Chile, or the United States
- Start
- Q1 2027 in Santiago; US start in the term sheet
Other platforms
| Platform | GPUs | GPU memory | NVLink | Power | Cooling | Smallest block | Available from |
|---|---|---|---|---|---|---|---|
| GB300 NVL72Rack scale | 72per rack | 20 TB288 GB per GPU | 130 TB/s1.8 TB/s per GPU | 132 to 140 kW155 kW peak | Liquid | 1 rack, 72 GPUs | Q1 2027, SantiagoOn contract. US on request |
| GB200 NVL72Rack scale | 72per rack | 13.4 TB186 GB per GPU | 130 TB/s1.8 TB/s per GPU | 120 to 132 kWper rack | Liquid | 1 rack, 72 GPUs | Q1 2027, SantiagoOn contract. US on request |
| HGX B3008-GPU node | 8per node | 2.3 TB288 GB per GPU | 14.4 TB/s1.8 TB/s per GPU | About 14 kWper node | Air or liquid | 1 node, 8 GPUs | December 2026, SantiagoJanuary 2027, US. On contract |
| HGX B2008-GPU node | 8per node | 1.4 TB180 GB per GPU | 14.4 TB/s1.8 TB/s per GPU | About 14.3 kWper node | Air or liquid | 1 node, 8 GPUs | December 2026, SantiagoJanuary 2027, US. On contract |
| HGX H2008-GPU node | 8per node | 1.1 TB141 GB per GPU | 7.2 TB/s900 GB/s per GPU | About 10.2 kWper node | Air | 1 node, 8 GPUs | Q1 2027, SantiagoOn contract. US on request |
| HGX H1008-GPU node | 8per node | 640 GB80 GB per GPU | 7.2 TB/s900 GB/s per GPU | About 10.2 kWper node | Air | 1 node, 8 GPUs | Q1 2027, SantiagoOn contract. US on request |
| Vera RubinRack scale | 72per rack | HBM4not yet published | Not published | Not published | Liquid | 1 rack | On request |