Loading...
Dedicated next-generation GPU capacity engineered around qualified power, direct-to-chip liquid cooling, high-speed fabric and a defined target RFS.
Capacity can be configured around current and next-generation NVIDIA rack-scale platforms, with 400G / 800G-ready Ethernet and RoCE architectures sized around the workload.
Compare geographic location, IT load, indicative GPU scale, network, cooling, target RFS and forward-reserved pricing — then define the final deployment around your workload.
Nistar pairs forward power and cooling capacity with deployable GPU, fabric and operating architectures. Compare the available deployment windows, then lock the final platform and BOM around your workload.
The entries below represent planned infrastructure capacity, not warehoused GPU inventory. Final hardware allocation, cluster topology and commercial terms are established with the tenant.
| Platform | Indicative GPUs | IT Load | Location | Network | Cooling | Status | Target RFS | Indicative Reserved Rate / GPU-hr | Term | |
|---|---|---|---|---|---|---|---|---|---|---|
GB300 Blackwell Ultra / Vera Rubin | ~2,520indicative | 5.4 MW | Texas | 400G / 800G-ready Ethernet / RoCE | Direct-to-chip liquid | Forward Reserving | Q3 2027 | GB300from $5.25 Rubinfrom $6.50 | 36–60 mo | Request |
GB300 Blackwell Ultra / Vera Rubin | ~4,968indicative | 10.5 MW | Texas | 400G / 800G-ready Ethernet / RoCE | Direct-to-chip liquid | Forward Reserving | Q4 2027 | GB300from $5.00 Rubinfrom $6.25 | 36–60 mo | Request |
GB300 Blackwell Ultra / Vera Rubin | ~20,664indicative | 43.5 MW | Texas | 400G / 800G-ready Ethernet / RoCE | Direct-to-chip liquid | Forward Reserving | Q2 2028 | GB300from $4.50 Rubinfrom $5.75 | 36–60 mo | Request |
GB300 Blackwell Ultra / Vera Rubin | ~18,576indicative | 39.1 MW | Iowa | 400G / 800G-ready Ethernet / RoCE | Direct-to-chip liquid | Forward Reserving | Q2 2028 | GB300from $4.50 Rubinfrom $5.75 | 36–60 mo | Request |
GB300 Blackwell Ultra / Vera Rubin | ~28,944indicative | 60.9 MW | Pennsylvania | 400G / 800G-ready Ethernet / RoCE | Direct-to-chip liquid | Forward Reserving | Q2 2028 | GB300from $4.25 Rubinfrom $5.50 | 36–60 mo | Request |
A reservation aligns compute procurement with the infrastructure required to commission and operate it. Final architecture remains workload-specific.
Final rack-scale platform selected around workload, availability and target RFS.
ConfiguredLow-latency east-west fabric with deployment-specific leaf-spine architecture.
800G-readyHigh-density facility design engineered around rack-scale accelerated compute.
LiquidCapacity and throughput sized around checkpoints, datasets and inference traffic.
Workload-ledOperating boundary can be configured around tenant orchestration and support requirements.
FlexibleSend the platform, GPU scale, preferred location and RFS date. Nistar can match the requirement against current forward-reserving capacity and define the technical and commercial path to acceptance.
No. These listings represent planned dedicated AI infrastructure capacity. Final GPU procurement and allocation are matched to the tenant's selected platform, scale, target RFS and definitive agreement.
The final count depends on platform, rack architecture, networking, storage, operating overhead and engineering headroom. The displayed quantities provide a planning-scale view of the corresponding IT capacity.
Nistar can structure a forward reservation around a Vera Rubin-class deployment, subject to hardware allocation, final OEM availability, engineering and definitive commercial terms.
No. Rates are indicative starting points for long-term dedicated reserved capacity and vary by platform, scale, term, procurement timing, location and operating scope.
Yes. The operating boundary can range from infrastructure-only delivery to managed cluster support, allowing tenant orchestration, Kubernetes and application software to remain within the customer environment.
Tell us the platform, scale and target RFS. We'll map the requirement against Nistar's forward capacity and define the path to deployment.