A petaflop of AI compute on your desk. Now available through Robust HPC — factory-built by NVIDIA, ready for local LLM fine-tuning and inference workloads up to 200B parameters.
Everything under the hood of the DGX Spark unit as sold and supported by Robust HPC.
Optional, on top of the standard unit.
Our team can integrate your DGX Spark into your existing workflow, network, and storage setup on-site — model environment setup, network/storage integration, and validation against your actual workload before we hand it over.
What customers ask most before ordering a DGX Spark.
No. DGX Spark is a fixed-configuration unit factory-built by NVIDIA. Robust HPC sells, supports, and (optionally) deploys it — it isn't run through our custom BOM/configurator process the way Robust MAX and Robust FLOW builds are.
DGX Spark carries a 1-year warranty through Robust HPC — shorter than the 3-year on-site warranty on our custom-built Robust MAX and Robust FLOW workstations, reflecting that it's a standard factory unit rather than a hand-assembled build.
Yes — custom configuration & deployment is available as an add-on. Our team can integrate the unit into your existing network and storage setup on-site. Email us to scope this out.
DGX Spark isn't yet on our online RFQ flow — email info@robusthpc.com and our team will confirm stock, pricing, and lead time directly.