Nvidia Unveils Neocloud Partnerships as DSpark Threatens LPX Attach Rates

On July 1, Nvidia launched an AI Compute Partnership offering credit support and revenue sharing to neocloud providers, with SharonAI and Firmus targeting 210,000 GPUs combined. This follows the June 30 market reaction to DeepSeek’s June 27 open-source DSpark release, which delivers 60-85% faster inference speeds and reduces demand for Nvidia’s specialized LPX decode racks alongside Rubin GPUs. While core data-center revenue hit $75.2 billion in the latest quarter and consensus fiscal 2027 forecasts remain near $343 billion, investors now weigh both the recurring-revenue potential of the new ecosystem play and the risk that software efficiencies and rival disaggregated architectures limit hardware attach rates.