Nvidia “Blackwell-Next” GPUs mentioned in Linux kernel patch - KitGuru

A recent Linux kernel patch has revealed references to NVIDIA's upcoming 'Blackwell-Next' graphics architecture, sparking industry interest. The update modifies the NVGrace-GPU VFIO driver to include ... Read original →

The 'Blackwell-Next' reference in the Linux kernel isn’t a consumer GPU teaser but a strategic move by NVIDIA to harden CXL-VFIO co-optimization for enterprise workloads. Technically, using DVSEC to validate GPU memory readiness forces upstream chipset and firmware vendors to align with 3nm EUV-era low-latency interconnect standards. Geopolitically, embedding such logic early in open-source stacks mitigates supply chain fragility amid tightening U.S.-EU AI chip export controls. Competitors like AMD and Intel will likely fast-track ROCm and oneAPI support for CXL 3.0 to challenge NVIDIA’s dominance in Grace-Hopper-style HPC deployments. Within 18 months, kernel-level pre-integration will become standard practice—not just blurring CPU-GPU boundaries but enabling CXL-native AI-as-a-Service architectures that redefine datacenter hardware procurement.