Standard Kernel Raises $20 Million to Automate GPU Kernel Optimization for AI Workloads

New updates have been reported about Standard Kernel.

Standard Kernel has secured a $20 million seed round to advance its AI-driven platform that automatically generates ultra-optimized GPU kernels for AI workloads, positioning the company at the core of efforts to extract more performance from rapidly expanding AI infrastructure. The round was led by Jump Capital with participation from General Catalyst, Felicis, Cowboy Ventures, Link Ventures, Essence VC, CoreWeave, Ericsson Ventures, and several prominent industry angels, giving Standard Kernel both capital and strategic alignment with key ecosystem players.

The company’s technology uses AI to produce hardware-specific, instruction-level GPU kernels that replace generic libraries with code tailored to individual models and chip configurations, enabling customers to boost utilization of expensive GPU fleets without changing their models or hardware. In partner tests on NVIDIA H100 GPUs, Standard Kernel has delivered end-to-end performance gains of roughly 80% up to 4x, in some cases surpassing NVIDIA’s own cuDNN library, underscoring the potential impact on training and inference economics at scale.

By targeting one of the most technically demanding parts of the stack, the firm aims to address the widening gap between accelerating hardware innovation and comparatively static systems software, where performance is often constrained by manual tuning. Investors such as Jump Capital and CoreWeave highlight that automating low-level kernel optimization could materially reshape how AI infrastructure scales, particularly as fleets grow and hardware diversity increases.

Standard Kernel plans to deploy the new capital to accelerate development of its autonomous kernel generation platform, deepen engagements with AI-native and enterprise customers, and move toward adaptive systems software that continuously improves alongside new models and chips. The team, which includes alumni from leading technical universities and contributors to open-source benchmarks such as KernelBench and Kernel Tree Search, is expanding hiring for engineers and researchers focused on AI systems that optimize AI itself, signaling an intent to build a core enabling layer for next-generation AI infrastructure.