Standard Kernel, a Mountain View, CA-based startup building AI systems that generate GPU software, has raised $20 million in a Seed funding round led by Jump Capital.
The round also saw participation from General Catalyst, Felicis, Cowboy Ventures, Link Ventures, Essence VC, angels, and strategic partners, including David M. Siegel, Jeff Dean, Jonathan Frankle, Michael Carbin, Sachin Katti, Walden Yan, CoreWeave, and Ericsson Ventures.
Read More:Option Circle Raises $3M in New Funding
The company plans to use the funds to speed up development of its autonomous kernel generation platform, grow partnerships with AI-native and enterprise customers, and keep advancing adaptive systems software that improves with new models and hardware.
Standard Kernel is a startup creating AI systems that automatically generate highly optimized GPU software. The company uses AI to produce specialized GPU kernels—the basic building blocks that control how efficiently models run. By working deep in the system and optimizing for specific chips, Standard Kernel replaces generic libraries with code tailored to each workload and hardware setup.
As global investment in AI infrastructure grows, companies are spending hundreds of billions on GPU clusters. However, much of this hardware isn’t used to its full potential. Getting the most out of modern accelerators requires deep knowledge of hardware, compilers, and low-level optimization. Today, most critical code is still manually written and tuned, making it hard to keep up with faster chips and more complex AI workloads.
In partner tests, Standard Kernel showed performance gains of 80% to 4× on workloads running on NVIDIA H100 GPUs, even outperforming NVIDIA’s optimized cuDNN library in some cases.
Kernel generation has become a common benchmark for large language models, but most methods focus on simpler or higher-level tasks. Creating hardware-optimized, instruction-level kernels that match or beat the best human-engineered code is still a challenge. Standard Kernel aims to automate this process, delivering peak performance on new hardware from day one without long manual tuning.
“What excites us about Standard Kernel is that they’re using AI to tackle one of the most manual and complex parts of the system,” said Saaya Pal, Partner at Jump Capital. “Hardware is advancing quickly, but the software to get peak performance hasn’t kept up. Automating instruction-level optimization could transform how AI infrastructure scales.”
“Standard Kernel is addressing one of the biggest challenges in modern computing, optimizing deep in the system where performance really matters,” said Brian Venturo, Co-founder and Chief Strategy Officer at CoreWeave. “As AI grows, advances in these underlying layers will shape the next generation of capabilities. That technical vision and the strength of the team are why CoreWeave Ventures is proud to invest in Standard Kernel as they build the future of AI systems.”
“Kernel generation is essential for boosting AI hardware performance and efficiency,” said Dylan Patel, founder of SemiAnalysis. “As AI hardware fleets grow and hardware types diversify, Standard Kernel becomes critical for smooth deployment.”
About Standard Kernel
Standard Kernel combines expertise in machine learning, computer systems, and hardware optimization. The team includes alums from MIT, Stanford, UIUC and SJTU, and has created widely used open-source tools like KernelBench and Kernel Tree Search.
Read More:Chowbus Raises $81M in Funding
