Luminal, a San Francisco AI infrastructure startup, has raised a US$25 million Series A led by Standard Capital. Existing investors including Y Combinator, Felicis, and SagaVC also joined the round. The company builds a search compiler and inference cloud that help AI models run across GPUs and ASICs.
Funding so far
Luminal raised a US$5.3 million Seed round in November 2025. The Series A brings its total disclosed funding to about US$30.3 million. The capital will fund engineering and go-to-market expansion.
Company context
Luminal was part of Y Combinator's Summer 2025 batch. Co-founder and CEO Joe Fioti previously worked as an Intel chip designer. The company says its public cloud now processes about 90 trillion tokens per month.
What the platform does
The compiler tests mathematically equivalent ways to run a model on a chip and picks the fastest one. Luminal's inference cloud lets developers deploy AI models without managing the hardware underneath. The service targets continuous workloads such as AI coding agents, voice agents, and image-editing tools.
A large share of Luminal's token volume runs on Amazon Trainium chips. The company is expanding that Amazon partnership to give customers alternatives to Nvidia hardware. It also plans to broaden support for additional AI models and chip platforms.