Luminal raises $25 M Series A funding to cut AI inference costs beyond Nvidia GPUs

Luminal, a San Francisco-based AI infrastructure company, has raised $25 million in a Series A funding round led by Standard Capital.
In a statement on Thursday, Luminal said the fresh capital will support additional engineering and go-to-market efforts as the company expands compatibility with more AI models and chip platforms. It also noted that all its seed investors, including Y Combinator, Felicis, SagaVC and Liquid 2, participated in the latest round with increased investments.
Luminal develops a search compiler and an inference cloud designed to help AI models run on graphics processing units (GPUs) and application-specific integrated circuits (ASICs). Its compiler evaluates mathematically equivalent methods of executing a model on a particular chip and selects the fastest option. Once a model has been optimised for a specific chip, the configuration can be reused, making the optimisation process valuable across recurring workloads.
Its inference cloud enables developers to deploy AI models without having to manage the underlying infrastructure. Luminal said the two products allow AI companies to work across different hardware platforms while reducing the costs of running their models.
A significant part of the AI industry currently relies on Nvidia’s GPU ecosystem. However, optimising models for alternative chips often requires specialised engineering expertise and extensive tuning. Luminal’s technology aims to give AI companies access to a wider range of hardware options beyond Nvidia’s ecosystem without compromising performance or cost efficiency. The company estimates the opportunity in AI infrastructure at $5 trillion.
Joe Fioti, co-founder and CEO of Luminal, said the company initially began developing its technology openly on GitHub as AI adoption continued to grow. He added that Luminal is seeing strong demand for more efficient ways to run models across different hardware platforms, emphasising that the cost of operating AI should not restrict who gets to build with the technology.
Since launching its public cloud following a $5.3 million seed funding round in November 2025, Luminal has scaled to processing 90 trillion tokens per month within four months. A significant portion of these tokens is processed using Amazon’s Trainium chips, supported by an expanding partnership with Amazon.
Luminal said the Series A funding will be directed towards further developing its compiler and inference cloud, increasing support for additional AI models and hardware platforms, and strengthening its go-to-market operations.




