Luminal raises USD $25 million to cut AI inference costs

Luminal has raised USD $25 million in Series A funding, led by Standard Capital, following an extension of the round with additional angel investors.

Existing backers Y Combinator, SagaVC, Liquid 2 and Felicis also participated.

The San Francisco AI infrastructure company develops a search compiler and inference cloud designed to help customers run AI models across graphics processors and application-specific integrated circuits. It will use the new capital for engineering, broader model and hardware support, and commercial growth.

Its pitch centres on a key constraint in the AI market: the cost and complexity of running inference workloads at scale. Much of the market relies on Nvidia's GPU ecosystem, while companies using other chips often face extra engineering work to tune models for different hardware.

Luminal says its compiler tests mathematically equivalent ways of executing a model on a given chip, then selects the fastest option. The resulting implementation can be reused for the same model and chip across repeated inference workloads.

The software is paired with a cloud service that lets developers deploy models without managing the underlying infrastructure. The aim is to give customers more hardware choice while reducing the cost of serving AI applications.

Since launching its public cloud after a USD $5.3 million seed round, Luminal says it has reached about 45 trillion tokens processed per month within four months. It compares that with the 400 trillion tokens processed by TogetherAI, which it describes as an approximately USD $8 billion company built over four years.

According to Luminal, a large share of those tokens is set to run on Amazon's Trainium chips through an expanding relationship with Amazon. That forms part of a broader effort to shift workloads towards alternatives to Nvidia hardware.

Luminal says its products are aimed at applications with very large inference demands, including AI coding agents, voice agents and image editing tools. These use cases require systems that can handle constant model responses across large request volumes.

Joe Fioti, Co-Founder and Chief Executive Officer of Luminal, said the company's early public development helped reveal demand for software that could run models more efficiently on different hardware.

"We started by building in the open on GitHub as AI adoption took off, and quickly saw how much demand there was for a more efficient way to run AI models across different hardware," said Joe Fioti, Co-Founder and Chief Executive Officer of Luminal.

"The cost of running AI shouldn't limit who gets to build it. We're building the infrastructure to make high-performance inference more accessible, giving developers more choice beyond NVIDIA's ecosystem without forcing them to compromise on speed or cost."

Investor view

The funding comes as investors continue to back businesses focused on the computing layer beneath generative AI applications. While much attention has centred on model developers and chipmakers, software companies that help manage cost and efficiency across hardware have become a growing area of interest.

Dalton Caldwell, Managing Partner at Standard Capital, said Luminal stood out for its ability to extract more output from a wider mix of systems.

"Inference is the most important market in the world right now, not to mention the fastest growing. We backed Luminal because of their deep technical expertise and demonstrated proof they can squeeze fantastic performance out of heterogeneous hardware. Luminal's ability to get the most out of GPUs that have already been purchased and powered is incredibly high leverage," said Dalton Caldwell, Managing Partner at Standard Capital.

The extended Series A financing brings Luminal's total funding to USD $25 million. The additional capital will support continued development of its compiler and cloud platform as the company broadens support for more models and compute systems.