Luminal, a startup that builds a compiler and cloud service for running AI models on hardware other than Nvidia's, has raised a $25 million Series A led by Standard Capital, according to an announcement the firm published Nov. 12, 2026. A compiler, in this context, is software that translates a model into instructions a specific chip can execute efficiently.
The announcement ran under the name of Dalton Caldwell, whom Standard's partner biographies list as co-founder and general partner. Some funding coverage describes him as managing partner.
Why Standard backed it
"We led its $25 million Series A because better software can unlock more useful computation from chips companies have already bought," Standard wrote in the post. The firm added that the founders wanted a fundraising process that left them free to keep building.
Details beyond that are thin. Individual check sizes, ownership stakes and any board changes were not available. Tech Funding News reported that returning investors took part but, in the material reviewed, did not say who they were.
The numbers don't quite agree
Not every outlet landed on the same figure. SiliconANGLE's headline cited $23 million, while Standard, TNGlobal, Tech Funding News and TelcoNews all reported $25 million. Tech Funding News put Luminal's total funding at $30.3 million. That squares with the earlier seed round, which TechCrunch reported at $5.3 million on Nov. 17, 2025, ahead of Luminal's own announcement the next day.
A Form D, the notice companies file with the Securities and Exchange Commission when they sell securities privately, was filed Oct. 7, 2026.
What Luminal sells

The company makes a compiler that tunes AI models for different kinds of hardware, then runs a cloud service on top of it. Its website lists licensed cloud and on-premises deployments, paired with engineering support, custom optimization and negotiated service-level agreements. Pricing information on the site, it should be said, is not evidence of revenue or contract terms.
Luminal has also published its own benchmarks, among them a post on reducing image-generation costs on AMD hardware. It is a company-produced test rather than independent validation, and the cost result hinges on the rental-price assumptions Luminal chose.
A May 18, 2026 blog post announced a partnership with Positron AI, and a later post promised a follow-up with more precise performance figures.
Felicis, a seed investor, wrote in November 2025 that typical GPU utilization runs 10% to 20%, compared with more than 80% with Luminal. Those are claims from an investor and the company, and they are nearly a year old.
A crowded corner of the stack

Luminal is hardly alone in trying to loosen Nvidia's grip on AI software. Modular, whose acquisition by Qualcomm has been completed, works on related problems, as does Standard Kernel, which has announced its own seed round. SCALE, for its part, publishes documentation for running code on non-Nvidia hardware.
Who the customers might be
In a Nov. 12, 2026 interview, Luminal co-founder Joe Fioti pointed to neoclouds (newer cloud providers built around GPU rental), foundation-model laboratories and inference providers as likely license buyers. That account comes from an auto-generated TBPN Digest transcript, which may contain errors.
Luminal did not disclose headcount or revenue in the material reviewed, so how much of that pitch has already turned into business remains an open question.
