Artificial intelligence startup Modular Inc. is said to be involved in discussions with investors including General Catalyst over a big funding round that would value it at about $600 million, according to an exclusive report late Friday by The Information.

It’s seeking to raise a huge amount of cash at a time when AI has emerged as one of the hottest trends in technology. Modular’s draw is that its software can offer companies building AI models an alternative to Nvidia Corp.’s expensive and in-demand graphics processing units, which are seen by many as the only real option for training AI.

Nvidia’s GPUs are in such high demand that the company can barely keep up, with the biggest cloud infrastructure providers and thousands of enterprises all keen to ramp up their AI capabilities in the wake of ChatGPT’s success. Nvidia is also a force to be reckoned with in the AI software market. Its CUDA software is used to write machine learning applications that can run only on its GPUs.

Modular gives enterprises an alternative through its software, which makes it simpler for developers to train and run machine learning models on chips designed by other companies, such as Intel Corp., Advanced Micro Devices Inc. and Google LLC. It also boosts the performance of those chips, meaning they can be used to train more powerful AI models.

According to the startup’s website, its software works by unifying the front end of popular AI frameworks such as PyTorch and TensorFlow using modular, composable components. This means developers can use custom hardware to train AI systems and later deploy them across any server or edge device.

In a 2022 interview with TechCrunch, Modular co-founder and Chief Executive Chris Lattner said the startup aims to streamline the AI development lifecycle by automating and standardizing developer workflows, similar to how DevOps streamlines traditional software development.

Holger Mueller of Constellation Research Inc. told SiliconANGLE that it was always just a matter of time until someone emerged to compete with Nvidia’s AI empire. “Because competing on the hardware side is so difficult, much of the competition is coming at the problem from the angle of software,” he explained. “Modular has built a software layer for building AI models that can be trained and run without Nvidia’s chips. Instead they can run on regular silicon CPUs from AMD and Intel. It’s not clear yet how well the platform will work, but it’s nice to see more competition arising.”

Modular’s founding team is said to be one of the main reasons why it is attracting attention from venture capitalists. Lattner boasts an impressive resume, having co-created the Swift programming language during his time at Apple Inc. Meanwhile, the other co-founder and chief product officer is Tim Davis, who previously led the development of Google Cloud’s machine learning application programming interfaces, compilers and runtime infrastructure for server and edge devices.

The Information reports that Modular is seeking funding now because it wants to capitalize on the high demand, not only for Nvidia’s GPUs, but a viable alternative to those chips.

Modular previously raised $30 million in a June 2022 seed funding round led by GV, with participation from Greylock, SV Angel and The Factory.

Image: Modular