AI chip startup mythic raises $125 million in bid to take on Nvidia

.

Chip startup Mythic Inc. has raised $125 million in a new funding round, money that will help support its attempt to challenge Nvidia Corp. in the lucrative market for artificial intelligence processors.

Mythic, which tapped Nvidia veteran Taner Ozcelik as its chief executive officer last year, raised the money from a group led by venture firm DCVC, according to a statement Wednesday. The investors also included New Enterprise Associates, Atreides Management, SoftBank Group Corp., Honda Motor Co. and Lockheed Martin Corp.

The company has a novel approach to crunching the large amounts of data needed to run AI software. Its technology relies on analog chips, which use far less power than the digital components that typically serve as the brains of computers.

Mythic is part of a pack of chip companies vying for a piece of the AI infrastructure market. The explosive growth of that industry has fueled Nvidia’s sales and turned it into the world’s most valuable company. But it hasn’t been easy for rivals to match that success, and Nvidia remains dominant in AI processors.

Ozcelik, who spent a decade at Nvidia, believes that Mythic’s technology has the ability to exploit a weakness in current systems: They draw too much power.

If a company can successfully field an alternative that requires much less electricity, data-center operators should embrace the approach, he said in an interview. After all, finding enough power has become a key challenge for companies adding AI computing capacity.

Mythic offers what it calls analog processing units. This type of chip uses much simpler arrangements for handling information and removes an element that typically increases power consumption, Ozcelik said.

Digital chips spend a lot of time — and energy — shifting information between the processing and memory circuitry. If processing can be done within the memory then that time and energy is no longer wasted, he said.

Mythic says its design mimics the human brain in being able to use a tiny amount of energy to achieve “exceptional powers of inference” — the stage of AI computation where models use their training to answer queries and make decisions. The basic math operation that conventional AI processors use in their AI calculations — what’s known as matrix multiplication — requires 1,000 times more energy, Mythic said.

Getting a foothold in the market will mean taking on Nvidia. Ozcelik, whose career was shaped by Nvidia co-founder and CEO Jensen Huang, said he respects his mentor’s vision — but also welcomes the challenge.

“We’re making it clear,” he said. “We’re going after Nvidia.”