Jonathan Ross left Google to launch next-generation semiconductor startup Groq in 2016. Today, the Mountain View, California-based firm said that it had raised $300 million led by Tiger Global Management and billionaire investor Dan Sundheim’s D1 Capital as it officially launched into public view. The new funding values Groq at more than $1 billion – though just how much more Ross declined to say

Over the past five years, Groq developed its first AI chip and got it into the market, with customers in both data centers and autonomous vehicles, with help from $67 million funding. It’s now working on a second-generation offering that it plans to get into production. “We have a lot more customer interest than we can service,” says Ross, the company’s chief executive officer.

Ross, 39, had previously helped invent Google’s tensor processing unit, or TPU, which underpins its machine-learning software for image recognition and other tasks. His cofounder, Douglas Wightman, also came from Google, but left Groq more than two years ago.

Groq’s chips are next-generation ones that are geared towards so-called inference tasks. They use knowledge from deep learning to make new predictions on data. Groq says that its chips, called tensor streaming processors, are 10 times faster than competitors. “It’s the most powerful one ever built,” Ross says. “It’s not just how many operations per second, but latency.”

Chipmakers have been in a race to power the rapid development in AI applications. Nvidia, whose chips were first invented for rendering video games, has been in the lead. Earlier this week it introduced a new chip, called Grace, designed to boost artificial intelligence and other high-performance computing.

“The nature of problems that need to be solved computationally has changed and changed in ways that is stressing the existing architecture,” says Andy Rappaport, a longtime founder and investor in semiconductors, who came out of retirement to join Groq’s board of directors last year. “We are back in this period of chaos, and those are the periods where the brand-name companies in computation get established.”

The complexities of machine learning — and applications like autonomous vehicles — require more computing power and speed. “With machine learning, it’s probabilistic. If this is likely happening, you should probably do this, and if that is happening you should probably do that. That is very hard for machines to deal with,” Ross explains. “When it’s probabilistic you have to complete all the possible computations and weigh each one a little bit, which makes it dramatically more expensive to do.”

“The nature of problems that need to be solved computationally has changed and changed in ways that is stressing the existing architecture.”

The ex-Googlers got Groq off the ground with early financing from venture capitalist Chamath Palihapitiya, who told CNBC in 2017 that he first learned of the Google chip on an earnings call. Since then, Groq has finished research on its AI chip and brought the technology (produced in an American foundry) to market. “What attracted me to Groq is that their architecture is really different, and it’s really different in ways that really matter,” Rappaport says. “They’ve thought through the problems that the existing architecture can’t solve.”

In a recent video call, Ross showed off the design of Groq’s chip, which appears simpler though it can perform one quadrillion operations per second. That design includes a single enormous processor with hundreds of functional units rather than a small programmable core that is replicated dozens or hundreds of times.

Ross says that Groq’s customers include large companies in financial services and autonomous driving, though he would not divulge any customer names. Having customers in both areas is “unusual,” he says, because most semiconductor companies will have to optimize their chips for one or the other.

With the new funding, Ross says he hopes to expand to new markets (“there are several promising ones not being served by GPUs or CPUs,” he says). He also plans to double the number of employees, to 250 people from 122, by the end of the year – his main reason for talking at all. “We really like to stay as quiet as we can,” he says. “The only reason we are [talking] is we need to hire more people. If we hire 28 people out of 2,000 applicants we need a lot of applicants. If it weren’t for that, we wouldn’t be saying anything.”

The Financial Times had previously reported that Groq and Tiger were in discussions about the funding.



READ NEWS SOURCE

LEAVE A REPLY

Please enter your comment!
Please enter your name here