Cerebras Systems’ IPO was a big success on Thursday, netting billions for itself, its founders and its big investors.
Among the big winners is major shareholder Benchmark, which owns 9.5% of the company. One of the firm’s general partners, Eric Vishria, has been on Cerebras’ board since 2016, the year the artificial intelligence chipmaker was founded, having co-led a $25 million Series A round.
But those billions only happened for Benchmark because Vishria met with the startup almost against his will, he told TechCrunch.
“It was five founders and a deck and it was our first hardware investment in 10 years,” Vishria said of that first meeting. “I was an entrepreneur for 18 months.” (Before becoming a VC, Vishria sold the social browser startup she co-founded, RockMelt, to Yahoo for a reported $60 million to $70 million in 2013.)
Benchmark is famously selective in the companies it chooses and supports hardware companies so infrequently that Vishria was kicking himself for giving Cerebras time.
“Why did I have this meeting?” he kept mumbling. At one point, he even texted his assistant, who manages his calendar, and remarked, “Why did you let me take this meeting?” Visria recalls.
But his grumpy attitude disappeared by the third slide, as co-founder and CEO Andrew Feldman laid out Cerebras’ big plans.
“The first slide is the title slide. The second slide is the team. And I said, ‘Oh, this team is really good.’ And the third slide is something like, ‘GPUs are really bad for deep learning. They happen to be 100 times better than CPUs. And as soon as he said that, a light bulb went off,'” Visria recalled. “I was like, ‘Oh, my God, of course. Why would a GPU be the right thing for AI?’
However, this was years before Google’s famous Transformer paper — the 2017 research that laid the groundwork for modern artificial intelligence — that eventually led to ChatGPT. Cerebras was showcasing a new kind of giant-sized chip designed for training artificial intelligence, something the processor world wasn’t ready to build.
Vishria was intrigued enough to discuss it with some Benchmark partners, who quickly told him they didn’t know enough hardware either. They said if he wanted that deal, he would have to bring in one of Benchmark’s original founders from the 1990s, who understood.
Undeterred, Vishria scheduled a meeting for Feldman to give founding partner Bruce Dunlevie, who talked to the founder about chip packaging and cooling, and more.
“Most of this meeting was like a dog watching TV for me,” Visria joked, because he understood very little. After the pitch, Dunlevie warned that what Cerebras was attempting would be difficult. Others have tried and failed. But he thought this team had a shot. However, he worried there would be no market for the chip.
Although Vishria didn’t fully understand the technology, he was convinced that if Cerebras “could make AI faster,” there would be a market for it, and this team had the chops to succeed, he said. They had previously sold a startup, SeaMicro, to AMD.
“The benefit of having a successful exit in the past is that it erases some of the uncertainty in the minds of venture capitalists,” Feldman tells TechCrunch. “We hadn’t just fallen off the back of a turnip truck. We were an experienced team.”
The material is hard
What followed was eight and a half years of grinding as Cerebras faced fight after fight to build its product.
Feldman and Cerebras co-founder and CTO Sean Lie had to invent new cooling methods to prevent a chip of this size from burning up while drawing power. They had to invent a machine that could drill 40 screws into the wafer at once without breaking it. And so on.
The Benchmark investor repeatedly thought to himself, “What are we doing?”
In addition, the material is expensive. At the point where the company raised half a billion dollars from a long list of investors, its brands were still under development. It had to rise again in the 2022 VC bear market.
“You don’t have much traction in the company yet, so yeah, that’s where it got really tough,” Vishria recalls.
But about 18 months ago, that all changed. Cerebras’ chips, designed for training and successfully manufactured by TSMC, the world’s largest contract chip maker, proved even better at inference — running AI models to generate answers, rather than teaching them in the first place. Just as this realization was made, the world of artificial intelligence had an insatiable thirst for this kind of computation. He had a large client and revenue.
Instead of another private round, Cerebras sought to go public in 2024, only to end up under US government scrutiny over national security concerns prompted by a large investment by its only major customer, Abu Dhabi-based cloud provider G42. Public investors were also not keen on its reliance on the G42 coupled with huge losses.
The delay was a blessing in disguise. Today, OpenAI and AWS are also big customers. Cerebras doubled revenue and reported a profit last year.
Vishria gives full props to the Cerebras team for “perseverance, resourcefulness, but also adaptability,” she says.
But that’s also a feather in the investor’s cap for finding a winner so far outside the company’s usual comfort zone. Benchmark owned 17,602,983 shares worth $3.3 billion at the IPO’s opening price of $185, and more than $5.3 billion if the first-day price is above $300. He can’t sell shares until a six-month lockup expires — a standard restriction that prevents insiders from selling immediately after a company goes public.
The company bought about 80% of those shares in early rounds for about $18 million, according to various disclosures, and Vishria confirmed to TechCrunch. It bought the rest in more expensive subsequent rounds, costing it about $250 million, Cerebras revealed in its S-1.
In total, the venerable VC firm spent perhaps $270 million on that stake, which is worth several billion or more, depending on how the stock price holds up.
VC firm employees get bonuses when investments yield big returns — like Vishria’s assistant, the one who regretted that first meeting? He laughed and said, “I think he’ll do well, very well.”
When you purchase through links in our articles, we may earn a small commission. This does not affect our editorial independence.
