Inside a cavernous room this week in a one-story building in Santa Clara, California, six-and-a-half-foot machines were buzzing behind white cabinets. The machines built a new supercomputer that went into operation just last month.
The supercomputer, unveiled Thursday by Cerebras, a Silicon Valley startup, is built using the company’s specialized chips, which are designed to power artificial intelligence products. The chips are distinguished by their size — about the size of a dinner plate, or 56 times the size of a chip commonly used for artificial intelligence. Each Cerebras chip packs the computing power of hundreds of conventional chips.
Cerebras said it built the supercomputer for G42, an AI company. G42 said it plans to use the supercomputer to create and operate artificial intelligence products for the Middle East.
“What we’re showing here is that there is an opportunity to build a very large supercomputer dedicated to artificial intelligence,” said Andrew Feldman, CEO of Cerebras. He added that his startup wanted to “show the world that this work can be done faster, it can be done with less energy, and it can be done with less cost.”
Demand for computing power and AI chips has skyrocketed this year, fueled by the worldwide AI boom. Tech giants like Microsoft, Meta, and Google, as well as countless startups, have rushed to roll out AI products in recent months after the AI-powered ChatGPT chatbot went viral for the chilling prose a human can generate.
But making AI products usually requires large amounts of computing power and specialized chips, leading to a fierce hunt for more of these technologies. In May, Nvidia, the leading maker of chips used to power artificial intelligence systems, said appetite for its products — known as graphics processing units, or GPUs — was so strong that its quarterly sales would be more than 50% of Wall Street estimates. The predictions brought Nvidia’s market capitalization to over $1 trillion.
“For the first time, we are seeing a huge jump in computer requirements” due to AI technologies, said Ronen Dar, founder of Run:AI, a Tel Aviv startup that helps companies develop AI models. That, he added, “created huge demand” for the specialty chips, and companies rushed to secure access to them.
To get enough of AI chips, some of the biggest tech companies — including Google, Amazon, Advanced Micro Devices, and Intel — have developed their own alternatives. Startups such as Cerebras, Graphcore, Groq, and SambaNova have also joined the race, aiming to break into the market once dominated by Nvidia.
The chips are set to play such a major role in artificial intelligence that they can change the balance of power between technology companies and even countries. On the other hand, the Biden administration has recently balanced restrictions on the sale of AI chips to China, with some US officials saying that China’s AI capabilities could pose a threat to US national security by strengthening Beijing’s military and security apparatus.
AI supercomputers have been built before, including by Nvidia. But it is rare for startups to create them.
Headquartered in Sunnyvale, California, Cerebras was founded in 2016 by Mr. Feldman and four other engineers, with the goal of building devices that accelerate the development of artificial intelligence. Over the years, the company has raised $740 million, including from Sam Altman, who leads the OpenAI artificial intelligence lab, and venture capital firms like Benchmark. Cerebras is valued at $4.1 billion.
Because the chips typically used to power AI are small—often the size of a postage stamp—it takes hundreds or even thousands of them to process a complex model of AI. In 2019, Cerebras built what it claimed was the largest computer chip ever built, and Mr. Feldman said its chips can train AI systems 100 to 1,000 times faster than existing hardware.
G42 in Abu Dhabi started working with Cerebras in 2021. It used a Cerebras system in April to train an Arabic version of ChatGPT.
In May, G42 asked Cerebras to build a network of supercomputers in different parts of the world. Talal Al-Qaisi, CEO of G42, said the cutting-edge technology will allow his company to create chatbots and use artificial intelligence to analyze genomic and preventive care data.
But the demand for GPUs was so high that it was hard to get enough to build a supercomputer. Mr. Al-Qaisi said Cerebras’ technology was available and cost-effective. So Cerebras used its chips to build a supercomputer for the G42 in just 10 days, Mr. Feldman said.
“The date range has been drastically reduced,” Mr. Al-Qaisi said.
Over the next year, Cerebras said, it plans to build two more supercomputers for the G42 — one in Texas and one in North Carolina — and, after that, six more distributed around the world. She calls this network the Condor Galaxy.
Still, startups are likely to find it difficult to compete with Nvidia, said Chris Manning, a computer scientist at Stanford University whose research focuses on AI, because the people who build AI models are used to using software that runs on Nvidia’s AI chips.
Other startups have also tried to enter the AI chip market, but many of them have “effectively failed,” Dr. Manning said.
But Mr. Feldman said he was optimistic. He said many AI companies don’t want to just lock in with Nvidia, and there’s global demand for other powerful chips like those from Cerebras.
“We hope this will push AI forward,” he said.