Inside a cavernous room in a one-story building in Santa Clara, California, this week the six-and-a-half-foot-tall machines scurried behind white cabinets. The machines built a new supercomputer that became operational just last month.
supercomputer, which was unveiled on Thursday cerebrasThe Silicon Valley start-up was built with the company’s specialized chips that are designed to power artificial intelligence products. The chips stand out because of their size—as small as the equivalent of a dinner plate, or 56 times larger than a chip typically used for AI. 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 planned to use the supercomputer to create and empower AI products for the Middle East.
“What we’re showing here is that there is an opportunity to build a very large, dedicated AI supercomputer,” said Andrew Feldman, Cerebras’ chief executive. He said his start-up wanted to “show the world that it can be done faster, with less energy, with less cost.”
The demand for computing power and AI chips has skyrocketed this year due to the worldwide AI boom. Along with tech giants such as Microsoft, Meta and Google, there are also myriad start-ups. Rushed to introduce AI products in recent months after AI-Powered ChatGPT Chatbot It went viral for the amazingly human prose it could produce.
But creating AI products typically requires vast amounts of computing power and specialized chips, prompting an intense search for more of those technologies. In May, Nvidia, a leading maker of the chips used to power AI systems, said appetite for its products — known as graphics processing units, or GPUs — was so strong that its quarterly sales were among Wall Street’s most expensive. Estimated to be more than 50 percent. forecast sent to nvidia Market cap rising above $1 trillion,
“For the first time, we are seeing a huge jump in computer requirements due to AI technologies,” said Ronen Dar, founder of start-up Run:AI in Tel Aviv, which helps companies develop AI models. This has “created a huge demand” for the specialty chips, he said, and companies have “rushed to ensure access” to them.
To get enough AI chips, some of the biggest tech companies, including Google, Amazon, Advanced Micro Devices and Intel, have developed their own alternatives. Start-ups such as Cerebras, Graphcore, Grok and Sambanova have also joined the race, aiming to enter the market that Nvidia dominates.
The chips are set to play such an important role in AI that they could shift the balance of power between tech companies and even countries. The Biden administration, for one, has recently weighted restrictions On the sale of AI chips to China, some US officials say China’s AI capabilities could pose a national security threat to the United States by bolstering Beijing’s military and security apparatus.
AI supercomputers have been built before, including those from Nvidia. But it is rare for a start-up to make them.
Cerebras, which is based in Sunnyvale, California, was founded in 2016 by Feldman and four other engineers with the goal of building hardware that accelerates AI development. Over the years, the company has raised $740 million, including Sam Altman, who lead venture capital firms such as AI lab OpenAI and Benchmark. Cerebra is valued at $4.1 billion.
Because the chips used to power AI are typically small – often the size of a postage stamp – hundreds or even thousands of them are needed to process a complex AI model. it occurs. In 2019, Cerebras unveiled its claim that it is the largest computer chip ever built, and Mr. Feldman has said that its chips can train AI systems 100 to 1,000 times faster than existing hardware.
Abu Dhabi-based company G42 started working with Cerebras in 2021. It used the Cerebras system in April to train the Arabic version of ChatGPT.
In May, G42 asked Cerebras to build a network of supercomputers in different parts of the world. G42 Chief Executive Talal Al Qassi said the cutting-edge technology would allow his company to build chatbots and use AI to analyze genomic and preventive care data.
But the demand for GPUs was so high that it was hard to get enough GPUs to build a supercomputer. Mr. Al Kaisi said that Cerebras’ technology was both available and cost-effective. So Cerebras used its chips to build a supercomputer for the G42 in just 10 days, Mr. Feldman said.
“The time scale has been reduced significantly,” Mr. Al Qassi 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 will be distributed around the world. It’s calling this network the Condor Galaxy.
Chris Manning, a Stanford computer scientist whose research focuses on AI, said the start-up may still find it difficult to compete against Nvidia, in part because AI model makers work on Nvidia’s AI chips. are used to using the software, he said.
Dr. Manning said that other start-ups have also tried to enter the AI chips market, yet many have “effectively failed”.
But Mr. Feldman said he remains hopeful. That said, many AI businesses don’t want to be locked in with just Nvidia, and there’s global demand for more powerful chips like Cerebras.
“We hope this will drive AI forward,” he said.