Brain Chips: the new technology that comes incorporated into new computers

SANTA CLARA, California — Inside a cavernous room in a one-story building in Santa Clara, 2-meter-tall machines whirred behind white cabinets. The machines made up a new supercomputer.

The supercomputer, unveiled on July 20 by Cerebras, a Silicon Valley startup, was built using the company’s chips for artificial intelligence (AI) products. The chips are the size of a dinner plate, or 56 times larger than a commonly used chip for AI. Each Cerebras chip contains the computational power of hundreds of traditional chips.

Cerebras said it had built the supercomputer for G42, which it said planned to use it to create AI products for the Middle East.

Andrew Feldman, CEO of Cerebras, said his startup wanted to “show the world that this work can be done faster, with less energy and at lower cost.”

Demand for computing power and AI chips has skyrocketed this year. Tech giants like Microsoft, Meta and Google, as well as countless startups, have rushed to launch AI-powered products after the AI-powered chatbot ChatGPT went viral for the human-like prose it could generate.

But making AI products requires significant amounts of computing power and specialized chips, leading to a rampant search for more of these technologies. In May, Nvidia, the top maker of chips for AI systems, said appetite for its products — known as graphics processing units, or GPUs — was so strong that its quarterly sales would exceed forecasts by more than 50 percent. Wall St. The forecast sent Nvidia’s market value above a trillion dollars.

To get enough of the AI ​​chips, some of the major tech companies — like Google, Amazon, Advanced Micro Devices, and Intel — have developed their own alternatives. Startups like Cerebras, Graphcore, Groq and SambaNova have also joined the race.

Chips are poised to play such a key role in AI that they could change the balance of power between tech companies and even nations.

AI supercomputers have been built before, including by Nvidia. But it’s rare for a startup to believe them.

Cerebras was founded in 2016 by Feldman and four other engineers. The company has raised $740 million and is valued at $4.1 billion.

Hundreds or thousands of tiny chips are required to process a complicated AI model. In 2019, Cerebras unveiled what it claimed was the largest computer chip ever built, and Feldman has said that its chips can train AI systems 100 to 1,000 times faster than existing hardware.

G42, an Abu Dhabi company, used a Cerebras system in April to train an Arabic version of ChatGPT.

Talal Al Kaissi, chief executive of G42 Cloud, a G42 subsidiary, said the technology would allow his company to build chatbots and use AI to analyze preventive care and genomic data.

Cerebras’ technology was available and cost-effective, Al Kaissi said. Cerebras used its chips to build the supercomputer for G42 in just 10 days, Feldman said.

“The time scale has been greatly reduced,” Al Kaissi said.

YIWEN LU
The New York Times