AI chip startup Cerebras nabs $250 million Series F round at over $4 billion valuation

New money follows Cerebras's expansion from a single system to a portfolio of products and cloud partnerships.

Cerebras Systems, the five-year-old AI chip startup that has created the world's largest computer chip, on Wednesday announced it has received a Series F round of $250 million led by venture capital firms Edge Capital via its Alpha Wave Ventures and Abu Dhabi Growth Fund. Returning investors participating in the round include Altimeter Capital, Benchmark Capital, Coatue Management, Eclipse Ventures, Moore Strategic Ventures, and VY Capital.

The new money brings Cerebras's total raised to $750 million, and the company says it has a post-money valuation of over $4 billion.

Said co-founder and CEO Andrew Feldman in prepared remarks, "The Cerebras team and our extraordinary customers have achieved incredible technological breakthroughs that are transforming AI, making possible what was previously unimaginable. 


See also: Cerebras prepares for the era of 120 trillion-parameter neural networks.


"This new funding allows us to extend our global leadership to new regions, democratizing AI, and ushering in the next era of high-performance AI compute to help solve today's most urgent societal challenges -- across drug discovery, climate change, and much more."

Cerebras, which competes with AI chip Titan Nvidia, as well as fellow startups Graphcore and SambaNova Systems, among others, is building whole computer systems and services wrapped around a set of components based on the world's biggest chip, what Cerebras calls its Wafer-Scale Engine, or WSE.

wse2-natalia.jpg

Cerebras Systems product manager for AI Natalia Vassilieva holds the company's WSE-2, a single chip measuring almost the entire surface of a twelve-inch semiconduor wafer. The chip was first unveiled in April, and is the heart of the new CS-2 machine, the company's second version of its dedicated AI computer.

Cerebras Systems
hc2021-cerebras-embargoed-8-24-slide-10.jpg

Cerebras added to its previously announced CS-2 AI computer with a new switch product, the SwarmX, that does routing but also calculations, and a memory computer containing 2.4 petabytes of DRAM and NAND, called MemoryX.

Cerebras Systems

In the spring, the company unveiled the second version of the WSE, packing 2.6 trillion transistors onto a surface taking up almost the entirety of a 12-inch semiconductor wafer.

The WSE-2 is housed in the company's dedicated rack-mountable CS-2 server computer. In August, Cerebras added to its offerings with a dedicated memory unit, MemoryX, and a dedicated high-bandwidth switch that connects multiple CS-2 units to the MemoryX, called SwarmX.


See also: Mammoth AI report says era of deep learning may fade, but that's unlikely.


In September, the company announced a partnership with cloud operator Cirrascale for those who want to rent the CS-2 before rather than purchasing outright.

Cerebras says its infrastructure portfolio will compute very large neural networks of perhaps trillions of parameters much better than computing offerings with less horsepower. The company also claims top results on non-AI problems such as complex combinatorial problems in physics and other basic scientific research. 

Cerebras started out with customers at the U.S.'s national research laboratories and has branched out to commercial enterprises as well. The customer list now includes Argonne National Laboratory, Lawrence Livermore National Laboratory, Pittsburgh Supercomputing Center () for its groundbreaking Neocortex AI supercomputer, EPCC, the supercomputing center at the University of Edinburgh, Tokyo Electron Devices, GlaxoSmithKline, and AstraZeneca.