This Stealthy Silicon Valley Startup Wants to Make a Big Splash in AI

A startup with no product is approaching unicorn status.

Nestled in the quiet outskirts of Silicon Valley, Cerebras Systems recently raised a whopping $60 million—the company’s third round already, according to data from PitchBook. With venture capital funding totaling well over $100 million, the startup is already valued at nearly $900 million.

The Los Altos-based firm is still in stealth, which is why you probably haven’t heard of it; the fledgling firm doesn’t actually sell anything yet. But it is developing chips for the artificial intelligence space, according to a profile in Forbes, and is poised to drive complex computing tasks for technology titans with billion-dollar budgets—Amazon, Apple, and Microsoft are the types of companies with deep hunger for such a product.

Deep learning models demand massive computing power, and that demand only increases as algorithms advance in complexity. Cerebras, founded last year, was built by veterans in the chip industry, including Andrew Feldman, formerly of SeaMicro, and Gary Lauterbach, formerly of Sun Microsystems.

Notable VC firms like Benchmark Capital and Open Field Capital have already poured money into Cerebras, lending further validation. The startup’s primary rival is apt to be Nvidia, whose shares have nearly tripled in the past year since expanding from graphics-processing chips to meet the new demands of AI computing. However, Cerebras is developing AI processing technology from the ground up instead of reworking GPUs, which the startup’s founders are convinced is only a half-baked solution.

Of course, Cerebras is not alone in its quest. Startups around the world are diving head first into the AI ocean in hopes of riding the waves of robotics, automation, and deep learning software. But many will sink, while Cerebras—with its financial safety net, its founder pedigree, and its proximity to big-ticket clients in the Valley—appears built to float.

When it finally decides to undock, that is.