Via zdnet.com
The AI chipmaker's wafer-scale processor promises faster inference than GPUs, but investors remain skeptical after a brutal post-debut selloff.
Cerebras Systems went public on May 13, 2026, at $185 per share, and for about 24 hours, everything looked incredible. Shares surged 68% on debut day, closing around $311. Then reality set in.
By August 2026, the stock had cratered more than 50% from its highs, at times trading below the IPO price. Now the company is leaning hard into its next-generation wafer-scale chip as the cure for what ails its stock chart, arguing that its specialized silicon will outpace traditional GPUs as AI agents become the dominant computing workload.
The chip that’s bigger than your face #
The company’s Wafer-Scale Engine 3, or WSE-3, measures 46,225 square millimeters. The WSE-3 packs 4 trillion transistors and delivers 125 petaflops of AI computing power. Cerebras claims the chip offers thousands of times greater memory bandwidth than leading GPUs, which translates to 10x to 15x faster token generation for AI inference tasks.
The company, founded in 2015 by Andrew Feldman, employs roughly 708 people and focuses exclusively on AI training and inference hardware. It has landed partnerships with OpenAI and AWS to deploy its inference capabilities.
The stock tells a different story #
Cerebras raised approximately $5.55 billion through its IPO, earning an initial valuation near $56 billion. The company raised its 2026 revenue guidance to between $880 million and $890 million.
On August 12, 2026, shares fell 16% in extended trading despite the company raising its revenue targets. Cerebras has seen its hardware revenue shift unpredictably, and operational challenges after going public have drawn scrutiny from analysts watching profit margins.
The Nvidia problem #
Cerebras’ argument is architectural. Traditional GPUs were designed for graphics rendering and adapted for AI. The memory bandwidth advantage is real and measurable, and for specific inference workloads, the performance gap could justify the switching costs for large cloud providers.
The partnerships with OpenAI and AWS suggest at least some major buyers agree. But partnerships and purchase orders are different things, and the market will ultimately judge Cerebras on whether those collaborations translate into recurring, growing revenue rather than one-time deployments.
What to watch from here #
For investors who bought at the IPO price, the stock would need to more than double from its August lows just to revisit its first-day highs. Revenue guidance of $880 million to $890 million is meaningful for a company of Cerebras’ size, but it needs to grow substantially to justify a valuation that was approaching $56 billion. Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our