Cerebras Shares Fall 10% After Record AI Chip IPO Debut

Cerebras Systems completed the largest U.S. tech IPO since Uber, raising $5.55 billion as investors bet on alternative AI chip architectures. The company’s wafer-scale processors are positioned as faster inference systems than traditional GPU-based infrastructure.

By Samantha Reed Published:

Cerebras Systems shares fell 10% on Friday following a volatile market debut after the company completed the largest initial public offering by a U.S. technology firm since Uber’s 2019 listing.

The AI semiconductor company sold 30 million shares at $185 each, raising approximately $5.55 billion in Thursday’s IPO. Shares surged during the first trading session before closing at $331.07, valuing the company at roughly $95 billion.

Cerebras develops large-scale AI processors and computing systems designed to train and run artificial intelligence models faster than conventional GPU-based infrastructure. The company has focused particularly on AI inference workloads, where models generate live responses for users in applications such as chatbots, enterprise automation systems, and real-time AI services.

Its core product is the Wafer Scale Engine 3, a processor built from an entire silicon wafer instead of assembling many smaller chips together. Cerebras argues the architecture delivers significant performance advantages over traditional GPU systems by reducing communication bottlenecks between processors.

The company has positioned its hardware as an alternative to Nvidia’s dominant AI accelerator ecosystem at a time when demand for AI infrastructure continues to surge globally.

Despite investor enthusiasm surrounding the IPO, some analysts remain skeptical about the long-term commercial potential of Cerebras’ wafer-scale approach. Analysts at investment banking firm Davidson described the technology as “niche-y” ahead of the offering and warned that the platform remains in the early stages of maturity.

The analysts said the architecture may outperform traditional systems in certain specialized workloads but could prove less flexible than established GPU-based infrastructure already widely adopted across the AI industry. Nvidia currently dominates much of the AI accelerator market through its CUDA software ecosystem and large-scale GPU deployments.

Nevertheless, the public listing also significantly increased the wealth of the company’s leadership team. CEO Andrew Feldman and CTO Sean Lie now hold stakes valued at approximately $3.2 billion and $1.7 billion, respectively.

AI & Machine Learning, Cloud & Infrastructure, News