Cerebras shares slide as Nvidia competition, lockup expiry weigh on AI chip outlook

Traders Union
All News
Financial News
Cerebras shares drop on pressure

Highlights

Cerebras shares fall nearly 20% this week and are now down over 50% from IPO highs as investors react to competitive and supply pressures.
OpenAI's choice to use Nvidia GPUs over Cerebras hardware for GPT-6.1 Sol 'Ultrafast' mode undermines Cerebras's positioning as a major Nvidia alternative in AI infrastructure.
Approximately 19.4 million newly unlocked insider shares, about 8% of Cerebras's outstanding total, come to market as its valuation drops to just over $39 billion from a $95 billion IPO peak.

OpenAI workload shift and newly unlocked shares pressure the stock

As first reported by CNBC, Cerebras shares tumble nearly 20% this week, extending losses to more than half from the initial surge that followed the company's IPO.

Research firm SemiAnalysis says in a Wednesday post on X that OpenAI uses Nvidia graphics processing units, rather than Cerebras hardware, to power the "Ultrafast" mode for GPT-6.1 Sol. That development weighs on sentiment because investors have backed Cerebras as a potential alternative to Nvidia in AI infrastructure.

The stock also faces added pressure from the expiration of post-IPO trading restrictions on insider holdings. According to the company's prospectus, as many as 19.4 million shares held by directors, officers, non-executive employees and other holders unlock on Wednesday, equivalent to about 8% of total shares outstanding.

Valuation reset highlights risks in AI inference market

Cerebras closes Friday at $166.43, leaving the company with a market capitalization of just over $39 billion. That marks a steep decline from the $95 billion valuation it reached at the end of its first trading day, when it came close to joining the small group of companies whose IPO debut values exceeded $100 billion.

The company makes large ASIC chips designed for inference workloads and leases computing capacity from its own data centers as a cloud service. In January, Cerebras signs a deal worth more than $10 billion with OpenAI to provide 750 megawatts of computing power through 2028.

Any loss of inference-related demand is especially significant because that segment is expected to become more important across the AI data center market. McKinsey says inference is set to overtake training as the dominant workload in AI data centers by the end of the decade, increasing the strategic importance of contracts tied to that activity.

Our earlier report on U.S. startup funding showed venture capital concentrating heavily in AI, with the biggest rounds flowing to assistants, cybersecurity, and compute/infrastructure platforms—often at multibillion-dollar valuations and with debt layered alongside equity. That overview highlighted how capital intensity is becoming a defining feature of the AI buildout, and why large financings don’t always translate into durable demand or stable valuations.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.