Quick Summary
Table of Contents
- Cerebras (CBRS) shares declined approximately 10% during their second trading session after an explosive 68% jump on debut day, with shares priced at $185 and finishing Thursday at $311.07.
- The AI chipmaker secured $5.55 billion through its public offering, selling 30 million shares at $185 apiece with an opening price of $350.
- Trading at nearly 200x its trailing sales based on $510 million in 2025 revenue, CBRS commands a valuation over seven times higher than Nvidia’s comparable metric.
- The company’s $24.6 billion order backlog shows significant concentration risk, with $20 billion linked to one cloud computing agreement with OpenAI featuring exclusivity terms and delay-based exit provisions.
- D.A. Davidson’s Gil Luria values the company at approximately $115 per share — roughly $25 billion — consistent with its contracted backlog worth.
Shares of Cerebras Systems (CBRS) retreated approximately 10% on Friday, marking a sharp reversal from Thursday’s sensational public market debut, as market participants began scrutinizing the company’s stratospheric valuation alongside concerns about its technological limitations and customer diversity.
CBRS launched at $185 per share on Thursday, surged to an opening price of $350, and momentarily touched $385 before regulators paused trading. The debut session concluded at $311.07 — representing a 68% first-day surge. Friday morning saw shares changing hands near $279.99.
The public offering generated $5.55 billion in capital, positioning it among the most significant AI-focused market debuts in recent years. When calculated on a fully diluted basis, the market capitalization exceeded $100 billion.
That represents considerable market optimism for an enterprise generating $510 million in 2025 revenues.
At present trading levels, CBRS commands a trailing price-to-sales multiple approaching 200x. For perspective, Nvidia trades at approximately 27x. Even when factoring in Cerebras’ anticipated growth rate, this disparity raises eyebrows.
The Technology Argument
Cerebras manufactures exceptionally large AI chips. Its primary product, the Wafer-Scale Engine 3 (WSE-3), measures 58 times larger than leading GPU alternatives according to company specifications and provides inference performance up to 15 times faster than Nvidia-powered systems.
The performance benefits are demonstrable. However, market observers highlight accompanying limitations.
D.A. Davidson’s Gil Luria observed that the chip’s unconventional dimensions have thus far restricted its application to smaller, less sophisticated models. Additionally, the unique size introduces production complexities regarding yield rates — consistently manufacturing sufficient functional chips at volume remains unverified for extensive implementations.
“The system may deliver higher speed for some applications, but it is also less flexible than current deployments of AI compute,” Luria stated.
CEO Andrew Feldman countered these concerns, informing Barron’s that Cerebras currently supports larger models through private hosting arrangements and anticipates public availability within weeks.
The Customer Concentration Issue
Cerebras disclosed a $24.6 billion order backlog as of year-end 2024. Management projects converting approximately $3.7 billion of this pipeline into recognized revenue across 2026 and 2027.
However, significant risk factors exist. Approximately $20 billion of that backlog — representing over 80% — originates from a single cloud infrastructure agreement with OpenAI.
This arrangement incorporates exclusivity provisions that may constrain Cerebras’ capability to engage with competing frontier AI laboratories. Furthermore, the contract contains exit mechanisms triggered by implementation delays, potentially reducing the backlog substantially.
Luria established a preliminary valuation near $25 billion for the enterprise — translating to roughly $115 per share — derived from the backlog magnitude.
Cerebras does maintain relationships with prominent AI industry players, including OpenAI, Amazon Web Services, Meta, and IBM.
Revenue expanded 96% year-over-year to $171 million in the latest reporting period. By comparison, Nvidia’s data center segment generated $62.13 billion in its most recent quarter, advancing 75%.
Source link
