Cerebras Systems: Down 23% Since My Sell, And Still Not Cheap

Cerebras Systems is framed as a Sell ahead of its Q1 '26 print, with the stock trading at around 100x trailing sales despite fundamentals that are said not to justify the post-IPO valuation. The article highlights extreme revenue concentration in Abu Dhabi and heavy reliance on OpenAI warrants as key risks. Nvidia and AMD are presented as more proven, diversified AI-compute alternatives at materially lower multiples.

Analysis

The key issue is not whether AI demand exists, but whether scarce compute supply will monetize through a hardware vendor with concentrated, quasi-financialized revenue streams versus incumbents with durable enterprise pull. In this setup, the market is implicitly assigning venture-style optionality to a company whose economics still look closer to a single-customer, single-program story; that usually compresses fast once post-IPO lockup enthusiasm fades and investors refocus on quality-of-revenue and repeatability.

Second-order, this backdrop is more constructive for the large-platform GPU suppliers than it is for smaller “next-gen” challengers. If capital allocation in the AI ecosystem remains disciplined, customers will continue to prioritize proven ecosystems, software support, and resale value over novelty, which reinforces the moat of the dominant accelerators and lowers the probability that a fringe player can take meaningful share without subsidization. That dynamic also matters for supply-chain vendors: the more the market doubts new entrant durability, the less it is willing to underwrite aggressive capacity commitments around them.

The catalyst path is asymmetrical over the next 1-3 quarters: any print that shows slower-than-expected concentration improvement or a need to lean harder on warrant-like economics will likely trigger multiple compression rather than a gradual re-rate. The contrarian angle is that the bearish case may be partially crowded already, but the stock can still underperform if the market decides the story is not just expensive, but structurally financing-dependent. In that regime, valuation doesn’t need a collapse in growth to break; it only needs growth to normalize from “venture wins” to “industrial cadence.”

For NVDA and AMD, the setup is not a direct earnings catalyst from this article, but a relative one: capital that is skeptical of speculative compute exposure tends to migrate toward the names with clearer share, clearer roadmap visibility, and better evidence of recurring demand. That makes the pair trade more attractive than a naked directional short on the smaller name, because the trade can work even if AI sentiment remains positive overall.

Key Decisions for Investors

• Prefer long NVDA vs. short CBRS on a 3-6 month horizon; target a 20-30% relative move if the market re-prices post-IPO AI hardware stories toward proven platforms.
• Initiate a small tactical short or put spread in CBRS into the Q1 '26 print; use a defined-risk structure because the stock can remain disconnected until the market sees evidence of revenue deconcentration.
• Add to AMD on dips rather than chasing CBRS volatility; the setup favors the second-tier incumbent with a clearer path to share gains if buyers rotate away from speculative names.
• If you want convexity, buy CBRS downside via 2-4 month puts or put spreads around the earnings window; the catalyst is a disappointment in quality of revenue rather than absolute revenue growth.
• Avoid broad shorting of AI hardware basket; pair any bearish CBRS view against NVDA/AMD to isolate company-specific execution risk and reduce exposure to continued sector beta.