Shares of AI chipmaker Cerebras sink following first earnings report since going public

Cerebras Systems, the AI chipmaker that went public last month in one of the most anticipated tech IPOs of the year, delivered a mixed first earnings report on Tuesday that beat on revenue but missed on earnings and spooked investors with a gross margin forecast that revealed the challenges of competing with Nvidia in the AI infrastructure market. Shares dropped roughly 10% in after-hours trading.

The numbers tell a nuanced story. Revenue for the quarter jumped 92% year-over-year to $193 million, ahead of the analyst consensus of $181 million. The company's net loss narrowed to $14 million from $23.9 million a year ago. But the loss per share of 22 cents missed the Street's target of a 16-cent loss, and the real concern was the forward guidance: Cerebras expects core gross margin to shrink to between 36% and 38% in the current quarter, down from 46.5% in Q1.

That margin compression is the key signal that investors are parsing. CEO Andrew Feldman insisted the quarter represented an "outstanding" start to the fiscal year, and the revenue forecast was solid — $194 million for Q2, representing 88% year-over-year growth and above the Street's $178 million consensus. For the full year, Cerebras is projecting $855.5 million to $865 million in core revenue, which would be 69% growth at the midpoint. These are not the numbers of a company in trouble.

But the margin story reveals the structural challenge of taking on Nvidia. Cerebras's entire pitch is built on its wafer-scale engine — a dinner plate-sized chip that packs far more static random-access memory than Nvidia's GPUs, delivering what the company claims is the fastest AI inference in the world. "Fast AI is more valuable than slow AI because it is more productive," Feldman said. The technology is impressive. The business model is still being proven.

The margin squeeze comes from multiple directions. Cerebras is investing heavily in its managed cloud service, which allows customers to run their models on Cerebras-powered servers without buying the hardware outright. That's a different business than selling chips — it's more capital-intensive, with lower margins on the revenue but potentially more predictable long-term. The company is also grappling with the accounting impact of its massive OpenAI deal, which includes warrants for 33.4 million shares. When those warrants vest, their value is recorded as a sales discount, or contra-revenue, effectively reducing reported revenue. During Q1, contra-revenue was negligible, but it's expected to grow substantially as the OpenAI contract ramps up. Needham analyst Quinn Bolton flagged that one milestone trigger could vest as early as this month, which would create a significant non-cash charge.

The OpenAI relationship is simultaneously Cerebras's biggest asset and its most complex accounting challenge. The $20 billion computing supply deal announced during the quarter is enormous by any standard, and it validates Cerebras's technology at the highest level of the AI market. OpenAI is using Cerebras's cloud offering to host its Codex-Spark coding model and plans to bring more advanced models, including GPT-5.5, to the service. That's a credible customer reference that money can't buy.

But the warrant structure means that the more revenue Cerebras generates from OpenAI, the more non-cash contra-revenue it has to recognize. It's a clever deal structure that gives OpenAI significant upside while protecting Cerebras's cash position, but it creates noise in the financial statements that makes it harder for investors to assess the underlying business.

Then there's the Amazon Web Services partnership. Cerebras announced during the quarter that its chips will soon be available in AWS data centers, which dramatically expands the company's addressable market by letting AWS customers access Cerebras compute without having to migrate to a separate cloud. This is the kind of distribution deal that could accelerate revenue growth significantly, and it directly challenges Nvidia's dominance in the major cloud platforms.

The stock's trajectory since its IPO tells its own story. After pricing at $185, shares opened at $350 and closed the first day at $311.07. The company raised more than $6 billion. But the stock has been on a steady decline since, and with Tuesday's after-hours drop, it's now trading around $202 — a 42% decline from its opening day high. That's not unusual for a newly public company in a competitive space, but it does reflect the market's uncertainty about whether Cerebras can sustain its growth trajectory while managing the margin compression inherent in transitioning from a chip company to a cloud services provider.

What This Means For You: If you're an investor, the key metric to watch isn't revenue growth — which is strong — but gross margin trajectory and the timing of OpenAI warrant vesting events. The contra-revenue charges will create volatility in reported numbers that doesn't reflect the underlying business health. If you're in the AI infrastructure space, the AWS partnership is the sleeper story here: Cerebras chips in AWS data centers could give companies that don't want to build dedicated infrastructure access to an alternative to Nvidia. And if you're watching the broader AI chip market, Cerebras's margin pressures illustrate a truth that's easy to miss in the hype: building hardware is expensive, running cloud services is expensive, and competing with Nvidia means you have to do both simultaneously while your largest customer's warrants keep hitting your top line. The technology is real. The business model is still a work in progress.