Cerebras Systems (NASDAQ:CBRS) is sitting on $25.4 billion in contracted future revenue. Its stock has been acting like that number doesn’t exist. According to CNBC, shares recently sank to a post-IPO low as worries about Nvidia’s competition collided with the first waves of insider shares coming out of lockup (the window after an IPO when early holders can’t sell).
The business is easier to explain than the stock move. Cerebras designs AI processors and makes money two ways: selling them as hardware systems, or renting out their computing power through its own cloud. Amid a broader sector focus highlighted in picks like the 10 Best AI Stocks to Buy in October, the company’s pitch centers directly on speed in inference, the work of running a trained AI model every time someone types a prompt. Is the slide a verdict on the business, or just a crowded exit door? The answer depends less on total revenue than on which half of the company is doing the growing.
Speed Is the Selling Point
Cerebras’ edge is that its systems answer AI requests fast, and big names are paying for it. OpenAI picked Cerebras as a launch partner for GPT-5.6 Sol, serving it at 750 tokens per second. The customer list also includes Block, Figma, GSK, and CrowdStrike, and the second quarter added new cloud deals with Cognition and Lovable.
Partnerships widen the reach. Cerebras is working with AMD and AWS on disaggregated inference, with AWS’s version expected on Amazon Bedrock in the first quarter of 2027. Renting out that capacity puts Cerebras in the same fight for AI workloads as neocloud providers like Nebius.
The Cloud Is Eating the Hardware
The second-quarter numbers show the mix shifting fast. Core revenue (management’s non-GAAP measure) hit $209.9 million, up 103%, while GAAP revenue came in at $180.1 million, up 74%. The real story sits underneath: GAAP cloud and services revenue jumped 281% to $126.0 million, while GAAP hardware revenue was $54.1 million, versus $70.3 million a year earlier.
That shift matters because cloud revenue recurs. And it’s backed by commitments, not hope. Cerebras has more than 600 megawatts of data center capacity live or under contract for delivery by the end of 2027, plus plans to boost manufacturing capacity more than tenfold this year.
Management raised its full-year core revenue outlook to $880 million to $890 million, and it plans to more than triple revenue in 2027. Core gross margin also improved to 40.6%.
What Happens If OpenAI Shops Around?
This is the worry hanging over the stock. Cerebras’ own release names OpenAI, G42, Mohamed bin Zayed University of AI, and AWS as significant customers. When a handful of buyers drive the business, a single purchasing decision can rattle the whole thesis, and CNBC pointed to Nvidia pressure as a key force behind the recent slide.
That concern deserves weight. Nvidia has the scale and the software ecosystem to compete for any speed-focused workload, a challenge facing other AI chip challengers too.
Still, the $25.4 billion in remaining performance obligations (contracted revenue not yet recognized) doesn’t vanish because a customer adds a second supplier. And names like Cognition and Lovable show the base is slowly broadening.
Paying Today for 2027
Here’s where the story gets demanding. Cerebras trades at 217.39 times forward earnings, as of October 9, against a sector average of 23.35. In plain terms, investors are paying about $217 for every dollar of expected profit, versus roughly $23 for the typical peer.
Profits, meanwhile, haven’t fully arrived. The second-quarter net loss was just $6.9 million. But GAAP tells a harsher story: a $450.5 million net loss and a 14.2% gross margin. The $8.6 billion in cash and investments from the IPO buys plenty of time, yet the valuation only works if the 2027 tripling shows up roughly on schedule. That premium stands in sharp contrast to established chipmakers trading near sector multiples.
Hedge funds jumped in quickly, with 78 holding the stock in the most recent quarter, up from 0 in the prior one. Bears are active too: short interest stands at 11.41% of the float, a sizable bet against the shares.
A Backlog Worth Betting On?
The evidence points to a business whose cloud engine is accelerating and largely contracted, not one losing its footing. The recent slide looks driven more by lockup supply and competitive headlines than by anything in the results.
But at more than 200 times forward earnings, there’s little room for slippage. The setup suits growth investors who can stomach volatility and customer concentration. Third-quarter core revenue landing below the $214 million to $216 million guidance, or a major named customer shifting capacity away, would make that premium much harder to defend.
READ NEXT: This 6.3% Dividend Yield Looks Too Good to Ignore. Is the Stock a Buy? and 10 Best 52-Week Low Technology Stocks to Buy According to Analysts.
Hedge Fund Resource Center