Cerebras Grew 92% - Investors Saw the Margins and Sold

Cerebras grew revenue 92% in Q1, beat analyst estimates by 7%, and still lost $8 billion in market cap on earnings day.

TL;DR

$193.4M in Q1 2026 revenue, beating the $180.8M Wall Street estimate and up 92% year-over-year
Q1 gross margin hit 47% but guidance drops to 36-38% in Q2 - a sequential fall of about 10 points
Full-year 2026 gross margin guided at 38-41%, compared to Nvidia at 70%+ and AMD at roughly 50%
Stock fell ~20% in a single session, now trading 47% below its $350 IPO-day opening price
10 analysts still hold "Strong Buy" ratings with an average price target of $299

The company went public in May in what was billed as the largest semiconductor IPO in history, raising $6.4 billion at $185 per share. Shares opened at $350 on day one and touched $386 within hours. Six weeks later, after the company's first earnings report as a public company, they traded near $182 - briefly dipping below the IPO price for the first time.

The revenue beat was real. The margin problem is also real.

The Numbers

Metric	Q1 2026 (Actual)	Q2 2026 (Guidance)	Full-Year 2026 (Guidance)
Core Revenue	$193.4M	~$194M	$855-865M
Revenue Growth (YoY)	92%	-	-
Core Gross Margin	47%	36-38%	38-41%
Core Operating Loss	-$3.5M	-	-28% to -32% of revenue
Net Loss	$14M	-	-
EPS	-$0.22	-	-

The EPS miss (-$0.22 versus -$0.16 expected) was modest. The margin guidance was the number that moved markets.

What the Numbers Say

Revenue Growth Has a Real Backstop

Within Q1's $193.4M total, cloud and services revenue grew 167% year-over-year to $79.8M. Hardware revenue grew 60% to $111.6M. The faster-growing segment is the one with more recurring characteristics, which is the right direction for any investor trying to model 2027 margin recovery.

The full-year guidance of $855-865M implies Cerebras needs to more than quadruple its current run rate in the back half of the year. That's aggressive. But the number isn't speculative. The $20 billion compute supply agreement with OpenAI - covering 750 megawatts of inference capacity through 2028 - provides committed demand that most chip startups would trade for. GPT-5.4 is already running on Cerebras infrastructure. OpenAI has been placing parallel bets on custom silicon from multiple vendors; the Cerebras relationship is its largest.

Stock trading chart on a tablet device showing market data Cerebras (CBRS) dropped roughly 20% on June 24, its worst single-session move since the IPO in May. Source: pexels.com

Margins Are Falling on Purpose, Says Management

CFO Bob Komin was direct about the cause. Cerebras is renting back capacity from a customer while simultaneously building out its own data center footprint. Those two costs together reduce gross margins by 10 to 15 percentage points in 2026. Basically, Q1's 47% margin was an artifact of a transition period, not steady-state performance.

CEO Andrew Feldman explained the structural constraint plainly during the post-earnings media tour: "We're trying to move at the speed of AI, and data centers move with the speed of real estate."

If the buildout finishes on schedule, margins should recover as Cerebras stops paying for rented capacity and runs infrastructure it owns. If it doesn't, the margin compression extends into 2027 and the current guidance becomes a floor rather than a temporary dip.

The IPO Math Has Corrected

The collapse from $350 to $182 isn't just bad sentiment - it reflects a genuine valuation reset. Investors who bought the IPO opening at $350 were pricing in Nvidia-adjacent gross margins at scale. The 38-41% full-year guidance breaks that assumption, at least for this year.

At $182, Cerebras trades at roughly 47x trailing twelve-month revenue of $603M. Nvidia trades at around 25x forward revenue. The premium still exists - it's just smaller. Market cap sits near $40 billion. At the IPO-day peak, it briefly touched $80 billion.

What the Numbers Don't Say

The report doesn't break out how much of the $20B OpenAI deal is executed capacity versus contracted future services. That gap matters. If OpenAI renegotiates or reduces its purchase commitments, the revenue runway implied by the guidance shrinks considerably.

There's also a structural tension in the OpenAI relationship that the earnings release doesn't address. Under the deal's terms, OpenAI holds warrants convertible to up to a 10% equity stake in Cerebras as procurement scales. That arrangement makes OpenAI both Cerebras' largest customer and a prospective top-five shareholder. The incentives usually point in the same direction, but "usually" isn't "always."

The earnings release also contains no detail on the AWS partnership announced earlier this quarter, which pairs Cerebras chips with Amazon's Trainium3 processors for what Amazon calls "disaggregated inference." That business hasn't contributed to revenue yet. When it does, the margin picture changes - but the timing isn't specified in any public document.

"We're trying to move at the speed of AI, and data centers move with the speed of real estate."
Andrew Feldman, CEO, Cerebras Systems

Stock market ticker showing red and green price data AI chip stocks have faced renewed scrutiny on margins as the IPO wave of 2025-2026 meets its first earnings cycle. Source: pexels.com

So What?

Ten analysts still rate CBRS a "Strong Buy" with an average price target of $299 - 64% above where shares trade today. Their thesis is that the margin compression is temporary: Cerebras finishes building its own data centers, stops paying for rented capacity, and gross margins recover toward or above the 47% level seen in Q1. The company's inference speed advantage - Feldman claims an order-of-magnitude lead over GPU alternatives, a claim the OpenAI relationship lends some credibility to - would then translate into pricing power and durable margin expansion.

That argument may turn out to be correct. What the Q1 report established is that it won't be resolved in 2026. Investors who bought the IPO opening at $350 are sitting on a 47% drawdown from the peak and deciding whether the $855-865M full-year revenue target confirms the investment thesis or whether it's the next number management will need to revise downward.

Sources: