I’ve covered Cerebras from different angles since its IPO. First, I looked at the sharp reality check that followed its massive $5.5 billion debut. Then I covered Morgan Stanley‘s first-ever price target for the stock.
When you use an AI tool, how much does it matter that you get an answer in seconds rather than waiting longer?
Cerebras (CBRS) sells chips that deliver that speed. When it went public on May 14, it priced its IPO at $185 per share, then opened at $350. By its first close, the company was valued at nearly $95 billion against 2025 revenue of $510 million. The stock closed at $311.07 that day. It now trades around $108.72, 65% below that first close and well below its IPO price.
But the business is growing fast. Third-quarter core revenue was $230.9 million, up 70% from a year earlier, and contracted future revenue reached $26.1 billion.
CFO Bob Komin said Cerebras “exceeded our guidance across the board” for a third straight quarter, and that its pipeline for high-speed AI inference grew more than tenfold in the quarter. However, a delay in its largest contract complicates the story.
Also Read: Cerebras Systems Inc. Q3 2026 Earnings Call: Live Updates of $CRBS Earnings, Stock Forecast
Cerebras beats estimates, but the cloud ramp is late
Shared with me at TheStreet, Morgan Stanley kept an Overweight rating and $178 price target in a note titled “Strong numbers overall as mix shifts from cloud to hardware.”
Morgan Stanley came away slightly more cautious about the story’s foundation, even as numbers beat estimates. The stock closed at $97.75 on Nov. 12, then rose 11.22% the next day to $108.72, which leaves the price target 64% higher.
More Artificial Intelligence:
- AI investments fail when companies ignore the hard part of transformation
- Morgan Stanley issues a fresh warning about AI stocks
- CrowdStrike gives companies a ‘kill switch’ as bots go off script
Core revenue beat the bank’s $216.3 million estimate, gross margin came in about 2 points above the 41 % estimate, and non-GAAP earnings per share were $0.12 versus an expected loss of $0.21.
December (fourth quarter) guidance according to Cerebras is far stronger than expected: core revenue of $310 million to $330 million against Morgan Stanley’s $268.1 million target, with gross margin near 48% and operating margin near break-even. Management still expects revenue to triple in 2027.
What’s delaying the buildout
The problem is buildings, not chips. Cerebras says silicon and hardware supply are sufficient, but it is behind on the first batch of its largest contract because sites, power, and finished data center buildings are not ready.
Cloud revenue is guided roughly flat for the fourth quarter, and most new capacity now arrives in mid-2027, not early 2027. Cerebras has 700 megawatts under contract for delivery by the end of 2027, up from 600, and expects to return to schedule in the second half of 2027.
Related: Cerebras stock faces sharp reality check after massive $5.5B IPO debut
Data intelligence firm Sightline Climate found that up to 11 gigawatts of 2026 capacity still has no sign of construction, and as many as half of the world’s data center projects due this year could run late due to power limits, equipment shortages, and community opposition.
I see the same challenge from the other side of the AI boom. Faster chips can make AI services more responsive, but they still depend on data centers with enough power and infrastructure to run them. Better technology alone cannot remove the physical bottlenecks holding the industry back.
Delays have a price. Reported contracted revenue already deducts estimated delay credits, which the customer can apply to future purchases. Cerebras has called the amount small. Morgan Stanley estimates $0.6 billion to $0.9 billion, about 3%.
If a shortfall lasts 90 days, either side can cancel the affected capacity, though Cerebras estimates less than 5% of committed capacity is at risk in 2026. There have been no cancellations to date. Morgan Stanley’s concern is that customers have other options for fast inference.
Morgan Stanley’s $178 target and what’s offsetting the delay
New business means doing more work. Cerebras signed a cloud deal with Microsoft, its third hyperscaler win in six months, alongside Amazon Web Services and Oracle. Microsoft revenue starts in the first half of 2027. Oracle hardware orders begin shipping in the fourth quarter.
Technology is also moving. Cerebras says its work with AMD can deliver up to five times more AI output from the same power; its CS4 system is shipping, and the next-generation CS5 chip is due in early summer 2027. Its pitch is speed: about 300 tokens, roughly words, per second per user, at a lower cost than slowed-down GPUs.
Morgan Stanley forecasts $952 million in core revenue in 2026, $3.00 billion in 2027 and $6.67 billion in 2028, with EPS of $0.08 and $1.81 for the first two years. Its $178 target applies 8 times 2028 revenue.
Pressure on the stock isn’t only operational. After lock-up expirations in late September and early November, executives including the COO and CFO filed notices to sell tens of millions of dollars in stock.
Morgan Stanley sees enough in new customers and margin recovery to support the shares here, but wants to see capacity deliveries catch up before getting more positive.
