Cerebras (CBRS.US) performance and guidance both surpassed expectations: Strong hardware sales made up for delays in cloud business deployment and maintained a target price of $178

The Zhitong Finance App learned that Morgan Stanley released a research report after the artificial intelligence (AI) chip maker Cerebras Systems (CBRS.US) released its latest quarterly earnings report, saying that the company's core financial indicators completely exceeded market expectations, and hardware sales performance was strong, driving the fourth quarter revenue and profitability guidelines significantly better than expected. However, as a core business that supports the company's long-term growth prospects, the cloud computing power deployment schedule for large customers lags behind contract agreements, causing some uncertainty about future revenue growth. Despite this, Morgan Stanley maintained the stock's “overweight” rating and target price of $178, with potential upside of about 82% compared to its closing price of $97.75 on November 12.

According to financial reports, Cerebras' core revenue for the third quarter increased 70.1% year on year to US$230.9 million, up 10% month-on-month, higher than market expectations of US$214.7 million, and also exceeded the company's previous guidance range of US$214 million to US$216 million. According to US GAAP, the company's third-quarter revenue increased approximately 55% year over year to US$210.8 million.

In terms of earnings, Cerebras's adjusted gross margin for the third quarter was 41%, higher than market expectations of 39.2%; the adjusted operating margin was -14.3%, significantly better than Morgan Stanley's previous forecast of -24.4%; adjusted earnings per share were $0.12, while the market previously expected a loss of $0.15 per share, while Morgan Stanley expected a loss of $0.21 per share.

By business, Cerebras' third-quarter core cloud computing and other service business revenue increased 272.3% year-on-year to US$144.9 million, up 13.5% month-on-month, but lower than market expectations of US$153.4 million. Core hardware business revenue fell 11.2% year over year to US$86 million, up 4.6% month-on-month, significantly higher than market expectations of US$61.2 million.

In terms of performance guidance, Cerebras expects core revenue for the fourth quarter to be between US$310 million and US$330 million, with a median value of US$320 million, up 38.6% month-on-month and 86.7% year-on-year, far higher than market expectations of US$274.9 million. The adjusted gross margin is expected to reach 47% to 49%, with a median value of 48%, higher than market expectations of 41.2%; the adjusted operating margin is expected to be between -1% and 1%, which is expected to achieve a break-even at the operating level.

Meanwhile, Cerebras raised its core revenue guidance for the full year of 2026 to US$940 million to US$960 million. The core gross margin is expected to be 42% to 44%, and the core operating margin is expected to be -10% to -8%. Management still expects revenue in 2027 to triple that of 2026.

Morgan Stanley analyst Joseph Moore pointed out that the financial data released by Cerebras this time showed strong performance, and new hardware shipments drove fourth-quarter revenue guidance far exceeding expectations. As early lower-priced hardware sales agreements expired one after another, and new cloud businesses received more favorable pricing, the company's gross margin improved faster than previously anticipated.

However, the analyst said that compared to a few months ago, Morgan Stanley is more cautious about the core investment logic that supports Cerebras' long-term growth, mainly due to delays in cloud computing power deployment progress for the company's largest customer.

Cerebras revealed that due to restrictions on data center site selection, power supply, and delivery of construction facilities, the first batch of cloud computing power deployment under the company's largest contract has fallen behind the original plan. Despite an adequate supply of chips and hardware, the data center infrastructure is not ready. As a result, the company expects cloud business revenue in the fourth quarter to remain basically the same as in the third quarter.

Morgan Stanley previously predicted that the first batch of computing power in the relevant contract will be fully put into operation in early 2027, but according to the latest information disclosed by management, most of the additional computing power is expected to be launched one after another starting in mid-2027.

Currently, Cerebras has signed a contract and plans to deliver a total computing power infrastructure capacity of 700 megawatts (MW) by the end of 2027, which is higher than the 600 megawatts disclosed in the previous quarter. The company expects to catch up with the contracted delivery schedule again in the second half of 2027.

It is worth noting that since the relevant cloud computing power deployment involves binding long-term contracts, delivery delays may cause Cerebras to face certain financial compensation and even risk cancellation of part of the contract capacity.

By the end of the third quarter, Cerebras' remaining performance obligation (RPO) reached $26.1 billion, up from $25.4 billion at the end of the second quarter. This data has deducted anticipated customer deductions due to delivery delays, and has not been factored into the new cooperation agreement between Oracle (ORCL.US) and Microsoft (MSFT.US).

Although Cerebras did not disclose the exact amount of the relevant deductions, the company stated that compared to the total amount of remaining performance obligations, this amount is small, and since the relevant deductions will be distributed during the remaining contract period, the impact on revenue and gross margin is expected to be limited.

Based on its delivery schedule model and associated cost assumptions, Morgan Stanley estimates that delivery delays may reduce Cerebras' remaining performance obligations by $600 million to $900 million, with a median equivalent of about 3% of its total remaining performance obligations.

The bank further stated that up to now, the relevant contracts have not been cancelled. However, according to the terms of the contract, if the delivery of insufficient computing power continues for at least 90 days, both parties to the contract have the right to cancel the affected portion of the capacity. Cerebras expects less than 5% of the total contracted capacity promised in 2026 to be at risk of cancellation.

Morgan Stanley believes that compared to financial compensation due to delayed delivery, what is more noteworthy is whether long-term contract revenue can be realized as expected. Although the relevant contract uses a “pay-as-you-go” arrangement, actual revenue still depends on Cerebras being able to reach the agreed delivery point. If computing power deployment continues to lag behind, it may not only delay revenue recognition, but also affect the amount of contract revenue that can eventually be realized.

Based on the slowdown in cloud computing power deployment, Morgan Stanley lowered Cerebras' cloud business revenue forecast for 2027, but it still expects the company to put into operation a cloud computing capacity of about 500 megawatts in the second half of 2027.

Meanwhile, new customers and hardware orders recently received by Cerebras are becoming an important force supporting short-term performance growth.

The company announced a new cloud computing cooperation agreement with Microsoft, and the related revenue is expected to begin contributing in the first half of 2027. Following Amazon (AMZN.US)'s AWS and Oracle, Microsoft became the third major cloud computing customer acquired by Cerebras in the past six months.

Additionally, orders for new hardware, including Oracle, will begin shipping in the fourth quarter of this year and will continue until at least the first half of 2027. Oracle Cloud Infrastructure (OCI) AI inference service based on Cerebras technology is expected to launch in the summer of 2027.

Morgan Stanley believes that establishing partnerships with Microsoft, Oracle, and AWS not only reflects the recognition of Cerebras technology by large technology companies, but also helps the company expand business opportunities and reduce its dependence on a single customer.

However, hardware business revenue still depends on the delivery time of individual large orders and the progress of customer data center construction, which is relatively unpredictable. This means that against the backdrop of delays in core cloud business deployment, Cerebras' short-term performance growth will rely more on volatile hardware sales.

In terms of technological progress, Morgan Stanley remains optimistic about Cerebras' competitive advantage in high-speed AI inference.

Using a wafer-level processor architecture, Cerebras can provide different performance and cost advantages from traditional graphics processing units (GPUs) in high-speed inference scenarios. Traditional GPUs can improve the response speed of a single user by reducing the number of requests processed simultaneously, but this also reduces the overall processing efficiency of the hardware, thereby driving up the unit inference cost.

In contrast, Cerebras believes that its technology can achieve high-speed inference while maintaining a high overall throughput, has better cost efficiency in application scenarios where each user generates about 300 tokens (tokens) per second, and its advantages are expected to expand as inference speed requirements increase further.

Additionally, a decoupled AI inference solution developed by Cerebras in collaboration with AMD (AMD.US) has begun to serve workloads in actual production environments. Cerebras said that under the same power consumption conditions, the solution can generate up to about five times the amount of tokens.

In terms of next-generation products, Cerebras' CS4 systems have begun to be shipped, and the next-generation CS5 processor has also entered the system testing phase and has begun to run AI models. It is expected to be put into production in early summer 2027.

Morgan Stanley believes that these technological advances are expected to further enhance the performance advantages and unit economic benefits of Cerebras after computing power facilities are put into operation one after another in the future.

Based on the latest results and business developments, Morgan Stanley raised Cerebras' 2026 core revenue forecast from US$886 million to US$952 million, and raised the 2027 core revenue forecast from US$2.88 billion to US$3 billion. At the same time, the 2028 revenue forecast remains unchanged, which is expected to reach US$6.67 billion.

In terms of profit forecasting, the bank raised Cerebras' adjusted earnings forecast for 2026 from loss of $0.46 to profit of $0.08, raised the 2027 adjusted earnings per share forecast from $1.03 to $1.81, and is expected to increase further to $5.54 in 2028.

Morgan Stanley also raised Cerebras' 2027 core gross margin forecast from 52.1% to 54.4%, and the 2028 core gross margin forecast from 58% to 59.3%. The bank expects that as business scale expands and profitability improves, the company's adjusted operating margin will reach 19.8% in 2027, which is significantly higher than the previous forecast of 10.9%.

For stock valuation, Morgan Stanley kept its target price of $178 unchanged. The target price was based on Cerebras's 2028 adjusted revenue of approximately $6.3 billion after deducting revenue related to customer warrants, and was valued at 8 times the market-sales ratio.

The bank believes that the addition of large customers, improved profitability, and advances in AI inference technology are sufficient to support Cerebras' current stock price, but if the stock is to return to its previous high level, the market still needs to establish stronger confidence in the progress of cloud computing power deployment for its largest customer.

Under an optimistic scenario, if the computing power capacity agreed in the contract can be put into operation faster than expected, and hardware demand further exceeds expectations, Morgan Stanley believes that Cerebras stock is expected to reach 330 US dollars. Conversely, if cloud computing power deployment continues to be delayed, device utilization falls short of expectations, and hardware demand weakens, the stock's valuation under a pessimistic scenario could drop to $40.

Morgan Stanley said that although the latest financial reports and profit forecast increases provide support for maintaining the “gain” rating, the bank will remain relatively cautious about Cerebras' long-term growth prospects until it sees that the cloud computing power delivery schedule has caught up again with the contract plan and related revenue begins to be realized as expected.