Blab Business
Cerebras Systems Secures $20 B+ Deals with OpenAI and AWS, but Valuation and Margin Risks Persist
Cerebras Systems Inc. (NASDAQ: CBRS) announced in March 2026 that it has struck multi‑year agreements with OpenAI and Amazon Web Services (AWS) worth more than $20 billion. The contracts, centered on high‑speed AI inference, promise to boost the company’s wafer‑scale chip business while leaving its valuation, profitability and margin outlook under scrutiny.
At the heart of Cerebras’ offering is the Wafer‑Scale Engine 3 (WSE‑3), the largest AI processor ever built. The chip spans 46,225 mm², contains roughly 4 trillion transistors, and can draw up to 25 kW of power. It is 58 times larger than a typical GPU and is produced exclusively by TSMC, the only foundry capable of handling such a wafer‑scale device. Its switched‑fabric network, which runs across the entire wafer, eliminates interconnect bottlenecks and gives it a distinct edge for inference workloads that demand low latency.
The OpenAI deal, signed in 2026, is a multi‑year commitment for more than $20 billion of Cerebras compute and includes warrants that allow OpenAI to purchase Cerebras stock at a discount. The AWS collaboration, announced on the same day, will pair Cerebras chips with AWS’s cloud infrastructure to deliver the fastest AI inference on Bedrock, Amazon’s generative‑AI platform. Both agreements are expected to create significant new revenue streams and reinforce Cerebras’ position as a challenger to Nvidia in the inference market.
Cerebras’ revenue mix remains heavily concentrated. In 2025, the Mohamed bin Zayed University of Artificial Intelligence accounted for 62 % of revenue, G42 for 24 %, and the newly signed OpenAI and AWS contracts are projected to make up the remaining 14 %. Its cloud‑services segment, which offers “AI inference cloud” and “AI training cloud” APIs, grew 178 % year‑over‑year in 2025, according to the latest earnings report. However, the company warned that margins in the cloud‑services line are likely to shrink by 10–15 basis points as it invests in additional capacity.
Despite the headline‑making contracts, CBRS remains unprofitable and its enterprise‑value to sales ratio sits at roughly 72×, the highest among AI‑hardware peers. The analyst who wrote the accompanying research notes that the company’s valuation is “high” and that the risk‑reward profile is unattractive in the short term. The research also highlights that capital expenditures are expected to rise sharply as Cerebras expands its data‑center footprint and develops new wafer‑scale chips.
From a risk‑management perspective, the company’s exposure to a small customer base, the capital intensity of wafer‑scale manufacturing, and the uncertainty surrounding the long‑term profitability of its cloud services are key concerns. The research analyst, who holds an MSc in Applied Risk Management and has experience in financial analysis at leading firms, recommends a hold rating for CBRS, citing strong long‑term growth potential but noting that the current premium valuation and unprofitability make the investment unattractive.
In summary, Cerebras Systems has secured two landmark contracts that could accelerate its growth and strengthen its competitive position against Nvidia and other AI‑hardware vendors. Yet the company’s high valuation, concentrated customer base, and expected margin compression pose significant risks. Investors will be watching the next earnings cycle for updates on revenue growth, margin performance, and any changes to the company’s capital‑expenditure plans.
The company’s next quarterly earnings report is scheduled for early October 2026, at which time analysts will assess whether the OpenAI and AWS deals translate into the projected revenue and margin improvements. Until then, the risk‑reward balance for CBRS remains a key consideration for investors and stakeholders.
At the heart of Cerebras’ offering is the Wafer‑Scale Engine 3 (WSE‑3), the largest AI processor ever built. The chip spans 46,225 mm², contains roughly 4 trillion transistors, and can draw up to 25 kW of power. It is 58 times larger than a typical GPU and is produced exclusively by TSMC, the only foundry capable of handling such a wafer‑scale device. Its switched‑fabric network, which runs across the entire wafer, eliminates interconnect bottlenecks and gives it a distinct edge for inference workloads that demand low latency.
The OpenAI deal, signed in 2026, is a multi‑year commitment for more than $20 billion of Cerebras compute and includes warrants that allow OpenAI to purchase Cerebras stock at a discount. The AWS collaboration, announced on the same day, will pair Cerebras chips with AWS’s cloud infrastructure to deliver the fastest AI inference on Bedrock, Amazon’s generative‑AI platform. Both agreements are expected to create significant new revenue streams and reinforce Cerebras’ position as a challenger to Nvidia in the inference market.
Cerebras’ revenue mix remains heavily concentrated. In 2025, the Mohamed bin Zayed University of Artificial Intelligence accounted for 62 % of revenue, G42 for 24 %, and the newly signed OpenAI and AWS contracts are projected to make up the remaining 14 %. Its cloud‑services segment, which offers “AI inference cloud” and “AI training cloud” APIs, grew 178 % year‑over‑year in 2025, according to the latest earnings report. However, the company warned that margins in the cloud‑services line are likely to shrink by 10–15 basis points as it invests in additional capacity.
Despite the headline‑making contracts, CBRS remains unprofitable and its enterprise‑value to sales ratio sits at roughly 72×, the highest among AI‑hardware peers. The analyst who wrote the accompanying research notes that the company’s valuation is “high” and that the risk‑reward profile is unattractive in the short term. The research also highlights that capital expenditures are expected to rise sharply as Cerebras expands its data‑center footprint and develops new wafer‑scale chips.
From a risk‑management perspective, the company’s exposure to a small customer base, the capital intensity of wafer‑scale manufacturing, and the uncertainty surrounding the long‑term profitability of its cloud services are key concerns. The research analyst, who holds an MSc in Applied Risk Management and has experience in financial analysis at leading firms, recommends a hold rating for CBRS, citing strong long‑term growth potential but noting that the current premium valuation and unprofitability make the investment unattractive.
In summary, Cerebras Systems has secured two landmark contracts that could accelerate its growth and strengthen its competitive position against Nvidia and other AI‑hardware vendors. Yet the company’s high valuation, concentrated customer base, and expected margin compression pose significant risks. Investors will be watching the next earnings cycle for updates on revenue growth, margin performance, and any changes to the company’s capital‑expenditure plans.
The company’s next quarterly earnings report is scheduled for early October 2026, at which time analysts will assess whether the OpenAI and AWS deals translate into the projected revenue and margin improvements. Until then, the risk‑reward balance for CBRS remains a key consideration for investors and stakeholders.
