Cerebras Systems Reports $584 Million Nine-Month Revenue and Amends MBZUAI AI Compute Agreement
AI-generated image: synthetic visual, not an actual depiction of events, people, or locations.
Key Takeaways
- Nine-month revenue reached $584.3 million, a 72% year-over-year increase.
- Cloud and services revenue for the third quarter rose to $141.9 million, surpassing hardware revenue.
- Amended MBZUAI contract substitutes inference systems for higher-cost training and inference clusters.
- Cash and investments totaled over $7.3 billion as of September 30, 2026.
- Research and development spending tripled to $573.5 million for the first nine months of the year.
Cerebras Systems Inc. filed its quarterly report for the period ended September 30, 2026, revealing a shift in its revenue mix toward cloud and services.
The filing included an amendment to a December 2025 purchase order with the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI). Under the revised terms, Cerebras will substitute a portion of the previously ordered inference-only systems for a cluster of CS-3 systems configured for both training and inference applications. While the total contract value remains unchanged, the number of individual inference systems will decrease to account for the higher unit cost of the training-capable hardware. The amendment specifies that this training and inference cluster is not intended for deployment or use within the United States or the United Arab Emirates, with delivery designated for a specific storage facility.
Cerebras' balance sheet showed a substantial increase in liquidity following its initial public offering, with cash and cash equivalents reaching $3.07 billion as of September 30, 2026.
The company identified several key customers in its filing, including OpenAI, G42, MBZUAI, and Amazon Web Services. Cerebras stated that its business strategy relies on the continued expansion of its customer base and the procurement of data center capacity on commercially reasonable terms. The company noted that its high-speed inference services are currently available through several partner platforms, including the AWS and Microsoft Marketplaces, IBM watsonx, and Hugging Face. The filing did not disclose the specific location of the storage facility for the MBZUAI hardware delivery.
