Cerebras Systems Inc.
Business Model Analysis
How Cerebras Makes Money
Annual Revenue: $510.0M
Cerebras generates revenue primarily through Wafer-Scale AI Accelerators, with roughly $510.0M in annual revenue.
Core Growth Engine: Cerebras' growth strategy centers on fast inference. It is building out leased data center capacity to serve OpenAI and other model providers, distributing its service through Amazon Web Services under a March 2026 agree...
Revenue Model & Pricing
Cerebras makes money in two ways. First, it sells hardware: CS-3 systems built around the WSE-3 chip, bought by sovereign AI programs, national labs, universities, and enterprises that want to own their compute. Second, and now larger, it sells cloud and services: customers pay to run models on Cerebras systems hosted in data centers it leases, either through per-token inference APIs or dedicated capacity contracts. The biggest of these is a deal for OpenAI to use up to 750 megawatts of Cerebras compute through 2028, valued at more than $20 billion. In Q2 2026, cloud and other services produced $127.7 million of core revenue versus $82.1 million from hardware. Cerebras also sells through partners such as Amazon Web Services and G42's Condor Galaxy supercomputers.
Core Revenue Streams
Cloud & Other Services (Fast Inference)
~61% of Q2 2026 core revenuePer-token API fees and dedicated capacity contracts for running AI models on Cerebras systems in leased data centers, led by OpenAI. $127.7 million in Q2 2026.
Hardware Systems
~39% of Q2 2026 core revenueSales of CS-3 systems and clusters to sovereign AI programs, universities, national labs, and enterprises. $82.1 million in Q2 2026.
Growth Strategy
Cerebras' growth strategy centers on fast inference. It is building out leased data center capacity to serve OpenAI and other model providers, distributing its service through Amazon Web Services under a March 2026 agreement, and pricing low-latency token generation at a premium to GPU clouds. IPO proceeds of $5.55 billion (more with the underwriters' option) and an expandable Morgan Stanley credit facility fund that capacity. A partnership with AMD, announced in 2026, adds another route to market, while sovereign AI customers in the UAE and elsewhere continue to buy CS-3 hardware.
How Cerebras Makes Money: FAQ
How does Cerebras Systems generate revenue across its business lines?
Cerebras generates revenue primarily through direct hardware system sales of CS-3 AI supercomputers, cloud-based AI computing subscriptions (AI Model-as-a-Service and Cerebras Cloud), Cerebras Inference API token billing, and long-term supercomputer development contracts with government and commercial partners.
What is Cerebras AI Model-as-a-Service and how is it monetized?
Cerebras offers cloud-hosted access to its CS-3 clusters, charging enterprise AI labs and corporate customers hourly, monthly, or multi-year subscription retainers to train custom LLMs, fine-tune open-source models, and run high-performance scientific simulations.
How does Cerebras price its high-speed AI Inference API?
Cerebras prices its Inference API based on token consumption (per million input and output tokens), offering developers ultra-low-latency real-time inference for Llama 3, Mistral, and custom architectures at prices significantly lower than GPU cloud providers.
How much does a Cerebras CS-3 system cost?
Cerebras does not publish list prices for CS-3 systems. They are sold as complete systems, often in multi-system clusters, under negotiated contracts with sovereign AI programs, labs, and enterprises.
How important are G42 and OpenAI to Cerebras' revenue?
G42 and the UAE's MBZUAI together provided 86% of Cerebras' 2025 revenue. Going forward, OpenAI's 750 MW agreement, valued at more than $20 billion, is expected to be the largest revenue source.
What is the Cerebras Software Development Kit (SDK) and compiler?
The Cerebras SDK and CSL compiler automatically map standard PyTorch and TensorFlow deep learning neural network graphs directly onto the 900,000 cores of the Wafer-Scale Engine, eliminating the need for engineers to manually write complex distributed parallelization code.
How does Cerebras Weight Streaming technology monetize multi-trillion parameter models?
Cerebras disaggregates memory from compute through its Weight Streaming architecture, storing model parameters on external MemoryX appliances and streaming layers sequentially to the CS-3 wafer, enabling large model scaling without cluster re-architecting.
Who are Cerebras's target enterprise and scientific research customers?
Target customers include national supercomputing laboratories (Argonne National Laboratory, Lawrence Livermore National Laboratory), pharmaceutical biotech companies (GlaxoSmithKline, AstraZeneca), financial institutions, and frontier AI research labs.
How does Cerebras make money?
Cerebras generates massive revenue by selling the CS-2 and CS-3 supercomputer systems (built around their giant chips) for millions of dollars each, targeting massive pharmaceutical companies, energy firms, and government defense labs.
Related Companies
Compare Cerebras Systems Inc. With Competitors
Related Monetization Mechanics
Compare revenue streams, unit economics, and business models
NVIDIA Corporation
Micron Technology, Inc.
Advanced Micro Devices, Inc.
Texas Instruments Inc.
Intel
Coreweave
Explore Cerebras
Core profile pages, founders and related research hubs for this company.