Cerebras Systems
Category: AI Infrastructure
A computer systems company that builds the world's largest and fastest AI processors to accelerate deep learning training and inference at massive scale. Cerebras Systems was founded in 2016. The company is led by Andrew Feldman. Based in Sunnyvale, USA. Team size: 700+. Total funding raised: $4.8B+. Latest round: IPO. Key investors include Tiger Global, Benchmark, Fidelity Management, Alpha Wave Global, 1789 Capital, G42, Coatue, Atreides Management, Altimeter Capital.
- Founded
- 2016
- Headquarters
- Sunnyvale, USA
- Team size
- 700+
- Total funding
- $4.8B+
Value proposition
Delivers orders-of-magnitude more compute performance, memory bandwidth, and communication speed than traditional GPU clusters by using a single wafer-scale chip, drastically reducing AI training time from months to minutes.
Products and solutions
Wafer-Scale Engine 3 (WSE-3): The world's largest AI chip with 4 trillion transistors and 900,000 AI cores., Cerebras CS-3 System: An AI supercomputer powered by the WSE-3, designed for easy integration into data centers., Cerebras Cloud: A high-performance AI-as-a-Service platform for training and inference., CSoft Software Stack: A comprehensive software suite that makes wafer-scale clusters appear as a single logical device to developers., Condor Galaxy: A series of massive AI supercomputers (CG-1, CG-2, CG-3) built in partnership with G42.
Unique value
Pioneered Wafer-Scale Integration (WSI), which keeps the entire silicon wafer intact as one giant processor rather than cutting it into hundreds of small chips, eliminating the 'memory wall' and interconnect bottlenecks.
Target customer
Hyperscale cloud providers, national research laboratories, pharmaceutical giants, sovereign AI initiatives, and Fortune 500 enterprises.
Industries served
Healthcare & Life Sciences (Drug Discovery), Energy & Oil/Gas (Seismic Modeling), Government & Defense, Scientific Research (National Labs), Financial Services, Generative AI & LLM Development
Technology advantage
The WSE-3 provides 52x more on-chip memory and 7,000x more memory bandwidth than a leading Nvidia H100, allowing massive LLMs to fit on a single chip and reducing the power-hungry data movement required in traditional distributed computing.
How they differentiate
Pioneered Wafer-Scale Integration (WSI) which keeps the entire silicon wafer as a single chip, eliminating the 'memory wall' and interconnect bottlenecks found in traditional GPU clusters.
Main competitors
NVIDIA, SambaNova Systems, Groq, Graphcore
Key partnerships
OpenAI: Multi-year Master Relationship Agreement for 750MW of ultra low-latency AI inference compute, valued at $20B+ over three years., AWS: First cloud provider for Cerebras disaggregated inference solution, available through Amazon Bedrock (March 2026)., G42: Multi-billion dollar partnership to build the world's largest distributed AI supercomputers., Mayo Clinic: Collaboration on generative AI for healthcare and longitudinal patient data analysis., Qualcomm: Strategic partnership to optimize AI models trained on Cerebras for high-efficiency inference on Qualcomm chips., Argonne National Laboratory: Long-term research partnership for scientific AI applications.
Notable customers
OpenAI, AWS, G42, Mayo Clinic, Argonne National Laboratory, GlaxoSmithKline (GSK), AstraZeneca, Meta, Cognition, AlphaSense, Notion
Major milestones
Completed IPO on Nasdaq May 14, 2026 (CBRS), raising $6.38B at $56.4B valuation, Signed $20B+ Master Relationship Agreement with OpenAI for 750MW of inference compute, Partnered with AWS as first cloud provider for disaggregated inference via Amazon Bedrock, Raised $1B Series H led by Tiger Global at $23B valuation (Feb 2026), Closed $850M revolving credit facility (Apr 2026), Revenue reached $510M in 2025 with GAAP profitability of $237.8M net income
Growth metrics
Revenue grew from $290M in 2024 to $510M in 2025 (+76% YoY); GAAP net income of $237.8M in 2025; Headcount ~900+ employees as of mid-2026.
Market positioning
High-performance AI infrastructure leader targeting sovereign AI, national labs, and massive-scale LLM training/inference.
Geographic focus
North America, Middle East (specifically UAE via G42 partnership), and Europe.
Patents and IP
Holds over 500 patents and applications globally covering wafer-scale manufacturing, thermal management for giant chips, high-speed interconnects, and specialized AI data-flow architectures.
About Andrew Feldman
Andrew Feldman is a veteran entrepreneur in the data center and networking space. Prior to Cerebras, he was the co-founder and CEO of SeaMicro, which pioneered energy-efficient microservers and was acquired by AMD for $334 million in 2012. Following the acquisition, he served as Corporate Vice President and General Manager at AMD. Earlier in his career, he was the VP of Marketing and Corporate Development at Riverstone Networks from inception through their IPO in 2001.
Official website: https://www.cerebras.ai