Cerebras
Developer of wafer-scale AI chips and supercomputers for inference and training.
Cerebras Systems builds the largest AI semiconductors using wafer-scale integration to power deep learning. Headquartered in Sunnyvale, California, it offers hardware, cloud APIs, and dedicated data centres for high-performance computing.
- WSE-3 chips measure 215mm squared, making them the largest AI semiconductors ever built.
- OpenAI signed a $10 billion deal for 750MW of compute capacity through 2028.
- IPO in May 2026 raised $5.55 billion, valuing the company at $95 billion.
- Condor Galaxy 3 system delivers 8 exaFLOPs of performance with 58 million cores.
- Revenue reached $510 million in 2025 with net income of $87 million.
What it is
Cerebras Systems is a technology company based in Sunnyvale, California, that designs and manufactures custom silicon for artificial intelligence workloads. Unlike traditional chipmakers that produce small dies, Cerebras uses wafer-scale integration to create massive processors that cover an entire silicon wafer. The company sells these chips as part of its CS-series supercomputers, which are rack-mounted appliances designed for high-speed deep learning. It also provides cloud access through application programming interfaces for both training and inference tasks, allowing users to rent compute power without owning the physical hardware.
How it came about
The company was established in 2015 by Andrew Feldman and four co-founders who previously worked at SeaMicro. They aimed to overcome the limitations of GPU clusters by building larger, more efficient chips. Early development faced significant engineering hurdles regarding cooling and packaging, costing hundreds of millions of dollars before a working prototype emerged in 2019. Subsequent releases included the WSE-2 and WSE-3 architectures. The firm secured substantial venture capital funding throughout the 2010s and 2020s, culminating in a major initial public offering on the Nasdaq in May 2026, which valued the company at $95 billion.
The compute
Cerebras relies on its proprietary Wafer Scale Engine architecture, which eliminates the latency bottlenecks found in standard GPU clusters by placing all memory and processing cores on a single piece of silicon. The latest WSE-3 chip features 900,000 AI-optimized cores and 4 trillion transistors. These processors use static random-access memory instead of dynamic memory to reduce access times. The company builds complete systems like the CS-3 supercomputer, which can be linked into clusters to handle models with hundreds of billions of parameters. Manufacturing is handled exclusively by TSMC.
Where the GPUs are
The company operates offices in Sunnyvale, San Diego, Toronto, and Bangalore. It plans to expand its physical infrastructure significantly with new data centres in Dallas, Minneapolis, Oklahoma City, Montreal, New York, and France. These facilities will house its supercomputers to provide direct cloud services. The company has also partnered with local entities in Japan and India to support regional operations and manufacturing supply chains.
Customers and deals
Major clients include the Mohamed bin Zayed University of Artificial Intelligence and G42, which account for the majority of recent revenue. In 2026, OpenAI signed a ten-year deal worth $10 billion for 750 megawatts of capacity. Amazon Web Services also agreed to purchase CS-3 systems for its Bedrock platform. Other notable customers span academia, healthcare, and energy sectors, including Sandia National Laboratories, Mayo Clinic, and TotalEnergies.
Money and ownership
Cerebras raised over $1.1 billion in its Series G round in late 2025, reaching an $8.1 billion valuation. This was followed by a $1 billion Series H round in early 2026. The company went public in May 2026 via a $5.55 billion IPO, trading under the ticker CBRS. Financial reports for 2025 showed revenues of $510 million and net income of $87 million, though equity remained negative at -$578 million due to ongoing investment in research and development.
Its place in AI
Cerebras holds a unique position in the AI hardware market by challenging the dominance of Nvidia and AMD with its wafer-scale approach. Its systems have set records for training large language models on single devices and demonstrated superior inference speeds compared to conventional GPUs. The company's technology supports critical research in drug discovery, nuclear simulation, and climate modelling. Its partnerships with sovereign wealth-backed entities like G42 highlight its role in national AI infrastructure strategies.
Questions and criticism
The company faces competition from established giants like Nvidia, AMD, and Intel, as well as hyperscalers such as AWS and Microsoft. Its reliance on TSMC for manufacturing creates a single point of failure in the supply chain. The high cost of individual nodes, reaching up to $3 million, limits accessibility. Additionally, the massive power requirements of its systems, drawing up to 25kW per node, pose challenges for data centre energy efficiency and cooling infrastructure.
Questions people ask about Cerebras
What makes Cerebras chips different from Nvidia GPUs?
Cerebras uses wafer-scale integration to place thousands of cores and terabytes of memory on a single silicon wafer. This design reduces latency and avoids the interconnect bottlenecks typical of GPU clusters that rely on separate chips connected by cables.
Who are the main customers of Cerebras?
Key customers include the Mohamed bin Zayed University of Artificial Intelligence, G42, OpenAI, and Amazon Web Services. The company also serves research labs like Sandia National Laboratories and healthcare institutions such as Mayo Clinic.
How much did Cerebras raise before going public?
Before its IPO in May 2026, Cerebras raised over $1.1 billion in its Series G round in September 2025 and another $1 billion in Series H in January 2026. Total private financing exceeded several hundred million dollars across multiple rounds.
Where is Cerebras headquartered?
Cerebras Systems is headquartered in Sunnyvale, California. It also maintains offices in San Diego, Toronto, and Bangalore to support its global operations and engineering teams.
Good reading and sources
- Official websiteOfficial site
- To Power A.I., Start-Up Creates a Giant Computer Chip (Published 2019)The New York Times, Article
- Cerebras continues 'absolute domination’ of high-end compute, it says, with world’s hugest chip two-dot-ohZDNET, Article
- Cerebras Systems on WikipediaEncyclopedia
- More photos on Wikimedia CommonsPhotos
- Wikidata recordOpen data
The facts on this page come from the Wikipedia article "Cerebras Systems", written by its contributors and used under the CC BY-SA 4.0 licence, and from its Wikidata record (CC0). EveryNeocloud rewrote the text in its own words, with AI assistance, and compares every page with the article to make sure no passage is copied. It may be reused under the same CC BY-SA 4.0 licence with a link back.
It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-15.