Cerebras — Publicly Disclosed Customers and Named Production Users

As of August 20, 2026. Ordered from most important to least important using a subjective blend of disclosed relationship size, apparent compute commitment, and the size/strategic importance of the customer organization. This is not a revenue ranking. “Production use” means the organization publicly uses Cerebras technology, but the available source does not prove that it is a directly invoiced customer.

Rank	Customer / end user	Public evidence and relationship	Evidence level
1	OpenAI	Cerebras’s prospectus identifies OpenAI as a significant customer. The companies disclosed a multi-year agreement worth more than $20 billion for 750 MW of Cerebras compute, with deployments beginning in 2026. Cerebras’s Q1 and Q2 calls describe OpenAI as an enormous customer and the main near-term cloud ramp. Announcement · SEC prospectus	A — explicit major customer / contract
2	Amazon Web Services (AWS)	The prospectus and earnings calls identify AWS as a significant customer. AWS is deploying CS-3 systems in its own data centers and plans to offer the combined AWS Trainium + Cerebras system through Amazon Bedrock. Announcement · Q1 results	A — explicit significant customer / deployment
3	G42 / Group 42 (including Inception/Core42)	Cerebras’s prospectus identifies G42 as a significant customer: it represented 85% of 2024 revenue and 24% of 2025 revenue. G42 and Cerebras built the Condor Galaxy supercomputer network; G42’s Inception also co-developed the Jais models on Cerebras. SEC prospectus · Condor Galaxy	A — explicit major customer / historical concentration
4	Mohamed bin Zayed University of Artificial Intelligence (MBZUAI)	Cerebras’s prospectus identifies MBZUAI as a significant customer and says it represented 62% of 2025 revenue. It is also a long-running model-development and sovereign-AI partner on Jais/Jais 2. SEC prospectus · Jais 2	A — explicit major customer / historical concentration
5	U.S. Department of Defense (including DARPA)	The prospectus names the U.S. Department of Defense among organizations for which Cerebras has trained models. DARPA separately awarded Cerebras and Ranovus a $45 million contract to develop a real-time HPC/AI platform. SEC prospectus · DARPA contract	A — named training customer and awarded contract
6	Meta Platforms	Meta selected Cerebras as the fast-inference provider for the Llama API; Cerebras later described its infrastructure as powering models from OpenAI, Cognition, and Meta. Llama API announcement	B — named production platform use; economics undisclosed
7	IndiaAI Mission / Centre for Development of Advanced Computing (C-DAC)	G42, Cerebras, MBZUAI, and C-DAC announced an 8-exaflop national-scale AI supercomputer hosted and governed in India for researchers, startups, enterprises, and government institutions. Cerebras summary	B — announced national deployment; direct payer unclear
8	GSK (GlaxoSmithKline)	The prospectus identifies GSK as a training customer and features a case study in which GSK uses CS-3 systems for biological language models and drug discovery. Customer spotlight · SEC prospectus	A — explicit customer / deployed system
9	AstraZeneca	Cerebras says AstraZeneca used a CS-1 to train models in roughly 52 hours versus more than two weeks on its prior GPU setup; Cerebras also named AstraZeneca in its published customer roster. Customer spotlight	A — explicit historical customer / system use
10	TotalEnergies	TotalEnergies Research & Technology USA selected and deployed a CS-2 in Houston for multi-energy research; Cerebras has called it its first publicly disclosed energy-sector customer. Customer spotlight	A — explicit customer / deployed system
11	CrowdStrike	CrowdStrike and Cerebras disclosed a strategic collaboration under which CrowdStrike runs more powerful Falcon AIDR models on Cerebras inference; Cerebras also standardized on CrowdStrike Falcon. Customer spotlight	B — reciprocal production relationship
12	Mayo Clinic	Mayo Clinic and Cerebras built and trained a genomic foundation model on Cerebras infrastructure for diagnostics and treatment-response prediction. Customer spotlight	A — named customer / co-developed production model
13	Notion	Notion uses Cerebras inference to deliver real-time enterprise search across a platform serving more than 100 million workspace users. Announcement	B — named production use
14	Mistral AI	Cerebras powers high-speed inference in Mistral’s Le Chat product; AWS and Cerebras also list Mistral among organizations using Cerebras for demanding inference workloads. Le Chat announcement	B — named production use
15	Perplexity AI	Cerebras describes Perplexity Sonar as an early production example of its real-time inference technology. Cerebras 2025 review	B — named production use; commercial terms undisclosed
16	Cognition (including Windsurf)	Cognition chose Cerebras to power the fast tier for SWE-1.6 and the SWE-grep family inside Windsurf. Case study	A — explicit inference customer / production deployment
17	Aleph Alpha	Aleph Alpha publicly said it chose Cerebras for foundation-model, multimodal-model, and sovereign-AI work. Customer spotlight	A — explicit customer choice / model-development work
18	AlphaSense	AlphaSense partnered with Cerebras to accelerate the Generative Search architecture behind its enterprise market-intelligence workflow. Case study	A — explicit customer case study
19	ZS	ZS integrated Cerebras CS-3 systems directly into its MAX.AI enterprise agentic-AI platform. Announcement	B — platform deployment / partnership
20	Jasper	Jasper used Cerebras’s Andromeda supercomputer to train and improve generative-AI models; Cerebras explicitly called Jasper the type of customer it builds for. Announcement	A — explicit historical customer / supercomputer use
21	Quora / Poe	Poe natively supports Cerebras-powered models for its users and bot builders, with usage billed through Poe. Announcement	B — named production platform use / channel economics unclear
22	Sandia National Laboratories	Sandia has a multi-year relationship with Cerebras and deployed the Kingfisher CS-3 cluster, initially four systems with eight planned, for NNSA national-security AI workloads. Deployment	A — contract / deployed systems
23	Argonne National Laboratory	Argonne deployed CS-1/CS-2 systems in the ALCF AI Testbed and used them for cancer, genomics, and other scientific workloads. Customer spotlight	A — explicit customer / deployed systems
24	Lawrence Livermore National Laboratory	LLNL integrated Cerebras into its supercomputing infrastructure for AI and simulation workloads. Customer spotlight	A — explicit customer / deployed system
25	National Center for Supercomputing Applications (NCSA), University of Illinois	NCSA deployed a CS-2 in its HOLL-I supercomputer for large-scale AI and machine-learning work. Customer spotlight	A — explicit customer / deployed system
26	University of Edinburgh / Edinburgh Parallel Computing Centre (EPCC)	EPCC has been a Cerebras customer since 2021, upgraded from CS-1 to CS-2, and later expanded to a CS-3 training and inference cluster. Expansion announcement	A — explicit repeat customer / deployed systems
27	Leibniz Supercomputing Centre (LRZ)	LRZ deployed a Cerebras CS-2 as part of its AI supercomputing environment for scientific users in Bavaria. Customer spotlight	A — explicit customer / deployed system
28	Pittsburgh Supercomputing Center (PSC)	PSC’s Neocortex supercomputer deployed two CS-2 systems for academic and scientific AI workloads. Customer spotlight	A — explicit customer / deployed systems
29	National Energy Technology Laboratory (NETL)	Cerebras included NETL in its public customer roster and features its Cerebras-based energy and computational-fluid-dynamics work. Customer spotlight	A — explicit customer roster / system use
30	nference	A CS-2 was installed at nference’s headquarters for biomedical NLP and self-supervised model training. Announcement	A — explicit customer / installed system
31	Tokyo Electron Device	Cerebras repeatedly included Tokyo Electron Device in its 2021–2022 public customer roster. The surviving public material is less clear on whether its role was end customer, reseller, or both. 2022 customer roster	C — historical roster; role may include channel partner
32	NinjaTech AI	NinjaTech uses Cerebras inference to power fast modes in its multi-agent AI workforce, including Fast Deep Coder. Customer story	B — named production use
33	StackAI	StackAI integrated Cerebras as the default inference choice for latency-sensitive enterprise-agent workflows. Case study	B — named production integration
34	Upstage	Upstage’s Solar 31B model runs on Cerebras Inference Cloud, and the companies are collaborating on production-ready AI offerings in South Korea. Announcement	B — named cloud use / partnership
35	Rox	Rox uses Cerebras inference for low-latency revenue-agent workflows across web, Slack, macOS, iOS, and its Command interface. Case study	B — named production use
36	Tavus	Tavus uses Cerebras inference for real-time digital-twin conversations and reports materially lower time-to-first-token and higher throughput. Customer story	B — named production use
37	Delphi	Delphi integrated Cerebras to power low-latency, multimodal “Digital Mind” interactions. Customer story	B — named production use
38	OpenCall	OpenCall uses Cerebras inference in AI contact-center agents for real-time triage, verification, scheduling, and payments. Customer story	B — named production use
39	Armis	Cerebras and Armis disclosed an application-security integration intended to accelerate vulnerability analysis and remediation within developer workflows. Partner spotlight	C — product partnership; customer economics unclear
40	Operant AI	Cerebras lists Operant AI as an integration for protecting AI applications, agents, and APIs in production. Cerebras cybersecurity overview	C — integration partner; customer economics unclear
41	Norby	Norby uses Cerebras for low-latency, voice-first AI companion interactions. Customer story	B — named production use
42	Sei AI	Sei AI uses Cerebras-powered agents for real-time voice and QA workflows at regulated financial institutions. Customer story	B — named production use
43	Tako	Tako uses Cerebras inference for interactive knowledge cards and real-time data analysis. Customer story	B — named production use
44	Alex	Cerebras publicly features Alex as a startup using Cerebras inference for AI-assisted iOS development. Startup customer page	C — featured startup user
45	Cognitive Computations / Dolphin	Cerebras publicly features Dolphin’s use of its infrastructure for efficient data curation. Startup customer page	C — featured startup user

Evidence key

• A: Explicitly called a customer, directly tied to a contract/purchase, or clearly has Cerebras systems deployed.
• B: Officially named production use, model deployment, or co-developed workload; direct billing and contract size are not public.
• C: Historical customer roster, featured user, or product/channel partnership where the commercial relationship is unclear.

Pure distribution, marketplace, supplier, and technology partners are not counted unless a named end-user workload is also public. That excludes Microsoft Marketplace, IBM watsonx, Vercel AI Gateway, OpenRouter, Hugging Face, AMD, Qualcomm, Carahsoft, Flex, Sanmina, TSMC, and WhiteFiber from the ranked customer list.