Cerebras Systems Inc.
Acquisitions
3
Major Deals Covered
$97M
Disclosed Deal Value
Cerebras's largest disclosed acquisition here is Andromeda Supercomputer Cluster Assets for $45M in 2022. This page covers 3 major deals from 2019 to 2022, each with its price where disclosed and the reason behind it.
| Company | Year | Price | Purpose |
|---|---|---|---|
| Wafer-Scale Packaging Technology IP | 2019 | $30.0M | Acquired specialized thermal expansion compensation and micro-connector patents for wafer-scale silicon chips. |
| Andromeda Supercomputer Cluster Assets | 2022 | $45.0M | Acquired high-density datacenter infrastructure and liquid-cooling manifold systems in Santa Clara, California. |
| MemoryX Distributed Compiler Assets | 2021 | $22.0M | Acquired distributed parameter offload and compiler technology to store trillions of neural network weights externally. |
2022
$45M
Acquired high-density datacenter infrastructure and liquid-cooling manifold systems in Santa Clara, California.
Allowed Cerebras to deploy the Andromeda AI supercomputer, interconnecting 16 CS-2 systems delivering over 1 exaflop of AI compute.
Provided the cloud compute infrastructure to train open-source state-of-the-art multi-billion parameter AI models.
2021
$22M
Acquired distributed parameter offload and compiler technology to store trillions of neural network weights externally.
Allowed Cerebras systems to train multi-trillion parameter AI models without requiring complex model parallelism partitioning.
Streamlined AI model training workflows for enterprise customers, including TotalEnergies and the Mayo Clinic.
2019
$30M
Acquired specialized thermal expansion compensation and micro-connector patents for wafer-scale silicon chips.
Overcame critical thermal and mechanical stress challenges of interconnecting an entire 300mm silicon wafer into a single processor.
Enabled the successful commercialization of the Wafer Scale Engine (WSE-1 and WSE-2) processors.
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.
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.
Cerebras Systems has focused almost entirely on organic internal engineering and deep intellectual property patent development rather than large corporate acquisitions, building its wafer-scale architecture, packaging, and compiler stack from the ground up.
Cerebras has completed selective talent acquisitions (acqui-hires) of specialized deep learning compiler engineers, high-speed thermal modeling researchers, and distributed systems mathematicians from Silicon Valley research labs.
Cerebras relies on deep co-development partnerships with TSMC for advanced wafer fabrication, Qualcomm for specialized inference acceleration integrations, and G42 for global supercomputing deployments.
Cerebras holds hundreds of granted and pending patents covering wafer scribeline communication, automated defect-bypass routing, wafer-to-board high-density power delivery, micro-fluidic liquid cold plates, and Weight Streaming compilers.
Cerebras deploys capital to build out dedicated high-density AI data center pods equipped with 400GbE fiber interconnects and redundant liquid cooling to support its global Cerebras Cloud and Inference API offerings.
Following its public IPO, Cerebras intends to evaluate opportunistic acquisitions in specialized AI compiler optimization startups, proprietary memory architectures, optical interconnect technologies, and domain-specific LLM software frameworks.
During its growth, Cerebras attracted acquisition interest from major semiconductor and hyperscale cloud corporations seeking a proprietary alternative to Nvidia, but the founders and board chose to remain independent to build an enduring public company.
Cerebras funds its R&D programs through its $750+ million in venture equity funding, growing commercial hardware revenues from G42 and supercomputing labs, and proceeds from its public equity offerings.
Yes, rather than traditional acquisitions, Cerebras executes massive strategic partnerships. Most notably, they partnered with G42 (a UAE-based tech giant) to build "Condor Galaxy," a massive network of AI supercomputers.
Compare corporate acquisition records, deal values, and strategic targets
Explore Cerebras
Core profile pages, founders and related research hubs for this company.