Cerebras Strikes Massive AI Agreement: Unveiling a 36 Exaflop Powerhouse

**Cerebras Wins Major Cloud Deal with G42 in Abu Dhabi**

The Wafer-Scale Engine company, Cerebras, has achieved a significant milestone in its quest to become a leader in AI computing. After years of development and innovation, Cerebras has won a major deployment deal with G42, an artificial intelligence and cloud computing company based in Abu Dhabi, United Arab Emirates. This partnership is expected to propel Cerebras into the top tier of AI server providers and solidify its success in the industry.

**Cerebras’ Journey to Success**

Cerebras has been making waves in the AI industry since the launch of its first wafer-scale engine for AI. The company has made notable contributions to the field, including open-sourcing large language model training and building its own supercomputer named Andromeda. Despite these achievements, Cerebras has struggled to break into the AI server business on a large scale. While it has seen some smaller wins, securing a major data center contract has remained elusive. However, this is about to change with the new partnership with G42.

**Cerebras Secures a Game-Changing Deal**

Cerebras has recently announced that it has won a deployment deal with G42 for nine interconnected 4-Exaflop supercomputers. G42, founded in Abu Dhabi in 2018, specializes in artificial intelligence and cloud computing. As part of the agreement, Cerebras Cloud will host and manage G42’s infrastructure for its internal applications. Additionally, both companies will have the opportunity to resell excess capacity to other AI developers and users.

The initial phase of the deployment is already complete, with 32 CS-2 systems in Sunnyvale, CA, providing 2 exaflops of AI power. Notably, it took only 10 days to set up the hardware and software for the first training run, highlighting the ease of deployment offered by the Cerebras platform.

**Big AI, Done Fast, Done Easy**

One of the key strengths of Cerebras’ platform is its ease of deployment. The company claims that it only requires 1200 lines of code to distribute an AI model across any number of CS-2 systems. This stands in contrast to the more than 28,000 lines of code required to distribute the same model across thousands of GPUs. While this claim has not been independently verified, it showcases Cerebras’ commitment to simplifying the work of distributed computing through its software and hardware integration.

**Expanding the Supercomputing Power**

The initial system deployed by Cerebras will soon double in size to 4-Exaflops and 54 million compute cores within 10 weeks. Two additional US-based 4-Exaflop systems have already been contracted and will be linked together with the first instance to create a total of 12-Exaflops. Furthermore, six additional systems will be added next year, establishing what Cerebras believes will be the largest AI supercomputer in the world.

**The Winning Factors**

So why did G42 select Cerebras for this significant deployment? The decision was likely influenced by several factors. Cerebras’ fast hardware, support for open-source foundation models, and the ease of scaling out the distributed models were crucial considerations. Additionally, the availability of hardware may have been a deciding factor, given that the NVIDIA H100 is reportedly sold out for a year.

**Implications for the Industry**

Cerebras’ successful partnership with G42 represents a significant breakthrough in the AI server market. This deal is likely worth hundreds of millions of dollars, surpassing the combined revenue generated by all AI startups to date. The question now is whether other mainstream cloud providers will follow suit and opt for a CS-2 based cluster over waiting for NVIDIA’s backlog to clear. As the competition in the AI landscape intensifies, waiting can result in financial losses and missed opportunities.

**Future Prospects**

The success of Cerebras and the demand it has seen from companies like G42 indicate that there may be other ambitious cloud service providers waiting in the wings. While G42 may not be widely known, its decision to partner with Cerebras highlights the growing interest in AI technologies and the need for powerful computing solutions.

In conclusion, Cerebras’ deployment deal with G42 is a game-changer for the company and the AI server industry. With its fast hardware, ease of deployment, and expanding supercomputing power, Cerebras is poised to become a major player in the world of AI computing. As the industry continues to evolve, it will be exciting to see how Cerebras and other innovative companies shape the future of artificial intelligence.


The opinions expressed in this article are solely those of the author(s) and should not be considered as financial advice or investment recommendations. Cambrian AI Research is honored to have numerous semiconductor firms as clients, including Blaize, Cadence Design, Cerebras, D-Matrix, Eliyan, Esperanto, FuriosaAI, Graphcore, GML, IBM, Intel, Mythic, NVIDIA, Qualcomm Technologies, Si-Five,, Synopsys, and Tenstorrent. The author(s) do not have any investment positions in the companies mentioned and have no plans to initiate any in the near future. For more information, please visit our website at [](

Leave a Reply

Your email address will not be published. Required fields are marked *

GIPHY App Key not set. Please check settings

A BlueYard Conversation with Juliette Chevalier (Aragon) on the Imperatives of Web3 Success

Navigating the Uphill Battle: Projections for Superstar Cities’ Urban Real Estate Demand Throughout the Decade