Andrew Feldman, the chairman of Cerebras, is actively involved in the development of the “CG-1,” a massive horizontal AI computer designed for the client G42, known as the Condor Galaxy 1. The industry is currently in high demand for Graphic chips from Nvidia, a prominent supplier of artificial intelligence technology. Despite this, there has been no significant surge in chip sales for competitors like Intel and Advanced Micro Devices. Nevertheless, this trend may be contributing to the exploration of innovative technological designs.
As per Andrew Feldman, the co-founder and CEO of Cerebras’ Artificial computing startup Systems, which offers the substantial Artificial computer CS-2 housing the world’s largest chip, there exists a potential alternative to Nvidia. Furthermore, Nvidia has expanded the storage capacity of its “superchip” Grace-Hopper for AI applications. Several years back, Feldman and his team embarked on the journey of selling computers to compete with Nvidia’s GPUs, leading to an unexpected evolution in their business model. Cerebras has transitioned into a more hybrid entity, catering to both individual system sales and the larger scale deployment of massively parallel systems. These systems are constructed over months and operated as an AI cloud computing service on behalf of clients.
According to Feldman, the transformation in Cerebras’ business model has been profound, with a shift towards clients preferring to utilize 16 machines for shorter durations rather than investing in one or two machines for extended periods. This shift aligns with the cloud service paradigm where flexibility and scalability are paramount. The recent unveiling of Cerebras’ latest achievement in the AI domain at the SC23 computation conference in Denver signifies their ongoing progress in this field. Additionally, Cerebras recently introduced a massive computer program featuring 27 million IoT “cores,” showcasing their commitment to cutting-edge technology.
The Condor Galaxy 1, a significant AI system developed for client G42, has been completed and is poised to revolutionize the utilization of supercomputing power in various creative applications. This innovative system, located at Colovore’s facilities in Santa Clara, California, represents a new frontier in cloud computing capabilities. The upcoming iteration, Condor Galaxy 2 (CG-2), will further enhance computing power with an additional 64 servers and four exaFLOPS. With a total of 576 CS-2 computers and 654,000 AMD CPU cores, the Condor Galaxy program is expected to achieve 36 exaFLOPs in its final configuration.
Feldman emphasizes the importance of a hybrid business approach, measuring success not only in system profits but also in the percentage of new customers opting for Cerebras’ cloud services without upfront hardware purchases. This shift reflects a broader industry trend towards flexible and scalable computing solutions tailored to the unique needs of diverse clients. The collaboration with industry leaders like GlaxoSmithKline highlights the versatility and impact of Cerebras’ technology in driving innovation across various sectors.
In conclusion, Cerebras’ relentless pursuit of excellence in AI computing, coupled with their innovative business strategies, positions them at the forefront of the industry. The ongoing development of the Condor Galaxy series underscores their commitment to pushing the boundaries of supercomputing and cloud services, paving the way for a new era of AI-driven solutions.
