Cerebras Systems, a leading AI computing startup, is navigating the chip shortage landscape by adapting its business model. CEO Andrew Feldman reveals a shift toward large-scale, cloud-based AI computing services, with the completion of the Condor Galaxy 1 (CG-1), a colossal AI computer built for Abu Dhabi-based client G42. The move signifies a departure from traditional hardware sales to more extensive cloud service offerings, presenting a unique approach to addressing the challenges posed by chip shortages and the growing demand for AI computing.
Adapting to Chip Shortages:
Cerebras Systems, aiming to compete with Nvidia, has experienced challenges in acquiring GPU chips amid the global shortage. The company is adapting its business model to focus on larger cloud-based solutions rather than individual system sales.
CEO Andrew Feldman notes a significant shift in the business landscape, with clients preferring cloud services for AI computing over purchasing individual machines.
Hybrid Business Model:
Cerebras' evolving business model combines individual system sales with the provision of massively parallel systems that the company builds over months and operates as a dedicated AI cloud computing service.
Customers are inclined to utilize cloud services for shorter durations across multiple machines, marking a departure from the traditional model of purchasing standalone machines for extended computing tasks.
Condor Galaxy 1 (CG-1):
The company unveiled the CG-1, a massive AI computer built for G42, an investment firm based in Abu Dhabi. CG-1 comprises 64 of Cerebras's CS-2 computers, making it a powerful system valued at approximately $150 million, offering an alternative to Nvidia's GPU chips.
The machine is hosted at the facilities of Colovore, a California-based hosting provider, and is designed for high-performance computing tasks.
Cloud Expansion and Future Plans:
Cerebras is witnessing increased demand for its cloud services, leading to the construction of the CG-2, the second version of Condor Galaxy. CG-2 will add another 64 computers and four more exaFLOPS of computing power.
The hybrid business model includes both hardware sales and the rental of cloud computing capacity, with a focus on attracting new customers to the cloud-based offerings.
Success Measures Beyond Hardware Sales:
Success for Cerebras is not solely measured by hardware sales; the rate of new customers adopting cloud services without upfront purchases is becoming a key metric.
Pharmaceutical giant GlaxoSmithKline, an early hardware customer, is now utilizing Cerebras's cloud for specific computing tasks, exemplifying the shift in customer preferences.
Condor Galaxy's Scientific and AI Contributions:
Condor Galaxy serves as a platform for various creative uses, including AI work and scientific research. Cerebras emphasizes the uniqueness of their system's capabilities in AI and supercomputing, setting it apart from competitors.
Cerebras Systems' strategic shift to a hybrid business model, combining hardware sales with cloud-based AI computing services, reflects a response to the challenges posed by chip shortages. The completion of Condor Galaxy 1 for G42 exemplifies the company's commitment to providing innovative solutions to meet the evolving demands of the AI computing landscape.