Abu Dhabis G42 and California-based Cerebras Systems have unveiled the worlds largest supercomputer for AI training

G42, an Abu Dhabi-based technology holding company, and Cerebras Systems, a California-based artificial intelligence chip developer, have presented the world's biggest supercomputer for AI training. Condor Galaxy, a network of nine linked supercomputers, provides a novel way to AI computation that promises to dramatically cut AI model training time.

Condor Galaxy 1 (CG-1) is the network's first AI supercomputer, with 4 exaFLOPs (a term for measuring computer speed equivalent to one quintillion floating-point operations per second) and 54 million cores. Two additional of these supercomputers, CG-2 and CG-3, are scheduled to be deployed in the United States early next year. With a total capacity of 36 exaFLOPs, this unique supercomputing network will change AI progress worldwide.

Condor Galaxy, according to Talal Alkaissi, CEO of G42 Cloud, a subsidiary of G42, will be utilised to address society's most urgent concerns in healthcare, energy, climate change, and other areas. The CS-2 system, powered by the world's largest and fastest AI processor, enables training huge models straightforward and quick by eliminating the complexities of distributed computing. CG-1, located in Santa Clara, California, connects 64 CS-2 systems into a single, user-friendly AI supercomputer with a training capacity of 4 exaFLOPs. Cerebras and G42 provide CG-1 as a cloud service, allowing users to benefit from the power of an AI supercomputer without having to manage or distribute models across physical systems.