UAE-based G42 has teamed up with US chipmaker Cerebras Systems to spearhead the deployment of an 8 ExaFLOPS supercomputer in India. This groundbreaking initiative was unveiled at the India AI Impact Summit held in New Delhi and aims to provide substantial computational power to educational institutions, government bodies, and small and medium-sized enterprises (SMEs) within the country. The project adheres to local data residency and security regulations, ensuring data remains within Indian borders.
This supercomputer will be powered by approximately 64 of Cerebras’ WSE-3 wafer-scale accelerators. Each accelerator delivers 125 petaFLOPS of sparse 16-bit compute, marking a significant technological advantage over traditional GPUs offered by companies like Nvidia and AMD. Notably, Cerebras’ unique approach utilizes on-chip SRAM, providing memory bandwidth that is about 1,000 times faster than Nvidia’s forthcoming Rubin GPUs.
The partnership includes contributions from the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) and India’s Center for Development of Advanced Computing (C-DAC). These collaborations are part of a broader trend of substantial investments in India’s AI infrastructure. Major commitments include Adani’s pledge of $100 billion, Reliance’s $110 billion investment, and US tech powerhouses Amazon, Google, and Microsoft, which have collectively committed roughly $70 billion.
Moreover, OpenAI’s partnership with Tata for boosting compute capacity and similar endeavors by AMD with TCS and Nvidia with Yotta, reflect the growing competition and dynamism within India’s AI sector. By aligning with India’s local governance frameworks and keeping data in-country, this project underscores a commitment to bolstering India’s position in the global AI landscape.

Collections

UAE-based G42 and US chipmaker Cerebras Systems are deploying an 8 ExaFLOPS supercomputer in India, announced at the India AI Impact Summit in New Delhi. Powered by approximately 64 Cerebras WSE-3 wafer-scale accelerators, each delivering 125 petaFLOPS of sparse 16-bit compute, the system claims memory bandwidth roughly 1,000 times faster than Nvidia's upcoming Rubin GPUs. The initiative targets educational institutions, government bodies, and SMEs while complying with Indian data residency regulations. Partners include MBZUAI and India's C-DAC. The project is part of a broader wave of AI infrastructure investment in India, with major commitments from Adani, Reliance, Amazon, Google, Microsoft, and collaborations involving OpenAI-Tata, AMD-TCS, and Nvidia-Yotta.

G42 and Cerebras Collaborate to Launch 8 ExaFLOPS Supercomputer in India