G42 and Cerebras Deploy 8-Exaflop Supercomputer in India as Sovereign AI Infrastructure Race Accelerates
Abu Dhabi-based tech company G42 and U.S. chipmaker Cerebras announced Thursday they will deploy 8 exaflops of computing power through a new supercomputer system in India, marking one of the largest sovereign AI infrastructure projects in the developing world. The announcement came on the sidelines of the India AI Impact Summit in New Delhi.
The move reflects a broader scramble among nations to secure domestic AI computing capacity as concerns grow about data sovereignty and technological independence. For CFOs at multinational firms operating in India, the project signals both opportunity—access to subsidized computing resources—and complexity, as data residency requirements increasingly dictate where AI workloads can run.
The system will be hosted entirely within India and comply with local data residency, security, and compliance regulations, the companies said. G42 India CEO Manu Jain framed the project as essential infrastructure for national competitiveness. "This project brings that capability to India at a national scale, enabling local researchers, innovators, and enterprises to become AI-native while maintaining full data sovereignty and security," Jain said in a statement.
The supercomputer will provide computing resources specifically for educational institutions, government entities, and small and medium enterprises—a departure from typical cloud infrastructure that prioritizes large commercial customers. Abu Dhabi's Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) and India's Centre for Development of Advanced Computing (C-DAC) are also participating in the project.
Andy Hock, chief strategy officer at Cerebras, positioned the deployment as accelerating India's capacity to train large-scale AI models tailored to local needs. "It will accelerate training and inference for large-scale models, enabling researchers and developers to build AI tailored to India's needs," Hock said.
The partnership builds on existing AI collaboration between the UAE and India. Last year, MBZUAI and G42 released Nanda 87B, a Hindi-English large language model built on Meta's Llama 3.1 70B model, designed to understand casual speech in both languages.
The announcement came amid a flurry of AI infrastructure commitments at this week's summit. Indian conglomerate Adani pledged $100 billion to build up to 5 gigawatts of data-center capacity in the country by 2035. Reliance separately committed $110 billion over the next seven years for gigawatt-scale data centers.
For finance leaders, the proliferation of sovereign AI infrastructure creates both planning challenges and potential cost advantages. Companies with Indian operations may gain access to subsidized computing resources, but will need to navigate increasingly strict data localization requirements that could complicate global AI strategies. The question isn't whether to use local infrastructure—it's how to architect AI systems that can comply with fragmented sovereignty requirements across multiple jurisdictions without duplicating costs.
The race to build national AI computing capacity shows no signs of slowing, with implications for how CFOs budget for AI initiatives in markets where governments are actively subsidizing—and regulating—the technology stack.
Responses (0 )