Yotta, a prominent data center operator, is gearing up to enhance its computing capabilities significantly, with plans to boost its graphic processor units (GPUs) to a staggering 32,000 units from Nvidia by March 2025. This ambitious move follows the company's recent orders for 16,000 high-powered GPUs from the renowned US-based chip giant, Nvidia. The CEO of Yotta Data Services, Sunil Gupta, emphasized the critical role of Nvidia's cutting-edge chips in powering the training of large language models and advancing artificial intelligence (AI) technologies, which have sparked fervent interest across industries worldwide.

Gupta revealed that Yotta has secured a "soft commitment" from Nvidia, with the chips slated for delivery by 2025. The company has already committed approximately $1 billion to procure 16,000 H100 GPUs and develop the requisite infrastructure, including servers, networks, storage, software, and additional layers. While Gupta refrained from disclosing the financial specifics of acquiring the additional chips, he underscored that the capital requirements for the project would be supported by the company's promoters, possibly supplemented by financing from banks.

Yotta's inaugural GPU-based data center, set to go live by May 15, 2024, is poised to offer substantial computing capacity for AI processing, driven by an initial batch of 4,096 GPUs. Gupta outlined the burgeoning market for high-performance computing capabilities in India, noting the early stages of sector development in the country. With the first slot of over 4,000 GPUs already reserved by enterprises, Yotta aims to not only serve the domestic market but also cater to global demand. Gupta emphasized the company's competitive pricing strategy, aiming to undercut prevailing rates in the market and position Yotta as a viable alternative to established hyperscalers like Amazon and Meta.