Zoho's Sridhar Vembu has entered the discussion surrounding AI's energy consumption, advocating for a "fundamental rethink" to ensure its viability for India. With data centers in the US causing electricity bills to surge, Vembu believes vastly more energy-efficient AI is needed, which fundamentally rethinks the computational aspect, to make the technology viable in India.
Vembu's comments come at a time of increasing concern about the energy demands of AI, particularly as India seeks to expand its digital infrastructure. He warned on X (formerly Twitter) that AI could overload India's power grid, potentially leading to rolling blackouts. He pointed to a 60% increase in electricity costs since 2023 in Athens, Georgia, blaming the spike on AI-driven data centers, stating, "Even if we could afford all the GPUs (not!), we cannot afford the electricity bill...We cannot afford to hurt households and factories".
Analysts agree that India's ability to sustainably scale AI hinges on breakthroughs in hardware efficiency and energy-smart AI architectures. India's data center capacity is projected to increase significantly, from 1.2 GW in 2024 to a potential 5 GW by 2030. By that time, AI workloads could consume 40–50 terawatt-hours annually, placing a significant strain on a power system already struggling to meet rising industrial and residential demand. Experts caution that any shortfall in renewable energy targets could create regional deficits, especially in AI-heavy zones, even as India ramps up its total generation capacity to 777 GW. If clean energy growth lags, the nation could face a 6 TWh annual gap, increasing the risk of blackouts during heatwaves or periods of high AI activity. Local grids may lack the capacity to handle the concentrated demand of data centers, raising the possibility of regional outages affecting homes and factories.
Vembu has been a vocal proponent of energy efficiency. He has emphasized the need for optimization, especially in software deployed across data centers, not just limited to AI, and even regular SaaS applications. According to Vembu, Zoho is actively working to enhance the efficiency of its technology stack, focusing on both compute resources and energy consumption.
Vembu's concerns align with broader discussions on AI's environmental impact and sustainability. Microsoft is developing a new AI chip cooling system, which it says is three times better than current systems, which could improve energy efficiency and lower costs in data centers. The company is exploring how to integrate the technology into future chip generations and data center operations and plans to continue collaborating with silicon manufacturers to scale the system. Micron has also announced a new generation of modular laptop memory, the LPCAMM2, which consumes 85% less power than DDR5 across CPU, NPU and integrated GPU hardware.
Zoho recently launched Arattai, a homegrown messaging app built on deep R&D. Arattai offers meetings, calling facility, Pocket storage, mentions, no ads, and no forced AI, built for privacy and simplicity.
While acknowledging AI's potential to accelerate learning, streamline workflows, and improve product experiences, Vembu has cautioned against over-reliance or misuse of AI, as it can reverse expected productivity gains. He also doesn't support using AI tools like ChatGPT or Gemini to create new content and has warned against copy-pasting AI-generated text as human-written. He sees AI as a valuable debating partner and uses it to critique different models to deepen his understanding.
