Hyperscale alone won't work for India: HP's Ipsita Dasgupta backs LLM–SLM hybrid strategy

At the AI Impact Summit 2026, Ipsita Dasgupta, Senior Vice President and Managing Director of HP for India, Bangladesh, and Sri Lanka, emphasized the need for India to adopt a hybrid strategy for artificial intelligence. She argued that a combination of hyperscale capabilities and smaller, distributed models is essential for the country’s AI progress. Dasgupta explained the importance of balancing LLMs (large language models), which demand extensive data centers and substantial computing resources, with SLMs (small language models) that can operate efficiently on local devices or edge infrastructures. She highlighted the staggering costs associated with building large-scale AI capacity, noting that a single gigawatt of data center power could require an investment of $50 to $60 billion, enough to supply energy to up to one million homes. Recognizing that not every AI application in India necessitates such vast infrastructure, she urged a focus on practical applications. Examples she provided included AI services designed for farmers, citizen engagement, vernacular support, and educational tools aimed at the 200 million students in India. Dasgupta also reflected on India’s distinctive development trajectory, referencing the Union Budget unveiled on February 1, which addressed both advanced technology and the livelihoods of millions in traditional sectors like coconut farming. She asserted that India has a unique capacity to adapt technology to its specific needs, often outperforming other nations, as exemplified by its digital payment innovations. In her closing remarks, she expressed confidence that AI will replicate this transformative impact, advocating for a dual approach that embraces the ambition of scaling up while also recognizing the value of small, localized models.