Will run AI queries on Indian data centres: Microsoft

Microsoft said it would allow customers of its productivity suite in India to have AI processing conducted within the country itself, starting by the end of 2025.

The firm said “in-country data processing for customers’ Microsoft 365 Copilot interactions [would be] available in 15 countries around the world,” with India, Australia, Japan and the United Kingdom getting the option by the end of the year.

The announcement is a tonal shift from large tech firms’ resistance over the last few years against proposed data localisation mandates, such as India’s in 2018. Since then, Microsoft and other firms have dramatically increased their data centre footprint around the world, with India seeing a surge of investments, and as regulated industries — such as finance — have been required to keep data within Indian borders, data residency has turned from cumbersome requirement to a sweetener for cloud and data centre providers.

Microsoft emphasised that regulations are at the heart of why it is making this offer, even though local data processing may have some marginal advantages in terms of “latency,” the time data takes to travel from one place to another. “This offer is designed to enable customers, particularly those in government and highly regulated industries, to access Microsoft 365 Copilot with an additional option for governance, security, and regulatory compliance,” the firm said.

Google, Microsoft and Meta have data centres in India, where a lot of their customers’ content and data is already stored. However, it is less common for these datacentres to be equipped with the relatively power-hungry graphics processing units (GPUs) on which AI models are deployed. ChatGPT maker OpenAI and Google have been working on setting up such datacentres in India.