Microsoft's GitHub has 13 mn Indian developers, to overtake US by 2027

Nadella said he’s “very excited” about Indian languages being integrated in GitHub co-pilot.

Addressing a house full of developers enthused to hear the Indian-origin top-deck leader, Nadella told developers it is now imperative to build their own AI applications.

“To do that… what we are doing is taking effectively the stack we have used underneath GitHub copilot (and) Microsoft 365 copilot underneath ChatGPT, and making it available to all of you, every layer of it, starting right at the infrastructure,” he said.

He elaborated that the tech giant has three operational data centres in India, with the fourth one being work-in-progress. “We'll keep expanding so that you have the compute resources you need,” Nadella told Microsoft’s developer community.

Microsoft is building low latency AI accelerators for training large language models. “So this is a great time to be in that combination of a hardware/software engineer, where you're thinking about system architecture,” Nadella said.

Developers at Microsoft are told to be “super-ambitious" as whatever they build today with the best models, will become 1/10th or 1/100th cheaper in about 18 months or so, he further said.

Nadella assured developers that generative AI or GitHub are not the last of model architecture breakthroughs. He said he sees a lot of “great stuff” happening in the small language model space too.

Small language models are compressed versions of AI systems to generate human-like text, usually with fewer parameters and computations when compared to large models.