A few years ago, India making its own chips and AI models sounded like a distant dream. In 2026, there are real numbers to show for it. The country now has more than 45,000 GPUs available for AI work, the government has approved ₹1,27,500 crore for Semicon 2.0, and three chip plants are reported to be in commercial production.
But there are big gaps too. Here’s a simple look at where India stands.
AI: computers that anyone can rent
Training AI models needs powerful chips called GPUs, and they are very expensive. Under the IndiaAI Mission, which has a budget of ₹10,371.92 crore, the government has pooled these chips and rents them out cheaply.
- Over 45,000 GPUs were available by June 2026.
- 237 projects had used subsidised computing power by August, adding up to 93.18 lakh GPU hours.
- 20 Indian AI models were picked from 506 proposals: 12 large models that understand text, images and more, and eight smaller language models.
- Researchers, startups, small businesses and students can reportedly use this computing power for about ₹65 an hour, according to the IT Ministry secretary, as reported by DD News.
Curious about who is building these models? We listed seven Indian companies working on their own AI and the AI startups to watch this year.
Chips: the first factories are up
Making semiconductors is much harder than using them. Here, India’s progress is slower but real:
- Semicon 2.0 has been approved with ₹1,27,500 crore. It covers chip design, equipment, new factories, packaging, research and training.
- 12 chip projects have committed more than ₹1.64 lakh crore. Three are already producing, with Micron, Kaynes and CG Semi reported to have started.
- At SEMICON India 2026 in September, PM Modi launched production at two projects, including a CDIL Semiconductor packaging unit in Mohali, Punjab, according to the India Semiconductor Mission.
- About 68,000 students have been trained in chip design across 315 universities, and 105 startups are working on chips.
- The first large chip-making plant under the newer push is expected to be ready in 2028.
What is still missing
A September analysis by Hello Entrepreneurs points out the weak spots. The biggest one: India’s AI push still runs mostly on GPUs bought from abroad. As the writer puts it, “Digital sovereignty, apparently, comes with an import invoice.”
Chip plants also cost huge amounts of money and take years to build. India still lacks depth in the most advanced chip-making, in specialised equipment and in chemicals. And it is competing with the rest of the world for the same engineers.
India is not yet in the same league as Taiwan, South Korea or the US. But instead of betting on one big factory, it is now building a full chain, from design to packaging.
What it means for you
If you are a student, chip design and AI engineering are two of the fastest-growing job areas in India right now. If you run a startup, the cheap GPU access can cut the cost of building AI products. And if you are an investor, remember that chip plants pay off over years, not months.
The same themes came up at this year’s India Mobile Congress, and we looked at an earlier milestone in India’s 5G and AI chip announcement.
Sources: Hello Entrepreneurs, DD News, India Semiconductor Mission and Compsmag.
