India's electronics ministry gave C-Dac a budget of 200 million dollars across five fiscal years for a domestic AI chip that does not depend on Nvidia. The budget runs from FY26 to FY30; FY26 and FY27 allocations are 275 crore and 280 crore rupees. The chip may not become a commercial product; meant for local data centres running public services. At SEMICON Taiwan 2026, TSMC's April Li said inference drives system growth, with utilisation below 40 per cent and data movement at up to 60 per cent of activity.
India's Ministry of Electronics and Information Technology has given the Centre for Development of Advanced Computing a target budget of 200 million dollars across five fiscal years for a domestic AI chip meant to give the country an option that does not depend on Nvidia. The project details are attributed to four senior officials. The budget covers the period from FY26 to FY30.[1]
The allocation to C-Dac for FY26 and the ongoing FY27 is given as 275 crore and 280 crore rupees. The project gained pace after the US decided in January 2025 to regulate global access to Nvidia's high-performance computing chips, a measure that placed India among the countries granted limited access. Officials say C-Dac was seen as the right fit because of the Microprocessor Development Programme it has run since 2017 and the experience it built with the Param supercomputer family. The chip may not become a commercial product; the intended use is academia, research workloads and local data centres running public services.[1]
At the IC forum of SEMICON Taiwan 2026, April Li, TSMC's AI and high-performance computing business development director, said those who build the most integrated systems will lead, and described inference as having stopped being a low-overhead task and as the main driver of system growth. The figures Li presented put the 2022 global inference token volume as having risen nearly 500 times, data movement at up to 60 per cent of typical system activity, and AI accelerator utilisation below 40 per cent. Li added that by 2030 AI packages could pass the 1 trillion transistor mark. The intended home for the C-Dac chip is local data centres running public services, the same inference-heavy class of workload Li described.[2], [1]
