India's government wants a homegrown graphics processor (GPU) within about 24 months, though it has not said who will build it.
S. Krishnan, secretary of the Ministry of Electronics and Information Technology, set out the target in a November 4 interview. It falls under India Semiconductor Mission 2.0. "This will not happen overnight," he said, putting the work at roughly 24 months. He described five layers India needs for AI, starting with energy and accelerator chips and running up through servers, AI models and applications.
Some lower layers exist. VVDN, Kaynes and Avalon already build servers, and the state-backed Centre for Development of Advanced Computing (C-DAC) designs its Rudra server range, then hands the technology to local manufacturers. The planned Tata Electronics fab in Dholera, Gujarat, is being equipped with help from ASML. It is expected to use 300mm wafers and eventually reach 50,000 wafers per month. Intel has also announced a partnership with Tata covering manufacturing and packaging.
The chips India has actually shipped are a different class. SHAKTI and VEGA are CPUs built on RISC-V, an open-source chip design standard, not GPUs. The VEGA ET301 on the Aries 3.0 board is a 32-bit chip running up to 100MHz with 26KB of SRAM, meant for sensors, wearables and Internet of Things devices. Government showcases on September 26 included a water-cleaning robot, a digital telescope and a forest monitor. A design incentive program has backed 24 companies that went on to raise private money.
That gap is the real story. Going from a sensor-grade microcontroller to a GPU is a leap, and the 2019 IIT Madras effort, which began with a software kit for processors that did not yet exist, shows how slowly the CPU side has moved. A fab also does not design a chip. With no designer or manufacturer named, 24 months reads as an ambition more than a plan.
The motive is easier to read. India's semiconductor market is about $50 billion and is expected to approach $100 billion by 2030, according to the industry association IESA. Krishnan also compared AI to the Y2K boom for India's IT services sector, which employs about 6 million people. A NITI Aayog paper estimates two million jobs could become redundant and 4 million could emerge. Y2K was a services windfall, though, and a GPU is a hardware bet with a very different cost of entry.