C-DAC Puts Its Homegrown AI Inference Chip Into Trial Production, With HCL Infosystems to Validate It

India's semiconductor ambition has spent most of 2026 living in cabinet notes and outlay figures. This month it took a quieter but more concrete step: the Centre for Development of Advanced Computing (C-DAC), the government's flagship computing lab under the Ministry of Electronics and Information Technology (MeitY), has moved its indigenous artificial-intelligence inference chip into trial production — and has picked HCL Infosystems to validate the design and put it through its paces.

It is a small line item next to the ₹1.27 lakh crore that the Union Cabinet has committed to the second phase of the India Semiconductor Mission, or Semicon 2.0. But it matters, because it is one of the first times an Indian public-sector effort has taken a home-designed AI processor out of the lab and toward something a customer could actually deploy.

What an inference chip actually does

Modern AI runs in two phases. Training is the compute-hungry process of building a model in the first place — the work that has made Nvidia's data-centre GPUs the most sought-after silicon on the planet. Inference is what happens afterwards: running that finished model against real-world data to answer a query, transcribe a call, flag a defect on a production line or route a citizen's request through a government portal.

Inference is where most of the day-to-day cost of AI ends up once a system is in production, and it is a market where a purpose-built, power-efficient chip can compete without matching the raw scale of a frontier training GPU. That is the opening C-DAC is aiming at. The chip is designed to execute trained AI models on live data across public services, domestic servers and national IT infrastructure — precisely the kind of workloads where India would rather not depend on imported accelerators and foreign cloud capacity.

Why HCL Infosystems, and why a tender

Rather than validate the part entirely in-house, C-DAC floated a tender on the Government e-Marketplace (GeM) and selected HCL Infosystems to oversee validation and performance testing. Under the arrangement, HCL Infosystems will check the chip's design and manage its performance through the trial-production phase, with the part then tested across multiple systems and computing architectures.

The choice reflects how the effort is meant to scale. A research lab can tape out a chip; turning it into something that boots reliably across real servers, drivers and software stacks is an industrialisation problem, and that is the gap the HCL Infosystems partnership is meant to close.

The 'Nvidia-class' framing — and a realistic clock

Trade coverage has framed the chip as India's eventual answer to Nvidia-class processors, which is ambitious shorthand for a part still in trial production. The government's own timeline is more measured: Union IT minister Ashwini Vaishnaw has said a production-grade indigenous AI chip is expected around 2029 or 2030, a horizon set when the Cabinet cleared Semicon 2.0. Trial production is best read as a decisive early milestone on that road, not the arrival of a shippable Nvidia rival.

That patience is warranted. Building a competitive AI accelerator is as much about the surrounding software — compilers, runtimes, model support — as it is about the silicon, and that ecosystem takes years to mature.

Building on an existing compute stack

C-DAC is not starting from nothing. It is the institution behind India's PARAM supercomputers under the National Supercomputing Mission, and it has been assembling an indigenous processor portfolio for years — including RISC-V based designs and, more recently, its Rudra server line and the ARM-based AUM high-performance CPU. An AI inference part slots into that stack as the piece aimed squarely at machine-learning workloads rather than general-purpose or HPC computing.

What to watch

The meaningful questions now are practical ones the trial is meant to answer: how the chip performs on real inference workloads, how efficiently it does so, and whether the software around it is mature enough for government departments and domestic server makers to actually adopt it. If the HCL Infosystems validation goes well, the more important follow-on will be the first deployments — the point at which an Indian-designed AI chip stops being a milestone and starts being infrastructure.

Sources

• C-DAC leads India's push to build an Nvidia-class AI chip — Communications Today
• C-DAC's Indigenous AI Inference Chip Enters Trial Run In Partnership With HCL Infosystems — Indian Defence News