India's Centre for Development of Advanced Computing (C-DAC) has moved its indigenous AI inference chip into trial production — the country's first serious homegrown attempt at a processor designed to run trained AI models against real-world data. IT Minister Ashwini Vaishnaw has set a target of a production-grade chip by 2029–2030, backed by the newly approved ₹1,27,500 crore Semicon 2.0 mission. The chip doesn't need to beat Nvidia globally to matter. It needs to give India a domestic alternative for government systems and critical infrastructure that can't stay permanently dependent on a single foreign supplier. That's the strategic frame — and it's closer to reality than most people realize.
TL;DR
- C-DAC's AI inference chip is in trial production; early trials have been "successful," per C-DAC's executive director S.D. Sudarsan. (LiveMint, Jul 17 2026)
- HCL Infosystems selected via GeM tender to validate chip design and performance.
- Production-grade target: 2029–2030, per IT Minister Ashwini Vaishnaw.
- Funded through the National Supercomputing Mission (not Semicon 2.0 directly).
- Part of Semicon 2.0's ₹1,27,500 crore cabinet-approved outlay (Jul 15 2026). (India Today, Jul 15 2026)
- Last verified: 2026-07-21. Chip specs, fabrication node, and manufacturing partner are not yet publicly disclosed — treat as volatile.
What Is an AI Inference Chip — and Why Is India Building One?
An AI inference chip is a processor designed to run an already-trained AI model against new, real-world data — generating responses, analyzing images, powering AI applications at scale. It is the workhorse chip that makes AI usable every day, as opposed to a training chip that builds models from scratch.
Think of it this way: training is building the brain; inference is using it. Nvidia's GPUs dominate both markets today, but inference is where the market is fragmenting fastest. Custom ASICs already capture an estimated 37% of data center inference deployments in 2025, while training remains Nvidia's fortress with over 90% share. (Plain English, 2025)
India's chip targets inference specifically — and that's the smart bet. Inference is more price-sensitive, less dependent on Nvidia's CUDA software moat, and growing faster as more trained models move into production. The global AI accelerator market is projected to exceed $200 billion in 2026, with Nvidia's share settling near 75% as custom silicon and competitors gain ground. (Silicon Analysts, Q1 2026)
C-DAC's chip is akin to Nvidia's GPUs, Google's Tensor Processing Units (TPUs), and Groq's Language Processing Units (LPUs) — a dedicated accelerator for AI workloads. The goal is a homegrown patent design that makes India less vulnerable to US export control regulations, which have previously restricted access to Nvidia's chips. (Communications Today, Jul 17 2026)
Who Is C-DAC and Why Are They Leading This?
The Centre for Development of Advanced Computing (C-DAC) was established in 1988 under the Ministry of Electronics and Information Technology (MeitY) — originally to develop indigenous supercomputers after India was denied access to such technology by the US. An institution built to create India's own supercomputers is now leading India's own AI chip. (LiveMint, Jul 17 2026)
C-DAC's role is expanding beyond this single chip. Amitesh Sinha, Additional Secretary at MeitY and CEO of the India Semiconductor Mission, confirmed that C-DAC will also be the nodal agency for implementing the Design-Linked Incentive (DLI) scheme under Semicon 2.0 — meaning it will validate and identify potential Indian chip designs that can support global workloads. (Communications Today, Jul 17 2026)
This builds on C-DAC's existing semiconductor IP work. In September 2024, C-DAC validated its second-generation Vega RISC-V processor for government and defense electronics procurement under the MeitY IndiaChips initiative. (GMInsights, 2026)
How Is India Testing the Chip?
C-DAC selected HCL Infosystems through a tender on the Government e-Marketplace (GeM) platform to validate the AI chip's design and performance. The testing process has three phases:
- Joint validation: C-DAC and HCL work together to validate the chip's performance baseline.
- Multi-architecture testing: The chip will be tested across multiple systems and computing architectures to ensure compatibility.
- Deployment target: If validation succeeds, the chip will be offered to run on domestic servers and IT infrastructure, powering public services and AI use cases.
The first government official involved told Mint: "Following this, C-DAC will work jointly with HCL to validate the performance of the chip, and the same will then be tested across multiple systems and computing architecture. Our eventual goal is to offer the chip to run on domestic servers and IT infrastructure, and power public services and AI use cases." (LiveMint, Jul 17 2026)
S.D. Sudarsan, Executive Director of C-DAC Bengaluru, confirmed progress: "MeitY has been highly supportive of all the work that C-DAC has pursued around indigenous silicon architecture, and work around the indigenous AI chip is progressing as per the expected timeline. Early trials have proven to be successful, and we're now in process of taking the project to the next level." (LiveMint, Jul 17 2026)
What Is the Timeline for India's AI Chip?
| Milestone | Target | Status | Source |
|---|---|---|---|
| 2nm GPU development begins | June 2025 (reported) | Confirmed | Communications Today, Jun 9 2025 |
| Early prototype / preview | End of 2025 | Reported | Communications Today, Jun 9 2025 |
| Trial production begins | July 2026 | Confirmed | LiveMint, Jul 17 2026 |
| Production-grade chip | 2029–2030 | Target set by IT Minister | LiveMint, Jul 17 2026 |
| Deployment on domestic servers | Post-2030 | Goal stated | LiveMint, Jul 17 2026 |
The earlier June 2025 reporting revealed that C-DAC Bengaluru had been allocated approximately $200 million for a 2-nanometre GPU development project, with a prototype preview targeted for late 2025 and full-scale production by 2029. Officials indicated the Indian GPU could cost "up to 50% less than what Nvidia currently retails its chips at." (Communications Today, Jun 9 2025)
Important caveat: C-DAC has not publicly disclosed the current chip's specifications — not the fabrication process node, nor the manufacturing partner. The trial production phase exists precisely to answer those unresolved questions before India commits to full-scale manufacturing. Treat all specs as unconfirmed until officially announced.
How Does This Fit Into India's Semiconductor Strategy?
India's AI chip effort sits at the intersection of three major policy initiatives:
1. Semicon 2.0 (₹1,27,500 crore)
The Union Cabinet approved Semicon 2.0 on July 15, 2026, with a budgetary outlay of ₹1,27,500 crore (₹1.27 lakh crore / approximately $15 billion). This builds on Semicon 1.0, which had a ₹76,000 crore outlay and approved 12 projects attracting around ₹1.64 lakh crore in combined investment. (India Today, Jul 15 2026; LiveMint, Jul 15 2026; ISM official site)
Semicon 2.0 covers chip design, fabrication, research, materials, equipment, and startups — with a new focus on incentivizing suppliers of raw materials (minerals, gases) used in chip manufacturing. The majority of Semicon 1.0 investment came from Tata Electronics and its semiconductor arm. (LiveMint, Jul 15 2026)
2. National Supercomputing Mission (NSM)
Here's a distinction that matters: C-DAC's AI chip funding comes through the National Supercomputing Mission, not Semicon 2.0 directly. Vaishnaw clarified that while Semicon 2.0 research grants will fund similar projects where India can own chip patents, the government-backed research organization's chip work is backed by the NSM — a mission steered jointly by the Department of Science and Technology (DST) and MeitY, implemented by C-DAC and IISc Bengaluru. (LiveMint, Jul 17 2026; C-DAC NSM project page)
The NSM has already produced indigenous technologies including the Rudra server, Trinetra HPC interconnect, and cooling systems — with 6,000+ Rudra servers under manufacture in Phase III. C-DAC's AIRAWAT AI supercomputer secured the 75th position on the Top 500 Global Supercomputing List in 2023. (C-DAC NSM project page)
3. Design-Linked Incentive (DLI) Scheme
C-DAC will serve as the nodal agency for the DLI scheme, which provides product design support, prototype development funding, and deployment-linked incentives. The scheme offers up to ₹15 crore per application for product design-linked incentives and 6%–4% of net sales turnover over 5 years (up to ₹30 crore) for deployment-linked incentives. (PIB, Jul 31 2024; ISM official site)
This matters because it connects India's broader sovereign AI ambitions — the push to move from the last mile to the first mile of AI development — to actual silicon infrastructure. You can't have sovereign AI without sovereign compute.
AI Inference Chips Compared: Where Would India's Chip Fit?
| Chip | Maker | Type | Target Node | Notes |
|---|---|---|---|---|
| H100 / H200 / Blackwell | Nvidia | GPU (train + inference) | 4nm / 4nm | ~80–90% AI accelerator market share; CUDA software moat |
| TPU v5 / v6 | ASIC (inference + train) | 5nm (est.) | Internal use + Google Cloud; not sold standalone | |
| Trainium 2 | AWS | ASIC (train + inference) | — | Internal use; challenges Nvidia on cost-per-inference |
| MI300X / MI350X | AMD | GPU (train + inference) | 5nm / 3nm | 10–15% merchant share; ROCm software catching up |
| Groq LPU | Groq (acquired by Nvidia) | ASIC (inference) | — | Ultra-low latency inference; acquired by Nvidia in 2025 |
| C-DAC inference chip | C-DAC (India) | ASIC (inference) | Undisclosed | Trial production; 2029–2030 production target |
Sources: Silicon Analysts, Q1 2026; AMD vs Nvidia analysis, Apr 2026; Communications Today, Jul 17 2026
India's chip enters a market that's bifurcating. Training remains Nvidia's fortress (90%+ share). Inference is fragmenting — with custom ASICs capturing an increasing share as mature models move into production. That's the wedge India is targeting.
Why Doesn't India Just Buy From Nvidia?
India currently imports over 90% of its semiconductor needs. (ETManufacturing, Mar 19 2026) The dependency creates three vulnerabilities:
1. Export control risk. In January 2025, the Biden administration released the "AI Diffusion Policy" — a global framework governing exports of frontier AI chips, cloud access, and model weights. The policy placed restrictions on advanced AI chip exports and capped GPU purchases for select nations. (CFR, Jan 13 2025; PCMag, 2025) A former US official noted that this executive order was "a key moment for India to start seriously weighing the idea of building its own chip." (Communications Today, Jun 9 2025)
2. Cost. Nvidia's H100 SXM costs approximately $3,320 to manufacture and sells for around $28,000 — an 88% gross margin. (Silicon Analysts, Q1 2026) Government officials project India's GPU could be priced "up to 50% less than what Nvidia currently retails its chips at." That pricing delta matters enormously for India's digital public infrastructure — which serves over a billion citizens at low per-user economics.
3. Strategic autonomy. As Ajai Chowdhry, chairman of HCL and co-founder of Epic Foundation, told Mint: "A domestic GPU patent based on the government-funded research bodies is imperative, especially seeing that almost all chips today are owned by the US." (Communications Today, Jun 9 2025)
For a deeper look at India's sovereign AI strategy beyond silicon, see our analysis of sovereign AI India's move from the last mile to the first mile.
Can India Actually Fabricate This Chip?
This is the hardest question, and the answer is: not yet at the most advanced nodes.
India does not currently have a domestic fabrication plant capable of manufacturing 2nm chips. The June 2025 reporting confirmed that once C-DAC's chip is developed, "we'll likely be manufacturing it at scale with Taiwan Semiconductor Manufacturing Corp (TSMC)." (Communications Today, Jun 9 2025)
But India's fab ecosystem is building. The Deloitte TMT Predictions 2026 report projects India will host 4–5 silicon fabs, 8–10 compound fabs, 1–2 display fabs, and 20–25 OSAT (Outsourced Semiconductor Assembly and Test) facilities by 2035 — supported by ISM and state-level incentives. By 2035, 60% of India's domestic semiconductor demand is expected to be met through local production. (ETManufacturing, Mar 19 2026)
India's semiconductor market is estimated at $45–50 billion in FY2024-25, growing at a 20% CAGR over the past three years. It's projected to reach $120 billion by 2030 and $300 billion by 2035 — driven by AI, automotive growth, and data center expansion. (Business Standard, Mar 18 2026; The Hindu, 2026)
The pieces — from funding to an actual chip on a test bench — are starting to come together. For context on how this fits into the broader compute arms race, see our coverage of the engineer building OpenAI's compute empire and the AI memory shortage that's become a geopolitical issue.
What About Other Indian Chip Breakthroughs?
C-DAC's effort isn't happening in isolation. India's semiconductor ecosystem is seeing parallel breakthroughs:
- IIT Bhubaneswar's spintronic chip — a fundamentally different approach to AI hardware that could reduce power consumption dramatically. See our deep dive: how IIT Bhubaneswar's spintronic chip could redefine AI hardware.
- C-DAC's Vega RISC-V processor — second-generation validated for government and defense procurement in September 2024. (GMInsights, 2026)
- India's electronics component manufacturing push — the 0% duty strategy to own the component layer. See: India's 0% duty strategy to own the electronics component layer.
These efforts are complementary. RISC-V provides the open-architecture foundation, spintronics explores beyond-silicon physics, and C-DAC's inference chip targets the most immediate commercial gap.
What This Means for You
For Indian startups and AI builders: A domestic inference chip — if it hits its 2029–2030 target and cost projections — could dramatically lower the cost of running AI models in production. If the 50% cost reduction holds, inference economics for India-built AI applications change fundamentally. Monitor C-DAC's AIRAWAT platform, which already offers AI computing access to academia and startups.
For government and public-sector technologists: The strategic value isn't about beating Nvidia on benchmarks. It's about having a domestic alternative for digital public infrastructure (DPI), government IT systems, and critical workloads that can't stay permanently dependent on one foreign supplier subject to export controls. This is infrastructure sovereignty, not a performance contest.
For investors and industry watchers: The funding chain is now clear — Semicon 2.0 (₹1,27,500 crore) + National Supercomputing Mission + DLI scheme — and an actual chip is on a test bench. But the gap between trial production and a production-grade chip running in a data center is still 3–4 years, minimum. Spec details remain undisclosed. The realistic timeline is late 2029 at the earliest.
For anyone following sovereign AI globally: India is one of the few countries attempting full-stack AI sovereignty — from sovereign AI models to sovereign silicon. The inference chip is the hardware layer of that stack. Whether it succeeds on schedule will determine whether India remains a customer in the AI hardware market or becomes a participant.
FAQ
Q: What is C-DAC's AI inference chip?
A: It's India's first indigenous AI inference processor — designed by the Centre for Development of Advanced Computing (C-DAC) to run trained AI models against real-world data. It's currently in trial production, with a production-grade target of 2029–2030. Unlike training chips (which build models from scratch), inference chips are the workhorse processors that power everyday AI applications.
Q: When will India's AI chip be ready?
A: IT Minister Ashwini Vaishnaw has set a target of 2029–2030 for a production-grade chip. Early trials have been reported as "successful" as of July 2026, but chip specifications, fabrication node, and manufacturing partner remain undisclosed. The timeline from trial production to deployment is typically 3–4 years.
Q: Will India's AI chip compete with Nvidia?
A: Not directly. India's chip targets the domestic market — government systems, public infrastructure, and Indian AI workloads — rather than competing with Nvidia globally. The strategic goal is reducing dependence on a single foreign supplier, not displacing Nvidia from the global AI accelerator market where it holds approximately 75–80% share.
Q: How much will India's AI chip cost?
A: Government officials have projected the chip could cost "up to 50% less than what Nvidia currently retails its chips at." However, this is a projection from June 2025 reporting, not a confirmed price. Final pricing depends on fabrication costs, manufacturing partner, and volume — none of which are finalized. Treat as unverified until official pricing is announced.
Q: Where will India's AI chip be manufactured?
A: India does not currently have domestic fabrication capability for advanced nodes (2nm class). Officials indicated in June 2025 that the chip would likely be manufactured at scale with TSMC (Taiwan Semiconductor Manufacturing Corporation). India's fab ecosystem is expanding under Semicon 2.0, but it will take years to reach advanced-node manufacturing capability domestically.
Q: How is C-DAC's AI chip funded?
A: The chip development is funded through the National Supercomputing Mission (NSM), not Semicon 2.0 directly. However, Semicon 2.0's research grants will fund similar projects where India can own chip patents. C-DAC also received approximately $200 million for a broader 2nm GPU development effort reported in June 2025.
Discussion
0 comments