The most discussed AI in India is the AI you can talk to. Chatbots, copilots, customer service agents, document assistants. It is the AI that demos well, that fits in a screenshot, that a marketing team can put in a launch video. It is also, in the long run, probably not the AI that matters most to the Indian economy.
There is another AI economy, and it does not make headlines. It runs inside nuclear reactors and aircraft engines. It lands spacecraft on the Moon. It optimises the placement of oil wells and the load on electricity grids. It designs the semiconductors that everything else runs on. It is harder to build, more dangerous to get wrong, and far more capital-intensive than anything in the chatbot economy. And India has been quietly building it.
You would not know this from the technology press. But you would know it from the Cypher stage, which over the past three editions has hosted one of the most comprehensive public records of Indian industrial and deep-tech AI anywhere. Twenty-six named industrial deployments appeared on that stage between 2023 and 2025. Read together, they reveal an adoption arc that the headline AI conversation has almost entirely missed.
Figure 1: The three-phase arc of industrial AI on the Cypher stage — from optimising existing plants to running safety-critical systems.
2023: The optimisation phase
In 2023, Indian industrial AI was about making existing operations better. The deployments described on the Cypher stage were overwhelmingly optimisation problems: how to run a plant more efficiently, extract more from a process, reduce waste in a supply chain.
Shell had the deepest presence. Kartik Ranganath spoke on harnessing AI to accelerate the energy transition. Chiranjib Sur presented the company’s road to scalable AI. A separate session covered efficient well-location optimisation using a Bayesian optimiser — the kind of unglamorous, high-value industrial AI that never makes a launch video but quietly changes the economics of an oil field.
Heavy industry was well represented. Kiran Shetty from Hindalco described the digital evolution of a metals business. Vinay Morje from Grasim covered AI in continuous manufacturing processes. Sreenivas Ramarao from Adani AI Labs spoke about large-scale industrial problem solving. Madhavi Kanumoory from Birla Carbon asked whether AI was a boon or a bane for industry. Prakash Hegde from L&T Construction described AI-powered transformation in one of India’s largest infrastructure businesses.
Mobility and aerospace were present but exploratory. Thameem Kamaldeen from Alstom covered AI-driven rail systems. Latha Chembrakalam from Continental Automotive discussed AI trends in mobility. Himagiri Gedela from Hindustan Aeronautics Limited described elevating aircraft production through strategic planning. And ISRO’s Radha Krishna Kavuluru presented Bhoonidhi, the agency’s earth-observation data platform, as a source of actionable AI insights.
The unifying characteristic of 2023 industrial AI was that it sat alongside existing operations rather than inside the critical loop. AI optimised the plant; it did not run it. AI analysed the satellite data; it did not fly the satellite. The technology was being trusted with analysis, not yet with control.
Figure 2: Industrial domain activity across editions. Nuclear, semiconductors and quantum appear only from 2025 — the newest and most consequential additions.
2024: The autonomy phase
Cypher 2024 marked a shift from optimisation to autonomy. The industrial AI deployments described that year were no longer about doing existing work better — they were about doing work that humans cannot do at all, or cannot do without machine assistance.
Space led the shift. Prateep Basu from SatSure spoke twice — on unlocking satellite data with AI, and on the opportunities and challenges of using AI on satellite imagery for enterprises. A separate session covered autonomous landing on the Moon and Mars using AI, the kind of application where there is no human in the loop because the speed-of-light delay makes real-time human control physically impossible.
Aerospace deepened. A session titled “From Concept to Sky” described transforming aircraft engines with AI — moving from the production-planning focus of 2023 toward AI embedded in the engineering of the machines themselves. Robotics entered the agenda, with Bharadwaj Amrutur presenting on 5G and large language models for the next generation of robotics applications. Manufacturing evolved too, with a session on GenAI reshaping the manufacturing-and-trading company landscape.
The 2024 industrial agenda also surfaced the first serious Indian AI-infrastructure conversation, with Sunil Gupta presenting on powering India’s AI-first ambitions with Shakti Cloud — the GPU and compute substrate that the rest of the industrial AI economy would need. Autonomy in the application layer and sovereignty in the infrastructure layer arrived in the same edition. That was not a coincidence.
2025: The safety-critical phase
Cypher 2025 was the year Indian industrial AI moved into the highest-stakes systems in the economy. The deployments described were not optimisations and not even autonomous applications in the 2024 sense. They were AI inside systems where failure is catastrophic, where regulators are watching, and where trust is the binding constraint.
Nuclear was the most striking arrival. Two sessions — one from a research scientist, Muthukumar Ganesan, and one from Dr Prasanna Kulkarni — addressed safety enhancements in nuclear reactors with AI, framed as a move from risk to resilience. Nuclear reactor safety is about as far from a chatbot as industrial AI gets: it is the domain where the cost of a hallucination is measured in something other than money. Its appearance on the Cypher stage is the single clearest signal that Indian industrial AI has crossed into the safety-critical tier.
Aviation followed the same logic. Rohit Pruthi, Head of AI Solutions at Rolls-Royce, presented on building model trust through AI explainability in aviation — the recognition that in safety-critical domains, a model that works is not enough; it must also be explainable to a regulator. Alstom India’s Puneet Srivastava returned the rail conversation to safety, with AI for smarter and safer railways.
Space matured from data to hardware. Suyash Singh, founder and CEO of GalaxEye, presented on harnessing AI for what the company describes as the world’s first hybrid imaging satellite. ISRO’s Nitish Kumar spoke on innovating in AI for space technology. The Indian space-AI story moved from analysing imagery in 2023 to engineering the satellites themselves by 2025.
Energy and mobility deepened into infrastructure. Dhanya Rajeswaran from Fluence India covered intelligent storage systems for smarter, greener grids. Arvind Gopalakrishnan from SUN Mobility presented AI platforms for EV infrastructure at scale. Mughilan Thiru Ramasamy, CEO of Skylark Drones, described agentic AI for what he called India’s new digital field force — drone fleets operating semi-autonomously across infrastructure and agriculture.
And then the deepest layer of all arrived: the compute substrate itself. Shashwath TR from Mindgrove Technologies presented on designing AI-optimised chips for India’s technology future — indigenous semiconductor design, the foundation on which every other AI application ultimately runs. And quantum appeared for the first time, with Sujoy Chakravarty from Quanfluence on quantum meeting AI, and Professor Arindam Ghosh from IISc on building a quantum-powered society and economy. The 2025 industrial agenda reached all the way down to the physics.
Figure 3: Twenty-six named industrial AI deployments on the Cypher stage, 2023–2025. Border colour indicates edition and phase.
Eight domains, one trajectory
Read across the deployments, India’s industrial AI is spread across eight domains, each with a distinct maturity profile.
Space and satellite — the most consistent. Present every year and deepening: ISRO’s Bhoonidhi in 2023, SatSure and planetary landing in 2024, GalaxEye’s hybrid imaging satellite and ISRO again in 2025. The progression from analysing data to building hardware is the clearest single-domain arc in the dataset.
Energy and grid — the early leader. Shell’s 2023 dominance, evolving into grid-scale storage AI with Fluence by 2025. Energy was where Indian industrial AI first proved its value, because the optimisation gains were large and immediately measurable.
Aerospace and aviation — the steady climb. From HAL’s production planning in 2023, to aircraft-engine AI in 2024, to Rolls-Royce’s explainability work in 2025. The trajectory tracks the move from operations to engineering to trust.
Heavy industry and manufacturing — the foundation. Strongest in 2023 (Hindalco, Grasim, Adani, Birla Carbon, L&T), then quieter on the public stage — not because adoption slowed, but because it became routine. When a technology stops being conference-worthy, it has usually become infrastructure.
Mobility and rail — the safety story. Alstom across multiple editions, Continental in 2023, SUN Mobility and the broader EV ecosystem by 2025. The conversation moved from mobility trends to railway safety to EV-charging infrastructure at scale.
Nuclear and safety-critical — the newest frontier. Absent until 2025, then arriving with two sessions. This is the domain that defines the leading edge of industrial AI trust, and its appearance is the strongest signal of category maturity in the entire three-year arc.
Semiconductors and quantum — the deepest layer. Also new in 2025: Mindgrove on indigenous AI chip design, Quanfluence and IISc on quantum. India reaching for the compute substrate and the next computing paradigm in the same edition is a statement about sovereign ambition.
Drones and robotics — the field force. Robotics in 2024, Skylark’s agentic drone fleets in 2025. The physical-world agent category that will likely expand fastest over the next two editions.
What the arc reveals
Three observations stand out.
One: industrial AI trust moved inward, toward the critical loop. In 2023, AI sat beside operations — analysing, optimising, advising. By 2025, AI was inside the critical systems — reactors, aircraft, grids, satellites. The defining shift of the three-year arc is not that more companies adopted AI; it is that they trusted it with progressively higher-stakes decisions. That inward movement is the real maturity signal, and it is invisible in the chatbot economy.
Two: the deepest layers arrived last and fastest. Nuclear, semiconductors, and quantum — the three hardest domains — all appeared for the first time in 2025, simultaneously. India did not climb the industrial-AI ladder rung by rung over a decade; it reached the top three rungs in a single edition. That compression mirrors what we observed in Indian finance, where the GenAI deployment cycle ran in 18–24 months rather than the 3–5 years that traditional ML took.
Three: sovereignty is the through-line. Shakti Cloud in 2024, Mindgrove’s indigenous chips and the quantum sessions in 2025, GalaxEye’s home-built satellite — the industrial AI story is also an Indian self-reliance story. Unlike the finance arc, where global captives led early, the industrial arc is dominated by Indian institutions, Indian startups, and Indian public-sector science from the start. ISRO, HAL, the nuclear research establishment, IISc. Industrial AI is where India’s sovereign-technology ambition is most visible.
The 2026 question
If the arc holds, Cypher 2026’s industrial agenda should show three things.
First, defence and space will deepen further. The combination of indigenous satellites, autonomous systems, and the national security dimension of AI makes this the domain most likely to expand. Expect named deployments from the space-tech startup ecosystem and, possibly for the first time, explicit defence-AI sessions.
Second, the semiconductor and quantum thread will move from single sessions to a dedicated track. Both appeared for the first time in 2025. Both have enough national-mission momentum behind them — the India Semiconductor Mission, the National Quantum Mission — to justify sustained programming by 2026.
Third, the safety-critical and explainability conversation will become a governance conversation. As AI moves deeper into reactors, aircraft, and grids, the binding question shifts from “can it work” to “can we certify it.” The 2026 sessions that matter most in industrial AI will be the ones about validation, certification, and regulatory trust — the same shift that finance is undergoing, arriving in industry one edition later.
The chatbot economy will keep the headlines. But the industrial AI economy — reactors, satellites, aircraft, chips, grids — is where the harder, more consequential, more uniquely Indian work is happening. The Cypher stage has been documenting it in plain sight for three years. It is, by some distance, the most important AI story in India that almost nobody is telling.
