IBM Deepens IIT Bombay and IISc Partnerships to Build Indic AI, AI Agents and Quantum-HPC Workflows
IBM has expanded its research collaborations with IIT Bombay and IISc, focusing on sovereign and Indic language models, agentic AI, Granite-based energy forecasting models and quantum-centric supercomputing.
IBM has widened two of its oldest academic partnerships in India. In an announcement dated 1 October 2026, the company said it is expanding its research collaborations with the Indian Institute of Technology Bombay (IIT Bombay) and the Indian Institute of Science (IISc), Bengaluru, to push work on agentic AI, sovereign AI, quantum computing and next-generation computing systems.
The two programmes are not new — IBM has worked with IIT Bombay since 2018 and with IISc since 2021 — but the next phase sets out a sharper agenda that maps closely onto India's national priorities: AI that works in Indian languages, AI that can run on infrastructure India controls, and the early plumbing for quantum computers to work alongside classical supercomputers.
IIT Bombay: Indic models and multimodal AI
At IIT Bombay, the expanded collaboration will focus on:
- Sovereign and Indic language model adaptation — adapting foundation models so they perform well in Indian languages and can be deployed under Indian control;
- Multimodal AI that reasons across text, images and other data types;
- AI infrastructure and knowledge retrieval, the systems that let enterprise AI find and use the right information; and
- multimodal systems for software programming education, human-AI collaboration and intelligent operations across hybrid cloud environments.
Ganesh Ramakrishnan of IIT Bombay described work across languages, modalities and domains in the company announcement.
IISc: agents, energy models and quantum-HPC
The IISc track is more infrastructure-heavy. Researchers will work on:
- Agentic AI workflows that orchestrate applications across hybrid cloud environments while balancing performance, cost and operational complexity;
- Lightweight time-series foundation models built on IBM's open Granite model family, aimed at energy analytics — forecasting, anomaly detection, load disaggregation and energy optimisation; and
- Quantum and high-performance computing (HPC) orchestration, including quantum-centric supercomputing workflows and next-generation quantum-HPC algorithms.
The energy piece is worth noting. India's power grid is absorbing large volumes of variable solar and wind generation, and utilities need better short-term forecasting and anomaly detection to keep it stable. Small, efficient time-series models that can run close to the data are a practical way to bring AI into that problem without massive compute bills.
Why "quantum-centric supercomputing" matters
Quantum computers will not replace classical machines; for the foreseeable future they will act as specialised accelerators inside larger workflows. The hard engineering problem is orchestration — deciding which part of a scientific calculation runs on a quantum processor, which runs on GPUs or CPUs, and how results move between them. IBM calls this approach quantum-centric supercomputing.
India has a direct stake in solving it. Under the National Quantum Mission, the country is investing in its own quantum hardware, and IBM has said Amaravati in Andhra Pradesh will host one of the first physical IBM quantum computers in India. Research groups that understand how to integrate such machines with national HPC facilities will be well placed to turn qubits into scientific results.
IBM Research India director Amith Singhee said the partnerships combine academic research with IBM’s scientific expertise.
What it means for India
Industry-academia partnerships of this kind do three things for the ecosystem. They give Indian researchers access to production-grade models and systems; they train students on problems at the frontier rather than on textbook examples; and they can support research that domestic startups may build on if outputs are released. Sovereign and Indic AI, in particular, is an area where India cannot rely on models tuned for English-speaking markets.
The open question, as with any such announcement, is output: published papers, open-source models and tools, and people trained. Those are the measures by which this next phase should be judged.
Sources
Image: IBM. Official eight-bar wordmark from the company newsroom, used to identify the research partner; not an image of a newly demonstrated AI or quantum system. Original source.