India's Tech Manufacturing Push Meets Its Hard Limits

IBM said on 2 October that it is widening two long-running research collaborations in India, one with IIT Bombay and one with the Indian Institute of Science, according to HPCwire.

The stated focus areas are agentic AI, sovereign AI, quantum computing and next-generation computing systems. The work is not new. IBM's link with IIT Bombay dates to 2018 and its link with IISc to 2021. What is new is the scope: sovereign and Indic language model adaptation, multimodal AI, AI infrastructure and knowledge retrieval, agentic workflows across hybrid cloud environments, lightweight time-series foundation models built on the IBM Granite family for energy analytics, and orchestration for quantum and high-performance computing. Dr. Amith Singhee, director of IBM Research India and CTO for IBM India and South Asia, said the next wave of computing will be shaped by advances in agentic AI, sovereign AI and quantum computing, according to the same HPCwire report. The company did not disclose financial terms.

That silence matters.

The gap between research announcements and factory floors is the central problem in Indian technology manufacturing policy right now, and the numbers on the ground are harder to move than the slide decks suggest. Rahul Bajoria, head of India and ASEAN economic research at Bank of America, told The Diplomat on 29 September that India's manufacturing sector has remained stuck at around 15 percent of GDP despite decades of rapid growth. He pointed to the colonial and planning-era roots of a weak industrial base, regulatory and compliance burdens that keep Indian firms small and push them into informality, and a pattern where regulatory progress has come from a handful of competitive states rather than the centre. Bajoria also flagged limited gains from China Plus One, growing Indian exports to China, a mixed record for production-linked incentives and the scaling up of the Apple ecosystem. The reforms he says are needed to lift manufacturing to 25 percent of GDP by 2047 are in electricity, labour and land.

Manufacturing stuck at 15 percent

That is a long list, and it lands against a backdrop of persistent supply chain disruption. Light Reading reported on 30 September that Omdia expects hardware delays to affect 60 percent of PC channel partners, with component constraints and geopolitical pressure keeping volatility high into 2027. Omdia also said 59 percent of organisations expect their AI budgets to rise by 10 percent or more next year, which raises the stakes on whether that spending produces returns.

Robotics is one place where the demand signal is real.

The Robot Report noted on 29 September that industrial robot installations doubled in the decade to 2024, according to the International Federation of Robotics, and that robot order value rose 6.6 percent year over year in the first half of 2026, according to the Association for Advancing Automation. Growth in non-automotive sectors offset a slowdown among automotive OEMs. India's own development finance record offers a partial counterpoint to the gloom. The Diplomat reported on 1 October that the New Development Bank had approved about $44 billion for 141 projects by the end of June 2026, with roughly $25 billion already paid out. India has used 32 of those approvals since 2016, worth close to $9.5 billion, in transport, social development and clean energy. Recent approvals include $300 million for a clean energy project involving REC Limited and an affordable housing project backed by HUDCO.

Capital and capability

On the defence side, procurement remains slow. The Diplomat reported on 30 September that India signed a letter of offer and acceptance for the Javelin anti-tank system in August, 16 years after the Army began looking for a next-generation fire-and-forget missile. A November 2025 Defense Security Cooperation Agency notification put the purchase at 100 Javelin rounds and 25 launch units at an estimated $45.7 million, against an overall requirement of more than 40,000 missiles.

Policy attention is not the constraint.

The Diplomat reported on 2 October that Prime Minister Narendra Modi spoke to US President Donald Trump on 28 September and described the call as productive, covering trade, defence, energy and critical technologies. The same piece noted that Trump signed the Lindsey O. Graham Sanctioning Russia and Iran Act of 2026 ten days earlier, giving Washington authority to impose tariffs of up to 100 percent on countries still buying Russian oil and gas, with India among the largest buyers.

None of this is settled by a research memorandum. arXiv published a paper on 30 September describing Sapien, a policy engine for stateful contextual policies in autonomous AI agents, which the authors say rules out 93 to 95 percent of attacks on AgentDojo and 62 to 85 percent on Toolathlon when an agent is fully hijacked. That is the kind of work IBM's India partnerships are meant to produce more of. Turning it into factories, grids and export earnings is a different problem, and Bajoria's 15 percent is the number that will not move on its own.

The near-term test is whether the AI spending Omdia describes converts into Indian capacity, or whether India keeps importing the hardware and exporting the talent. Policy papers do not answer that. Capital expenditure does.