AI pushed as lean lever

India’s push to lift manufacturing from roughly 16%-17% of gross domestic product to 25% is increasingly being framed as a productivity problem, not just a capacity problem, with investors and industry executives pointing to artificial intelligence on factory floors as the lever. (economictimes.indiatimes.com) (pib.gov.in) Anirudh Garg said in a recent post that moving India’s manufacturing share from 16% to 25% of GDP “won’t be driven by capacity alone” and argued that AI is already being used to optimize costs, improve yields and predict failures. (sotwe.com) That argument lines up with a broader industry discussion in India this month. The Economic Times reported on April 13 that PwC India’s Vinod Kumar and JSW Group’s Srihari Kaninghat described AI as a tool for cutting material costs, improving uptime and pushing decisions closer to the shop floor. (economictimes.indiatimes.com) The backdrop is a target India has talked about for years. A Press Information Bureau release from March 25, 2025 said manufacturing contributed 17% of India’s GDP at the time, while industry groups reiterated the long-running 25% goal. (pib.gov.in) Outside official messaging, the data are less flattering. The World Bank’s latest series lists India’s manufacturing value added at 13% of GDP in 2024, showing how far the country remains from the 25% benchmark even after years of “Make in India” rhetoric. (worldbank.org) That gap helps explain why executives keep stressing output per worker, machine and kilowatt instead of only new plants. Business Standard reported in March that a PwC India-Observer Research Foundation study estimated AI could create $135 billion to $150 billion in value for India’s manufacturing micro, small and medium enterprises by 2035. (business-standard.com) The sales pitch is straightforward: use software to catch defects earlier, tune processes faster and spot machine trouble before breakdowns halt a line. Mint reported on April 14 that large Indian manufacturers are now talking about “agentic” AI not just for automation, but for planning, logistics and day-to-day operating decisions. (livemint.com) But the bottleneck is not only software. Garg said companies also have to change culture to attract the engineers and operators who can deploy these systems beyond pilots, a point echoed in recent coverage that ties AI adoption to reskilling and day-to-day integration rather than one-off demos. (sotwe.com) (livemint.com) A parallel warning is emerging in semiconductors, where lab tests and real-world use are drifting apart. Semiconductor Engineering reported that customers increasingly want qualification testing to use the same test content and data ports used in production or field operation, because latent damage and silent errors can surface months after a chip passes standard checks. (semiengineering.com 1) (semiengineering.com 2) That is why engineers are pushing harder on sustained workload simulation — long, realistic stress runs that mimic how hardware actually behaves after deployment, not just in a controlled lab window. Semiconductor Engineering has described simulation as a way to close the gap between optimization and manufacturability as AI systems, advanced packaging and changing workloads make old qualification methods less representative. (semiengineering.com 1) (semiengineering.com 2) The thread tying these debates together is that India’s manufacturing ambition now rests on whether factories can get more output from existing lines while proving that AI-driven systems will hold up under real operating conditions. (economictimes.indiatimes.com) (semiengineering.com)