The UAE led global AI adoption in early 2026, with 70.1% of working-age adults using AI tools as per Microsoft’s Global AI Diffusion Q1 2026 Report. AI adoption in the Global North is accelerating at twice the pace of the Global South accentuating a widening technological divide. One which is driven by systemic challenges around access to frontier models, electricity, internet connectivity, and digital skills. Despite growing AI awareness through the India AI Impact Summit, India ranks 64th globally with a 17.6% adoption rate, reinforcing the Brookings report’s key message to view AI diffusion as a deliberate policy challenge instead of leaving it to market forces.
(Sign up for THEdge, The Hindu’s weekly education newsletter.)
India has AI users. It needs AI builders
India’s training ecosystem produces people who use AI, not people who build it. More than 90% of Indian employees already work with generative AI tools, yet a Quess Corp study found an 82.9% shortage in the deeper GenAI skills needed to take models into production. That gap is well paid: Scaler’s India AI Workforce Report 2026, independently assessed by B2K Analytics, found median salaries for its programme graduates doubled to ₹20 lakh, a 104% rise, with the top quartile clearing ₹45 lakh.
Demand is moving the same way. The fastest-growing AI role is a Forward Deployed Engineer, who connects models to real business workflows, tests whether they perform, and keeps them running in production. Global postings for these engineers, with salaries ranging from 150-500K USD, rose nearly 800% over 2025, but India’s supply, at roughly 250 open roles, stays tiny. India adds AI users quickly and builds far too few of the engineers who can put those models to work.
Restricted access to deployment infrastructure
Deployment depends on access, and access is tightening. U.S. export controls on advanced chips and restrictions on frontier-model access mean a workforce trained only to call imported models through an API holds a capability that can be revoked from abroad. India’s options narrow to two, with both demanding engineering depth: build and maintain models domestically—Sarvam’s $234 million raise signals appetite—or deploy open-weight models locally, which increasingly means Chinese stacks such as Qwen and DeepSeek, with the path dependency that follows. Tool literacy buys neither resilience nor sovereignty.
The university as India’s diffusion engine
This is where academia, not the private bootcamp, must lead. Universities are the only institutions that can produce deployment-capable graduates in volume rather than a handful of Ph.D.s, and their role in diffusion runs well beyond publishing research. Between 2021 and 2025, Chinese universities revoked or suspended about 12,200-degree programmes and introduced roughly 10,200 new ones — many of them in AI, robotics, semiconductors, and advanced manufacturing — together affecting more than 30% of the country’s degree offerings.
UAE universities have been central to its lead in AI adoption through aligning their programmes with applied AI, building dedicated institutions such as the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), and tying academic training to the national AI strategy. The broader lesson is that universities matter not only as research centres but as institutions that set the speed and quality of a country’s AI adoption.
Four functions matter most within this context.
Curriculum: prune outdated programmes on industry cycles and rebuild degrees around production practice; data engineering, model fine-tuning, evaluation and MLOps, not theory alone. Embedded learning: borrow the dual-study and co-op model so every degree includes supervised work shipping real systems, through industry-linked capstones and apprenticeships. The campus as a living lab: a university that deploys AI across its own teaching hospitals, libraries and administration gives students production experience on real infrastructure and demonstrates diffusion in miniature. Faculty and compute: faculty exchanges with firms, retraining of existing professors, and pooled compute close the resource gap that keeps most campuses theoretical.
India’s path forward
First, a nationally coordinated AI talent mission—NITI Aayog projects a $250 to $300 billion services-revenue shortfall by 2035 without one. Closing it requires shifting from billing for effort to delivering outcomes — “human + agent + platform” models, in the roadmap’s language — which is precisely the capability the FDE layer provides and that India currently lacks.
Three tracks need to run in parallel. The first is employer-led competence building: the TCS-Anthropic partnership, announced in June 2026, gives 50,000 associates access to Claude through enterprise licensing, with certification through TCS iON and LTM’s AI 1000 programme that aims to train 1,000+ AI-certified engineers, including Forward Deployed Engineers. Second, practice-based credentials, set jointly by academia and industry, that test production competence rather than tool familiarity.
Third, higher-education reform that holds universities accountable for deployment-ready graduates, with curricula refreshed on industry cycles, faculty embedded in real systems, and continuing-education tracks. Fourth, engineer and researcher pipelines funded as separate investments, because no single pathway produces both. India’s market has proven it can scale literacy. India must strengthen coordination mechanisms, which can translate widespread participation into deeper diffusion resilience.
(The author is an Energy and Emerging Technologies expert.)
Published – June 22, 2026 08:30 am IST