The India AI Impact Summit 2026 concluded at Bharat Mandapam with the endorsement of the New Delhi Declaration by 88 countries and international organisations—among them the United States, China, the United Kingdom, France, Germany, Japan, Brazil, Russia, the European Union, and IFAD. The moment marked more than the conclusion of a diplomatic gathering. It signalled a subtle recalibration in the global geography of artificial intelligence governance. It represented an important repositioning of India in the global conversation on artificial intelligence.
This was the largest AI summit convened to date and the first of comparable scale hosted in the Global South. Previous summits in advanced economies tended to emphasise AI safety, regulatory oversight, and long-term existential risk. The New Delhi edition broadened the lens. Its focus shifted from containment to deployment, from frontier anxiety to developmental application, from laboratory debates to societal transformation.
That shift is not rhetorical. It reflects a structural transition in the nature of AI itself.
Artificial Intelligence is no longer an experimental frontier discipline confined to research labs and venture-backed startups. It is becoming economic infrastructure—akin to electricity, telecommunications, or the internet. Its influence spans productivity, scientific discovery, financial systems, public administration, healthcare delivery, education, defence and creative industries. Nations are therefore no longer merely regulating AI; they are positioning themselves within its value chain.
For India, the summit was not symbolic diplomacy. It was strategic positioning.
NORMATIVE LEADERSHIP IN A FRAGMENTED WORLD
The New Delhi Declaration is organised around seven pillars: collaborative development, democratic diffusion of datasets and compute, trustworthy systems, AI for science, human capital, resilient infrastructure and inclusive economic growth. It is not legally binding. That design is deliberate.
In a geopolitical environment marked by strategic competition, particularly between major technology powers, binding supranational regimes are politically unrealistic. Normative convergence often precedes enforceable architecture. By framing shared principles without triggering sovereignty anxieties, the Declaration seeks to build a broad coalition of intent.
Its most consequential idea is “democratic diffusion.” Today, frontier AI capabilities are concentrated among a handful of firms and countries. Access to advanced semiconductors, hyperscale compute clusters and large proprietary datasets determines who can train state-of-the-art models. Without diffusion, technological concentration risks hardening into structural inequality.
to showcase India’s AI startup ecosystem and responsible AI vision
Prime Minister Narendra Modi’s articulation of MANAV—human-centric AI rooted in the ethic of Sarvajan Hitaya, Sarvajan Sukhaya—supplied philosophical coherence. The term reframes AI from an instrument of dominance to an instrument of collective welfare. It implicitly challenges the assumption that technological progress must be zero-sum.
Yet normative aspiration must be anchored in material capability. The frontier reality of AI development is defined by scale.
SCALING LAWS AND THE ECONOMICS OF INTELLIGENCE
Modern AI progress is governed by empirical scaling laws. As model parameters increase, training data expands and compute budgets grow, performance improves in relatively predictable ways. Over the past five years, leading large language models have expanded from billions to trillions of parameters. Training runs now require tens of thousands of advanced GPUs operating in parallel for weeks or months.
The cost is immense. Frontier model training can require investments exceeding several billion dollars when accounting for hardware, energy and engineering expertise. Energy consumption for a single large training run can reach gigawatt-hours. Inference—the generation of outputs for millions of users— also demands sustained computational throughput.
This compute intensity has created concentration. Companies such as OpenAI and Anthropic operate with backing from hyperscalers and semiconductor leaders capable of financing vast GPU clusters. Advanced AI accelerators depend on sophisticated fabrication processes, dominated by a small number of global manufacturers.
India does not yet compete directly at this frontier tier. Its semiconductor ecosystem remains nascent at advanced nodes. Public R&D expenditure remains below 1% of GDP. The number of globally leading AI researchers, while growing, is still modest compared to the United States or China.
But frontier dominance is only one dimension of AI leadership.
ADOPTION AT SCALE: INDIA’S DIFFERENTIATED PATH
India’s comparative advantage lies not in frontier model training but in deployment at population scale.
With roughly 900 million internet users and among the highest digital transaction volumes globally, India constitutes one of the largest real-world laboratories for AI applications. Scale provides feedback loops: models improve faster when exposed to diverse, high-volume usage contexts.
Surveys suggest that nearly 90% of Indian enterprises report some degree of AI integration, significantly above global averages. Applications span credit risk modelling, logistics optimisation, agricultural yield prediction, precision marketing, educational personalisation and telemedicine diagnostics.
Voice interfaces illustrate contextual adaptation. India’s linguistic diversity creates barriers for text-based systems. Speech-driven AI expands accessibility dramatically. Startups such as Sarvam AI have fine-tuned open-source models on Indian language corpora, enabling strong performance in translation, document analysis and conversational assistance across multiple scripts.
This application-first strategy reflects structural pragmatism. Constraints in capital and compute have encouraged efficiency and localisation rather than pure scale escalation. India’s innovation is therefore often frugal, domain-specific and socially embedded.
The result is a distinct trajectory: not artificial general intelligence as an immediate objective, but applied intelligence embedded within everyday systems.
THE COMPUTE BACKBONE: DATA CENTRES AS STRATEGIC ASSETS
AI systems depend on compute. Compute depends on physical infrastructure—data centres, fibre connectivity, cooling systems and reliable electricity.
India’s data-centre capacity has expanded rapidly, reaching approximately 1.3 gigawatts of installed capacity, nearly triple the level in 2020. Though still significantly smaller than the United States (approximately 39 gigawatts) or China (around 9-10 gigawatts), the growth rate is notable.
Major domestic and international players are investing aggressively. The Adani Group has articulated large-scale digital infrastructure ambitions. NTT Data remains a leading operator in Indian markets. Global hyperscalers such as Alphabet Inc. and Microsoft have announced multi-billion-dollar commitments to expand AI and cloud infrastructure in India.
THREE STRUCTURAL DRIVERS UNDERPIN THIS EXPANSION
First, data localisation policies increasingly require sensitive data—particularly financial data—to be stored domestically. As India’s digital public infrastructure generates vast volumes of transactions, domestic processing capacity becomes strategically valuable.
Second, competitive federalism has accelerated infrastructure growth. States including Maharashtra, Karnataka, Tamil Nadu and Telangana offer preferential electricity tariffs, land facilitation and renewable-linked incentives to attract investment.
Third, cost and demand dynamics favour expansion. India’s growing digital economy generates sustained demand for low-latency services. Proximity to compute reduces operational costs for startups and enterprises alike.
Critics argue that data centres create limited long-term employment once construction concludes. That critique overlooks ecosystem effects. Data centres anchor cloud ecosystems, AI startups, cybersecurity services, semiconductor design firms and renewable energy integration projects. They represent the physical substrate upon which higher-order innovation rests.
In the AI era, control over compute infrastructure is not peripheral—it is foundational.
ENERGY, SUSTAINABILITY AND THE SEMICONDUCTOR QUESTION
AI workloads are energy-intensive. As both training and inference demands expand, electricity consumption will rise sharply. Globally, concerns are emerging about grid strain and carbon emissions associated with AI data centres.
India’s expansion must therefore integrate renewable energy at scale. Several states are incentivising data centres linked to solar and wind capacity. Aligning AI infrastructure growth with sustainability objectives is not merely environmentally prudent; it is economically strategic, given rising global scrutiny of energy-intensive technologies.
THE DEEPER STRATEGIC CHALLENGE LIES IN SEMICONDUCTORS
Advanced AI accelerators depend on cutting-edge fabrication nodes and intricate supply chains. Even partial domestic capability in design and packaging would strengthen resilience. India’s semiconductor initiatives, including fabrication incentives and design-linked programmes, represent early steps toward reducing structural dependency.
Technological sovereignty in the AI era is inseparable from semiconductor capability.
RESEARCH DEPTH, CAPITAL FORMATION AND TALENT RETENTION
Infrastructure and adoption alone cannot substitute for research depth.
India must expand doctoral training in machine learning, computational neuroscience, robotics and AI ethics. University-industry collaboration requires strengthening. Long-horizon public laboratories capable of foundational research—beyond immediate commercial returns—are essential if India seeks to shape rather than merely apply global AI trajectories.
Risk capital for deep-tech ventures must mature further. Frontier research is capital-intensive and uncertain; it requires patient investors.
Talent retention remains critical. India produces one of the world’s largest engineering cohorts annually, yet leading AI researchers often migrate to ecosystems offering advanced facilities and compensation. Creating competitive research environments domestically is central to reversing that flow.
These are structural challenges, but they are addressable through sustained policy coherence.
DIPLOMACY, PAX SILICA AND STRUCTURAL CONVERGENCE
The Summit demonstrated India’s ability to convene diverse geopolitical actors. In a fragmented technological order, bridge-building carries strategic value.
The signing of Pax Silica symbolised convergence—of technology, finance and skills. “Silica,” the elemental foundation of semiconductors, evokes the material substrate beneath digital intelligence. AI leadership requires integrated ecosystems: capital markets capable of funding infrastructure, educational systems producing specialised talent, regulatory clarity enabling innovation and international partnerships facilitating supply chains.
Normative leadership without structural depth risks symbolism. Structural investment without normative clarity risks fragmentation. The Summit attempted to align both.
FROM INTENT TO IMPLEMENTATION
The AI Impact Summit represents a transition in global discourse. AI governance is moving from theoretical risk management toward practical capacity-building. Nations are asking not only how to regulate AI, but how to integrate it responsibly into economic and social systems.
India’s distinctive proposition lies in scale, affordability and inclusion. Its experience with digital public infrastructure has demonstrated the ability to deploy technology to hundreds of millions at low cost. Extending that philosophy to AI—guided by MANAV’s human-centric ethos and strengthened by compute infrastructure—could yield a developmental model distinct from frontier concentration.
Global AI leadership will not be defined solely by who trains the largest model. It will also be defined by who ensures that intelligence is diffused responsibly, embedded sustainably and aligned with human welfare.
The applause at Bharat Mandapam marked consensus.
The real test lies ahead—in research funding allocations, semiconductor fabrication milestones, renewable integration strategies, university reforms, startup financing mechanisms and cross-border partnerships.
The New Delhi Declaration is voluntary. The delivery, if India seeks enduring influence in the AI century, cannot be.
The summit may have ended. The harder work—institutional, technological, diplomatic—begins now.
*The writer, a Harvard educated civil servant, is a former Secretary, Ministry of Information & Boradcasting, Government of India. He also served on the Central Administrative Tribunal and as Secretary General of ASSOCHAM. He commands extensive expertise in the fields including Media and Information, Industrial and Labour Reforms, and Public Policy.
