India’s AI ambition faces structural limits

A MOMENT at the India AI Impact Summit revealed more than the declaration itself. A university pavilion presented what it described as an indigenous robotic system. Within hours, it was identified as a commercially available Chinese robot dog. The exhibit was withdrawn. An apology followed. The episode was minor. The signal was not. India's sovereignty narrative in AI still rests on a global stack it does not control.

The circulated Delhi Declaration affirms that AI's benefits should be shared by humanity. It is a sound principle. It is also voluntary. The text "takes note" and "recognises". It creates no binding obligation. In economics and diplomacy, commitments without enforcement shape tone, not incentives.

This was the fourth in a sequence of global AI summits. Delhi was the first summit to be hosted by a Global South country. That mattered symbolically. But influence in AI will not be determined by sequence. It will be determined by control.

Most AI coordination today operates through soft law — standards, benchmarks, procurement rules and reputational pressure. The issue is not whether Delhi produced a treaty. It is whether it created tools that can shape behaviour. The Declaration established platforms and frameworks. None requires reporting. None mandates verification. No institution is clearly tasked with implementation.

The summit revealed a deeper divide. The UN argued for coordinated oversight and warned that without shared access to compute, developing economies risk exclusion. It is estimated that roughly $3 billion could provide meaningful compute access for developing countries. The US delegation rejected global governance. European representatives spoke of strengthening regulatory guardrails.

China — central to AI capability and chip supply chains — was not structurally embedded in the governance frame emerging from Delhi. India described itself as a bridge between advanced economies and the Global South. A bridge, however, requires stability on both sides.

The sharpest contradiction emerged on India's own stage. Leading AI firms acknowledged the risks of concentration and called for international coordination. The companies shaping the frontier were urging oversight. Their home government was declining it. That divergence — between corporate calls for coordination and sovereign resistance — captures the gap between convening and governing. Bringing actors together does not align their incentives.

The structural imbalance lies in the AI stack. The economy rests on three layers: compute, models and applications. India is strong at the application layer. UPI, Aadhaar and multilingual platforms such as BHASHINI operate at a national scale. They are credible templates. At the model layer, domestic firms have begun launching language models suited to Indian contexts.

Compute remains the constraint. India's expansion of its GPU pool is serious policy. Frontier clusters elsewhere operate at a much larger, dedicated scale. Compute is not merely hardware. It encompasses supply chains, export controls, energy capacity, cloud access and pricing power. Application strength built on externally controlled compute remains a form of dependence.

Two clear messages follow.

For the Global South, the Declaration is a beginning, not a solution. Inclusion will depend less on the language of communiqués and more on who controls access to compute, sets evaluation standards and defines procurement rules. India's digital public infrastructure is replicable. But replication requires resources most developing economies do not yet command.

For advanced economies, the message is different. The firms driving AI's frontier are urging coordination. Governments remain divided. If major powers do not converge on governance architecture, fragmentation will follow. Developing economies will absorb the costs.

India's parallel decision to join a US-led AI supply-chain initiative underscores the balancing act. Alignment may strengthen access. It also reduces room for neutrality. The test will be what India secures in return — influence over standards, access to compute or flexibility within export control regimes.

The summit's most consequential silence concerned labour. Its theme was "Welfare for all, Happiness of all." India's technology services workforce — roughly 5.8 million people — had no structured presence in the summit's power forums. The IMF estimates that about 40% of jobs globally could be affected by AI. Entry-level coding, testing and administrative work — the base of India's IT pyramid — is precisely where generative tools are reshaping demand. Labour appeared in commitments as a policy concern. It did not appear as a represented interest.

Prime Minister Modi's MANAV framework, moral systems, accountable governance, national sovereignty, accessible AI and valid oversight sets out the right ambition. The next phase will determine whether those principles translate into operational tools: audit capacity, supply-chain strategy, compute diffusion and legal clarity. That is not a criticism. It is the natural progression from declaration to implementation.

Delhi demonstrated convening power. It showcased strength at the application layer and ambition at the model layer. It placed India at the centre of the discussion. What it did not demonstrate is structural leverage over the variables that determine who governs AI: compute access, standards and binding rules.

A host sets the seating plan. A bridge connects systems. A passenger uses what others design. India is not merely a passenger. Whether it becomes a rule-shaper, rather than the world's most effective convener, is the question the next summit will sharpen.