India's sovereign artificial intelligence push is entering what analysts describe as a deeper infrastructure phase, with more than 1,700 AI-native companies in the country having collectively raised approximately USD 5.5 billion in equity funding to date, according to a report titled "India and the Sovereign AI Shift" released this week by Tracxn, the technology-sector data and analytics firm. The cumulative figure reflects the breadth of India's AI ecosystem — extending across enterprise AI, vertical applications, consumer platforms, and, increasingly, the infrastructure layers of the AI stack.

The report's central analytical observation is that India's AI ecosystem is transitioning from an application-led growth model — in which Indian start-ups primarily built consumer and enterprise applications on top of globally available foreign models and foreign-controlled cloud infrastructure — to a more infrastructure-backed and compute-intensive framework, supported by policy coordination and capital formation. This evolution is being driven simultaneously by the ₹10,372 crore IndiaAI Mission, by accelerating private-sector investment in compute infrastructure, and by a sharpened policy and industry consensus around sovereign capability — the latter substantively reinforced by the events of the past several days in the global AI sector.

The funding trajectory

The Tracxn analysis traces a distinct trajectory in Indian AI funding over the past four years. Total AI funding into Indian companies peaked at approximately USD 1.1 billion in 2022 — the high-water mark of the global AI-investment wave following the public emergence of large language models. Funding moderated through 2023 and 2024, in line with the broader global venture capital slowdown that affected technology investment across categories. The recovery began in 2025, with USD 856 million in funding, and has continued into 2026, with USD 626 million already raised in the year to date.

Set against the global AI funding picture, the Indian numbers are modest in absolute terms. Global AI funding has, by Tracxn's tracking, surpassed USD 473 billion cumulatively, with OpenAI, Anthropic, and Elon Musk's xAI accounting for an estimated USD 170 billion together — roughly 36 per cent of the global total. The concentration of capital at the very top of the frontier-model layer reflects the substantial scale economies and compute requirements of building general-purpose foundation models from scratch. India's positioning has, by necessity, been different — focused not on direct head-to-head competition with the largest US foundation model laboratories, but on building domestic capability across the wider AI stack at sustainable capital intensity.

The IndiaAI Mission

The principal policy lever driving the structural shift identified in the Tracxn report is the IndiaAI Mission, the Government of India's flagship national programme for AI capability development, formally launched with a financial allocation of approximately ₹10,372 crore. The Mission's central operational mechanism is the structured allocation of GPU compute capacity to selected domestic companies engaged in foundational or specialised AI model development — directly addressing what has historically been one of the principal capital barriers to large-scale AI work in India.

Under the Mission, twelve companies building foundational and specialised models have been allocated structured GPU capacity. The model parameter sizes currently being trained in India under this framework range from 2.9 billion parameters at the smaller end to 105 billion parameters at the larger end, with several teams using mixture-of-experts architectures designed for inference efficiency. The mixture-of-experts approach has emerged as an important architectural choice for jurisdictions with constrained compute access, because it allows substantially larger total parameter counts to be deployed without correspondingly large active-parameter inference costs.

Among the most visible beneficiaries of the IndiaAI Mission's compute allocation framework is Sarvam AI, the Bengaluru-based foundational model company, which has been allocated 4,096 NVIDIA H100 GPUs and supported through approximately ₹99 crore in compute subsidies. The Sarvam allocation is, by any measure, a substantial commitment of compute resources by Indian standards — and reflects the scale of public-private coordination that the Mission framework is enabling.

Moving up the AI stack

What the Tracxn report and the IndiaAI Mission framework together reflect is a deliberate, multi-year effort to extend Indian participation up the AI value chain — from applications and services, where India has been strong for years, into the underlying infrastructure and compute layers, where Indian participation has historically been thin.

The earlier phase of India's AI participation was characterised by start-ups building consumer-facing applications or enterprise services on top of foreign-developed foundation models, accessed via foreign-controlled cloud infrastructure (primarily AWS, Microsoft Azure, and Google Cloud Platform). This was a sustainable model commercially, but it left the Indian AI ecosystem structurally dependent on foreign technology platforms whose access terms, pricing, and policy decisions were ultimately set in Washington, Seattle, Mountain View, or Redmond.

The current phase, on the Tracxn analysis, reflects deeper participation across upstream layers — including compute provisioning through Indian data centre operators, foundational model development supported by IndiaAI Mission compute allocations, and emerging semiconductor-design and fabrication-related activity supported by the Production Linked Incentive (PLI) schemes for semiconductor manufacturing and design.

The Anthropic Fable and Mythos episode

The pace of this transition has been substantively sharpened by the events of the past several days. On 12 June 2026, the US Commerce Department issued an export-control directive instructing the AI company Anthropic — one of the principal foundation-model developers globally — to suspend access to its two most advanced models, Claude Fable 5 (publicly available for enterprise and general use) and Mythos 5 (restricted to a limited group of trusted-organisation users), for all foreign nationals, whether inside or outside the United States. Anthropic complied with the directive at short notice and disabled access to both models worldwide.

In its public statement, Anthropic noted that it disagreed with the US government's determination — which it understood to be based on the discovery of a potential jailbreak technique that might allow users to bypass safety guardrails in Fable 5 and access the more powerful cybersecurity capabilities of the underlying Mythos model. Anthropic publicly stated that the cybersecurity capability in question is "widely available from other models" including OpenAI's GPT-5.5, and that the action did not adhere to the principles of a "transparent, fair, clear, and grounded in technical facts" statutory process. The company has nonetheless complied while engaging with the White House on the question.

For Indian technology stakeholders, the episode crystallised concerns that had been articulated for some years but had not previously had so concrete a real-world instance attached to them. Until the Fable and Mythos directive, US export controls on AI had primarily focused on hardware — high-end GPUs and chip-fabrication equipment. The 12 June directive marked the first time that AI software itself had been subjected to a comparable export-control regime, treating the model weights as the strategic asset rather than the silicon they run on.

The industry response

The response from Indian technology leadership was both swift and substantively aligned on the policy direction. Mr Sridhar Vembu, founder of Zoho Corporation, framed the episode in the most direct terms — describing technology as "the ultimate weapon" and arguing that "globalisation is dead and Bharat must find her own way ahead." Mr Vembu has, over recent years, consistently advocated for Indian development of smaller open-source models, expanded domestic R&D investment, and reduced dependence on foreign technology platforms.

Mr Siddarth Pai, founding partner at the venture capital firm 3one4 Capital, echoed the broader argument, calling for India to invest in semiconductor design alongside open-source model development. The investor Sandeep Mall, separately, emphasised the need for India to build foundational AI models rather than relying on services built on foreign platforms.

The position is also being articulated at the level of major Indian industrial leadership. At the India AI Impact Summit 2026 in February, Mr N. Chandrasekaran, Chairman of Tata Sons, framed AI as the "infrastructure of intelligence" and called for a full-stack national strategy spanning compute, data, models, and applications. The framing reflects the conceptual shift that the IndiaAI Mission and parallel private-sector commitments are operationalising.

The data centre build-out

Reinforcing the policy and start-up activity is a substantial wave of investment in AI-ready data centre infrastructure by India's largest industrial groups. Reliance Jio is developing multi-gigawatt, AI-ready data centres, including a facility at Jamnagar in Gujarat that is expected to add more than 120 megawatts of capacity in the second half of 2026. Adani Enterprises has separately committed to invest approximately USD 100 billion by 2035 in AI-enabled, renewable-powered data centre infrastructure.

The economic logic underpinning this build-out is distinctive. Both Reliance and Adani benefit from substantial backward integration into renewable power generation, which means their data centres can be located adjacent to power plants, reducing transmission losses and shielding them from rising grid electricity prices. As one independent market analyst observed, "with their backward integration, renewable-powered data centres are simply the cheapest option for them in the long run." The intersection of low-cost renewable power, integrated industrial groups, and growing AI compute demand creates a structural advantage for Indian data centre investment that few other markets globally can replicate at scale.

The honest reading

For all the momentum, several structural gaps remain candid in the Indian AI picture. India does not yet have large-scale advanced-node semiconductor fabrication capability — meaning the high-end GPUs that Indian foundational model developers train on continue to be imported, primarily from Nvidia, and to be subject to US export-control oversight. India's first major fab projects, under the PLI semiconductor schemes, are at advanced planning and early-construction stages but will not produce frontier-node chips in commercial volumes for some years.

The funding-scale gap between Indian AI investment and the frontier US laboratories also remains substantial. Indian cumulative AI funding of approximately USD 5.5 billion represents a meaningful national capability investment, but it is roughly an order of magnitude smaller than the cumulative funding raised by OpenAI alone, and substantially smaller than the combined frontier-model investment in the US and China. Closing this gap meaningfully — to the extent that India can compete at the frontier-model layer — would require sustained increases in either private capital flowing into Indian AI, public investment, or both.

The talent challenge is also real. India produces a large number of competent AI engineers and researchers, but the very top tier of research talent has historically tended to migrate to US-based laboratories, where compensation, compute access, and peer-research density are substantially superior. The IndiaAI Mission, by allocating compute resources to domestic teams, is addressing one of the principal factors in this dynamic; the others — research culture, peer density, and compensation — will take longer to address.

What the moment establishes

What the past several days have established, taken together with the cumulative picture the Tracxn report describes, is the following: India's AI ecosystem has reached a meaningful scale, the policy framework is now actively supporting the build-out of upstream capability, the private-sector commitment to the supporting infrastructure is substantial, and the strategic case for sovereign AI capability has been substantively reinforced by the visible vulnerability of foreign-platform dependence.

The Anthropic episode is, in this sense, less a turning point than a clarifying moment. The arguments for Indian sovereign AI capability — articulated by Mr Vembu, Mr Chandrasekaran, the IndiaAI Mission team, and a range of others over the past several years — have not changed substantively in their content. What the Fable and Mythos directive has done is to make the concrete, operational reality of platform-dependence visible in a way that academic and policy argument alone had not achieved. A category of foreign capability that Indian researchers and companies had been using as part of their working stack disappeared, at short notice, on the basis of a regulatory determination made in Washington. The clarifying effect of this has been substantial.

The road ahead

The immediate horizon for India's sovereign AI build-out involves several parallel work streams. The IndiaAI Mission will continue to roll out compute allocations to additional companies and to refine the model-development support framework. Private-sector data centre investment by Reliance, Adani, and other groups will continue to add capacity over the next several years. The semiconductor PLI projects will continue to move from planning to construction to production over a multi-year horizon. The foundational and specialised model development work across Sarvam AI, Krutrim, and the other supported teams will continue to mature, with successive model releases expected through 2026 and 2027.

For now, what the Tracxn report and the wider context establish is a substantive baseline: USD 5.5 billion of cumulative equity funding, a flagship national mission with structured compute allocations, multi-billion-dollar private-sector infrastructure commitments, and a policy environment increasingly oriented around full-stack sovereign capability. The harder questions — of whether India can close the frontier-model gap, of whether semiconductor capability can be built at frontier scale, of whether the research-talent retention picture can be improved — remain open. But the institutional and capital foundation, on the evidence of the past several months, is now substantively in place.