India’s artificial intelligence strategy may be approaching an important inflection point. Having spent the first phase of the IndiaAI Mission building shared compute capacity, subsidising GPU access and supporting indigenous foundation models, the government is now considering a much larger intervention: becoming an anchor investor in the country’s frontier AI ecosystem.
The government is weighing an anchor investment of around ₹15,000–20,000 crore in a proposed National Frontier AI & Compute Fund (NFAICF), according to an Economic Times report. The proposed vehicle could finance frontier-model developers, GPU clusters, specialised data centres and other capital-intensive AI infrastructure.
The proposal was discussed at a closed-door meeting involving AI companies, investors and Electronics and IT Secretary S. Krishnan.
The crucial qualification is that the proposal remains under deliberation. The structure, overall corpus and government contribution have not been finalised, and the NFAICF is not yet an announced government scheme.
Yet the scale being contemplated is significant. An anchor commitment approaching ₹20,000 crore would be nearly twice the entire five-year outlay of ₹10,371.92 crore originally approved for the IndiaAI Mission in March 2024.
From subsidising compute to financing AI champions
The IndiaAI Mission was conceived around seven pillars covering compute, foundation models, datasets, applications, skills, startup financing and safe and trusted AI. The government has already made substantial progress in building shared infrastructure. Official data released in August showed that more than 45,000 GPUs had been brought into the shared compute ecosystem by June 2026, while 237 projects had received subsidised compute covering 93.18 lakh GPU hours.
India has also moved beyond merely talking about sovereign models. Twenty indigenous foundation-model proposals—12 large multimodal models and eight small language models—have been identified for support. Models from Sarvam AI, BharatGen, Gnani and others have already emerged from this programme.
The proposed NFAICF would address a different problem. Subsidised GPUs can lower the cost of experimentation and model training, but they cannot by themselves solve the capital requirements of companies attempting to build frontier-scale AI businesses.
Frontier AI increasingly resembles infrastructure development as much as conventional software entrepreneurship. Large models require sustained spending on accelerators, data centres, electricity, networking, storage, research talent, datasets and inference capacity. The investment horizon can also be much longer than that preferred by conventional venture capital.
This creates a financing gap that becomes particularly acute when Indian companies compete with global technology groups possessing enormous balance sheets and access to hyperscale computing infrastructure.
Patient capital as industrial policy
A sovereign-backed anchor fund could therefore represent a transition from AI promotion towards AI industrial policy.
One structure reportedly being examined is a SEBI-registered Category-I alternative investment fund, with investment decisions entrusted to an expert committee. Such an architecture could allow the government to provide catalytic capital without directly selecting and managing companies through the administrative machinery.
The larger objective would presumably be to crowd in institutional and private capital. If government money serves as an anchor rather than the entire corpus, participation from domestic financial institutions, sovereign funds, pension capital, family offices and strategic investors could potentially create a substantially larger pool of long-duration capital.
That distinction matters. The state does not necessarily need to finance every AI company. Its more useful role may be to absorb part of the early technological and infrastructure risk so that private investors are willing to deploy capital at greater scale.
The approach would also complement the existing IndiaAI Startup Financing pillar, whose stated purpose is to provide risk capital across the startup lifecycle and bridge funding gaps from prototyping to commercialisation.
Compute is becoming strategic infrastructure
The proposal also reflects a broader change in how governments view computing capacity. Advanced compute is increasingly being treated as strategic infrastructure, comparable in some respects to semiconductor fabrication, telecommunications networks and energy systems.
IndiaAI already describes common compute as a mechanism for providing affordable access to startups, researchers, academic institutions and government organisations. The government has simultaneously been expanding domestic AI infrastructure and pursuing semiconductor self-reliance as interconnected elements of technological sovereignty.
The NFAICF could deepen that strategy by financing not merely access to GPUs but ownership and development of the underlying AI stack.
That could have consequences beyond generative AI. Domestic compute capacity will increasingly matter for defence applications, scientific research, weather modelling, drug discovery, robotics, autonomous systems, cybersecurity and digital governance.
Governance will determine whether the fund works
But deploying ₹15,000–20,000 crore of public capital into a fast-moving technological sector also creates substantial governance challenges.
Frontier AI investments are inherently risky. Technologies become obsolete rapidly, hardware cycles are short and today’s leading model architecture can quickly be displaced. The government would therefore need to avoid turning the fund into either a conventional subsidy programme or a mechanism for protecting selected companies from market discipline.
Investment decisions would require technically sophisticated, professionally independent management. Clear rules would also be needed on valuation, conflicts of interest, follow-on investments, intellectual property, exit mechanisms and the government’s rights as an investor.
Most importantly, “sovereign AI” should not become synonymous with technological isolation. India’s advantage lies in combining domestic intellectual property, compute infrastructure and talent with access to global technology and markets.
If designed well, the proposed fund could fill one of the largest missing pieces in India’s AI architecture. India has already created subsidised compute, datasets, model-development programmes and a growing talent base. What frontier AI increasingly requires, however, is something harder to provide: large quantities of capital willing to wait.
The NFAICF proposal suggests that New Delhi is beginning to recognise that the next phase of the AI race will be determined not only by algorithms and GPUs, but also by who can finance them at scale—and for long enough.
