Lightspeed's Mohapatra: India Needs a $50-100bn Bharat Fund to Own Equity of the AI Age

India needs a sovereign wealth fund (SWF) to invest in artificial intelligence (AI) and deep tech globally, and we need it now. This is not a matter of choice or preference. This is a matter of national importance and is directly correlated to India’s global relevance in the coming decades.

Let’s start with macroeconomics 101. We want what every growing nation wants: higher gross domestic product (GDP) and higher per-capita GDP. GDP per capita is basically a proxy for productivity and productivity has two levers: you either produce more with the same people (via automation) or you redeploy the same people to higher-value work (via specialisation).

AI is unusually good at both. That is why it is fundamentally deflationary. It lets organisations squeeze more output from the same people, and in many roles, it also raises the ceiling on what 'one good person' can do.

India’s growth story has always been sold as a demographic story. A young country, an expanding workforce, a long runway to move people from low-productivity work to higher-productivity work. It is a clean narrative that AI is about to make fragile, because AI attacks the one thing demographics give you for free: lots of human hours.

AI-based productivity gains won’t come everywhere at once but are arriving fast in in some categories: software development, back-office work, basic content creation, call-centre work and so on.

StackOverflow’s 2024 software developer survey with 65,000-plus respondents plotted by country and years of work experience shows that India has the sharpest and youngest peak around the two-year mark. That is exactly the band of work AI will overwhelm first: glue code, boilerplate, quality assurance, maintenance work, etc. India is exposed in a way we have not internalised yet. So, what does all this have to do with a SWF?

Capital and Labour

Historically, capital and labour have been tied together. For example, a logistics company can buy 100 vans, warehouses and more, but without human drivers, mechanics, staff, all they have is metal and concrete on their balance sheet.

Productive capital needed lots of human time to produce gains and these gains were then split between wages (labour) and profits (capital /shareholders). Classic Adam Smith economic theory. With how fast AI is moving, we may be able to scale output with much less additional human labour than before, especially in digital parts of the economy.

This was speculative a few years ago, but today AI company Cursor is reportedly near $500mn in annual run rate with approximately 300 people. Another AI company, Anthropic’s chief product officer Mike Kreiger has said that ~90% of the code produced by most teams is now written by AI. When one firm serves millions with a relatively small workforce, the profit share can rise and the labour share can fall even if lots of labour exists elsewhere.

In this future, the capital to own 'self-replicating' assets—robots that build other robots, code that writes other code—becomes a true substitute for labour, and the returns increasingly accrue to whoever owns the capital, not labour.

This has been happening in India over the past 50 years but is about to accelerate. And because a lot of the AI wealth is being created in private markets, and these private markets are often offshore, access itself becomes a new inequality engine. Not just between citizens of a nation, but between nations themselves. Put bluntly: you cannot buy your way into the best AI private deals through your retirement account. But a sovereign can.

Globally, SWFs are already behaving like this. Qatar’s QIA recently participated in Elon Musk’s xAI’s $20bn funding round. The UAE’s MGX is an equity holder in OpenAI’s Stargate data-centre project. When the wave is capital hungry, those with patient capital show up early, get board seats, get allocation and compound for decades. This is why we should stop treating SWFs as vanity projects only petro states do with windfall rents.

For India, an SWF is a strategic instrument to secure ownership in the 'rails' of the AI economy while we still can. India should create a Bharat SWF in the $50–100bn range. But how?

Financing an SWF

We do not have an oil windfall, do not run current-account surpluses and cannot finance this with fresh borrowing, given India’s debt constraints. The structure must be Indian: it must reallocate and monetise existing public wealth.

A pragmatic funding design could involve a limited and rule-based transfer from excess forex reserves into a long-duration investment fund. Other avenues could be quasi-royalties on digital public infrastructure for non-Indian entities operating in India or proceeds from sale of non-strategic public enterprises or residual stakes in illiquid assets. Long-duration domestic pools (Life Insurance Corporation, Employee Provident Fund Organisation) could also be allowed to participate as limited partners (LP) into specific sleeves of the SWF, subject to strict risk limits.

The mandate for this fund should be 60–70% run as a return portfolio with a deliberate tilt to AI, semiconductor, data centres, cloud, cyber, biotech and climate. This builds institutional credibility and generates non-tax income. The remaining 30–40% should be explicitly designed to build Indian capability. This is where you do things like co-invest in AI leaders with structures that create real spillover into India (R&D, compute clusters, etc). This sleeve could also take LP positions in Indian deep-tech funds or take targeted bets on domestic AI, data-centre, defence or semiconductor companies.

The Bharat SWF should be 100% state-owned under Indian law. It needs an independent, technocratic board, with a majority of directors who have deep investment and technology expertise. Government representation should be non-executive oversight with no veto on investment decisions within mandate. Fund performance and costs should be made transparent.

Time is Now

India is large enough to be strategically relevant, but too fiscally constrained to fund deep-tech bets purely from the Budget. We are precisely the country that will benefit most from converting idle or non-strategic public wealth into a professionally managed, return-seeking sovereign fund.

The bigger risk is that India does nothing or does it too late.

AI assets are already too expensive. The world is dividing into those who own the equity of the AI age through capital and those who merely rent it via labour. We have the engineers and the market. An SWF will ensure the value created by our talent around the world returns to us.

(The writer is partner, Lightspeed India)