Within the next few days I realised one of the major things I didn’t take into account. Finances. Making a drone would require some serious money, and it was hard to find someone who could take the initial risk of funding the project. Then my mind went into its overthinking mode. What about the money required after the initial prototype is built? How will I arrange that? Who will fund my dreams?

The Ecosystem Before the Money

The next morning, I walked into SIIC, IIT Kanpur’s startup incubation centre.

I expected to find funding.

Instead, I found an ecosystem.

That distinction matters.

One of the biggest mistakes first-time founders make is assuming incubators exist to write cheques. They don’t. They exist to increase the probability that your startup survives long enough to deserve one.

Both founders on our latest podcast began from the same place. They didn’t choose IIT Kanpur because it had the largest grants. They chose it because deep tech needs far more than money.

A drone startup needs laboratories before investors.

A medical device startup needs testing facilities before customers.

Both need professors who have solved similar problems, seniors who have already built companies, talented interns who can join for a fraction of industry salaries, and a manufacturing ecosystem that is already familiar.

The founders repeatedly made one point that most startup advice misses.

If you are building software, you can rent a flat in Bengaluru, open your laptop, and start coding.

But if you are building hardware, your office is only a small part of the company. The real office is the machine shop, the testing lab, the fabrication workshop, the research facilities and the people around them.

That is why almost every successful Indian deep-tech founder seems to emerge from an incubation centre attached to a research institution. IIT Kanpur, IIT Madras, IIT Bombay, IISc and several others have gradually evolved from providing office space to providing access to equipment, fundraising support, industry mentors and even their own venture funds. The ecosystem has matured significantly over the last five years.

Then came the most surprising part of the conversation.

The founders argued that for many first-time deep-tech entrepreneurs, the first meaningful capital doesn’t come from venture capitalists at all. It comes through incubators.

Because incubators often become the most trusted source of capital at initial stages.

Government departments, CSR programmes and funding agencies are far more comfortable backing a startup that is being reviewed, mentored and monitored by a reputed incubator than one operating out of a rented apartment. That trust becomes the bridge between an idea and its first prototype.

Only after that bridge is crossed does the real funding journey begin. That journey starts with grants. And today, for India’s deep-tech founders, there are more of them than ever before.

Money Exists. Most Founders Just Don’t Know Where to Look.

I had always assumed there were only two ways to fund a startup.

Turns out there is a third.

The government has built an entire ladder of grants for technology founders. The trick is knowing which rung you stand on.

If you are still exploring an idea and need time to work on it full-time, there are Entrepreneur-in-Residence (EIR) fellowships. They provide a monthly stipend so that founders can leave their jobs and build without worrying about rent for a while.

Once the idea becomes a prototype, programs like NIDHI PRAYAS step in. As Nikhil Kurele put it, “There are schemes like NIDHI PRAYAS... where you can actually go and apply for your prototype, whatever TRL (Technology Readiness Level) level your technology is available.” These grants are specifically meant to bridge the gap between an idea and a working prototype, while also giving access to mentors, fabrication labs and incubators.

Then there are sector-specific grants. Biotechnology startups can apply for BIG (Biotechnology Ignition Grant). Defence startups have iDEX and ADITI. Mining technologies have PRISM. CSR funds routed through incubators often support startups solving social problems as well.

Nikhil gave perhaps the best perspective during our conversation.

“I have seen startups... we got ₹10 lakh to build our prototype. I know a company in the medical device space that got ₹2.5 crore at the incubation centre level to develop its prototype.”

The message was simple.

Don’t start by asking, “Who will invest in me?”

Start by asking, “Which grant fits the stage I’m in?”

Because long before venture capital enters the picture, India already has a surprisingly well-built funding pipeline.

One word of caution though!

Don’t imagine these grants as free money waiting to be claimed.

There is paperwork. There are reviews. There are committees. Some schemes move surprisingly fast through incubators, while others can take months because multiple ministries and expert panels are involved. Both founders were honest about that. Bureaucratic delays do exist. But they also argued that complaining about them rarely helps. As Nikhil put it, “There is no point of whining... you have to take whatever best that you can squeeze from them.”

Ramakrishna added another perspective that every founder should remember. Government grants are meant to reduce risk, not eliminate it. He advised founders to treat them as an accelerator rather than the foundation of the business.

“We were more relying on private capital actually, but RDI Fund has come as a bonus to accelerate... and attract the private capital in a shorter span.“

Or as Nikhil put it even more simply while talking about early-stage grants:

“Don’t plan big things in the early stages... consider it as a bonus and a buffer.”

That is probably the healthiest way to think about government support.

Use grants to build the technology, but build a business that can survive even without them.

The biggest missing piece, however, wasn’t prototype money in India’s deep-tech struggles.

It was the patient capital. And that’s exactly where the RDI Fund enters the story.

What Exactly Is the RDI Fund?

The Research, Development and Innovation Fund is a ₹1 lakh crore initiative approved by the Union Cabinet in July 2025 and formally launched by the Prime Minister at ESTIC 2025 on November 3. It sits under the Department of Science and Technology and is routed through the Anusandhan National Research Foundation. Its job is simple to state and hard to execute. Fund technologies at TRL 4 and above, cover up to half the project cost, and do it through long-term, low-interest loans or equity, so that private capital feels safer entering deep tech.

Ramakrishna explained the thinking behind it. The government wanted to “Reduce the burden on private capital entering into this space,” and the fund was built on a straightforward split. “We will do 50% of it.”

For EndureAir, this wasn’t the first pitch. The problem statement was a 200 kg payload drone for high-altitude and maritime terrain, the kind of logistics support Siachen and Tawang badly needed. Ramakrishna called it “First of its kind across the globe.”

The evaluation, according to Nikhil, moved through three filters. Team, then technology, then market. “They spend like I would say 40% or 30% of the time on the problem statement and the solution,” he said, before checking whether the founding team was equipped to solve it.

Timing mattered as much as substance. There was no fixed deadline, but Ramakrishna admitted the process quietly rewarded speed. “It’s a first come first serve basis.” His team submitted the best version of the proposal they could, “Not waiting for maybe the perfect proposal.”

Once shortlisted, founders answered detailed written queries before a 10 to 12 minute presentation followed by 20 to 25 minutes of questions. Nikhil’s advice for anyone applying was to over-prepare rather than trim. “Don’t cut corners or shortcuts here.”

The money itself came as a hybrid, part loan, part optional equity, at 3% simple interest, with a two year moratorium and five year repayment.

But neither founder treated RDI as the finish line. Ramakrishna was clear about what it actually was. “RDI Fund has come as a bonus to accelerate the private capital in a shorter span.”

The prototype could now be built. The real test was still the market.

Back to the Room

I sat back in my room, the same room where I had yelled “Eureka!” months ago. Except this time I wasn’t overthinking about money.

Grants exist. Incubators exist. RDI exists. The government has built a ladder that didn’t exist for founders a decade ago. Nikhil built Noccarc on it. Ramakrishna is building EndureAir on it. The excuse of “no funding in India” doesn’t hold the way it used to.

But sitting there, I realised something the founders had already learned the hard way. Money was never the real problem. It was only the loudest one.

The real problem is more nuanced and harder to see. It is whether you actually understand the soldier standing in Siachen, the patient in an ambulance with no doctor, the district hospital with no specialist.

Grants can fund a prototype but they cannot fund the years of sitting with a problem until you actually feel it.

Both founders said the same thing in different words. Product wins. Trust wins. Everything else is a bonus.

So maybe the real question was never “Who will fund my dreams.” It was always the one I asked myself before any of this began. Do I understand the problem well enough to deserve the money at all?

That plan I made in my room? It was never fool-proof.

It was just the first draft.

-By Sanidhya Agarwal
Editor, The Rocket Media

This confession does not simply disappear. It gets stored, broken into tokens, and becomes part of the data that improves future AI models. Nobody in San Francisco knows this person’s name. But together, millions of conversations reveal something valuable. They show exactly how a heartbroken twenty-two-year-old in a Tier-2 Indian city talks about grief at one in the morning.

That’s the core of what’s happening. AI companies are collecting something valuable from ordinary Indians: their unfiltered emotional lives, shared through a chat window at their most vulnerable moments. They collect it for free. They use it to build better AI systems. Then they sell those improved systems back to India through paid subscriptions. Actually, if you think deeply, this pattern is quite familiar to us. Remember how in the colonial era, raw cotton was bought from us at a minimum price and then sold to us as finished cloth at premium prices. But now, the raw material is no longer cotton. It is human emotion.

A market nobody is being subtle about

India is not just another market for AI companies. It is one of the most important ones. ChatGPT now has around 100 million weekly active users in India, making it OpenAI’s second-largest market after the United States, and the largest market in terms of student users. Speaking at India’s AI summit earlier this year, Sam Altman described the country’s opportunity in a single phrase: a “full-stack AI leader.”

Google moved quickly as well. It partnered with Jio to offer tens of millions of young Indians eighteen months of Gemini Pro for free (a service worth about ₹1,950 a month). The offer proved especially popular with users between eighteen and twenty-five years old. Within months, Google expanded it to subscribers on cheaper Jio plans as well. Some economists would describe this as ecosystem building rather than extraction. They would argue that this is simply the cost of entering a market as large as India. That argument deserves to be taken seriously, and we will return to it later.

Meta barely needed a separate launch. It simply added Meta AI to WhatsApp, an app already used by more than 535 million Indians every month. By May, WhatsApp’s global head said that India had become the world’s largest market for Meta AI, with more monthly active users than any other country.

When Sundar Pichai spoke in Delhi this year, he called AI the “biggest platform shift of our lifetime.” It was more than a statement about technology. It reflected India’s growing importance, not just as a market to sell into, but as a place that helps shape the future of AI.

The machine that turns confessions into code

To understand why this matters, it helps to see how a private conversation becomes part of a better AI system.

When you correct a chatbot, give its answer a thumbs up or thumbs down, or rewrite your question because the first reply missed the point, that feedback does not disappear. It becomes a training signal. AI companies use millions of these signals to build what are called reward models. They then use those models to train their AI systems to produce better responses.

This is not speculation. It is a well-documented part of how modern AI is built. It is also the process that helped transform GPT-3 from a research model into the conversational assistant that millions of people now use every day. The idea was simple. Human reviewers repeatedly compared two answers and chose the better one. Over time, the model learned what people considered a good response. None of this works without a constant stream of real conversations, and real people showing the model, one correction at a time, whether it truly understood them.

For a market as large as India, this supply of human feedback has a company, a workforce, and a business model behind it.

Karya, a data-labeling company, has built a network of more than 32,000 workers across rural and semi-urban India. They transcribe speech, evaluate AI responses, and label text in Indian languages for clients including Microsoft, Google, and Anthropic. Workers earn around five dollars an hour, nearly twenty times India’s minimum wage. As co-founder Manu Chopra argues, “the wages that exist right now are a failure of the market.” Karya proves that data work can be paid fairly. The problem is that it remains the exception.

Most users, however, are not paid at all. Every prompt, correction and conversation they have with AI becomes valuable interaction data that helps improve future models. OpenAI and Google both state in their privacy policies that user conversations may be used for training by default, unless users actively opt out.

That is how the system works. A conversation becomes a training example. The training examples make the model slightly better. The better model becomes the next paid subscription.

Why India has so much to give

That system depends on a constant flow of deeply personal conversations. India provides an unusually large supply of them, for one difficult reason.

The country has a severe shortage of mental health care. Between 83 and 86 percent of people who need mental health support do not receive it. By some estimates, India has only 0.3 to 0.75 psychiatrists for every 100,000 people, far below the World Health Organization’s recommended minimum of one. Into that gap comes a free chatbot that is always available, always patient, and always ready to listen.

It is not surprising that many people turn to it. A survey of young Indians conducted last August found that 57 percent had used AI chatbots for emotional support. They spoke about loneliness, anxiety, relationship problems, and thoughts they might once have shared with a close friend. Among school-aged respondents, that figure rose to 88 percent during periods of intense anxiety.

Researcher Alisha Butala says these were not conversations about homework or careers, but about real emotional struggles. Chatbots appealed because they were always available, never interrupted, and never seemed impatient. For many, they became a substitute for support that an already strained mental health system could not provide.

Yet 58 percent also worried about what would happen to the information they shared. They kept talking anyway because they felt they had few better options.

That is why India is more than just a large market for AI companies. It provides something uniquely valuable: millions of honest, emotionally charged conversations that continue to train AI systems every day.

An old trade, wearing new clothes

Nick Couldry has a name for this pattern. He is the co-author of Data Grab: The New Colonialism of Big Tech and How to Fight Back, and he argues that today’s data economy is built on a familiar assumption: if data exists, it is there to be collected, and whoever reaches it first has the right to use it. As he has put it, “That was what colonialism was about, just taking everything.”

The comparison is not entirely new. During colonial rule, India exported raw cotton at very low prices and imported finished cloth at a much higher cost. Coffee-producing countries still sell raw beans for very little, only for those same beans to return later as expensive cups of coffee after being processed elsewhere. The pattern has existed for centuries. What has changed is the raw material. Today, it is not cotton or coffee leaving the country. It is human emotion, shared freely through a phone in Kanpur, and returning later as a paid monthly subscription.

Follow the money

The pattern is already visible elsewhere in India’s digital economy. In 2024–25, Google India reported ₹34,742 crore in advertising revenue, but over 92 percent was paid to Google’s Singapore entity as the cost of acquiring the advertising inventory it sold in India. The older model of value extraction has not disappeared. AI is simply built on a different raw material.

OpenAI, valued at $852 billion with annual revenue of around $25 billion, is built on people talking to its models. Roughly one in every nine ChatGPT users now comes from India, yet the company’s Mumbai entity has a paid-up capital of just ₹1 lakh.

Most of India’s 100 million ChatGPT users are currently on the free ChatGPT Go plan. But once they begin paying, ChatGPT Plus costs ₹1,999 a month, almost the same as in the United States, despite the vast difference in average incomes.

What does the person who provided the raw material receive in return? Very little. Researchers who have tried to estimate a “data dividend,” a payment to individuals for the value of their personal data, usually arrive at a surprisingly small number. For most people, it amounts to only a few dollars a year. Institutional data, however, is treated very differently. AI companies have signed licensing agreements with publishers worth millions of dollars. In News Corp’s case, the deal is reportedly worth more than $250 million over five years. The pattern is familiar. The person who produces the raw material captures very little of its value. The organization that collects, combines, and processes it captures most of it. This time, the difference is even starker. Many people do not even realize that the exchange has taken place.

The gap between who is protected and who is not extends beyond individual users. Earlier this July, Microsoft’s CEO, Satya Nadella, described what he called the “Reverse Information Paradox.” His argument was that every prompt, every correction, and every improvement a company makes while using an AI system can become valuable training data for the model provider, unless that organization keeps its learning process within its own boundaries. His advice to businesses was straightforward: control your own learning loop, keep your corrections inside your own systems, and do not allow that knowledge to improve someone else’s models for free.

It is the same mechanism described earlier in this article. Conversations are logged, turned into tokens, and used to improve future AI systems. The difference is who receives the warning. Large companies are told to protect their knowledge. Ordinary users are not.

India has already accepted the basic idea behind this debate, but only for one group of people. In December 2025, a government working paper proposed a system called “One Nation, One License, One Payment.” Under the proposal, AI companies would receive a single, mandatory license to train on copyrighted Indian works such as books, journalism, and music. In return, they would share a portion of their global AI revenue with a new collecting body that would distribute payments to creators. The details remain controversial, but the principle is clear. If your work helps train a commercial AI model, you should receive compensation.

That principle has not been extended to everyone else. The twenty-two-year-old sharing the details of a breakup at one in the morning is also contributing something valuable. In many ways, that conversation may be even more useful for teaching an AI model how to respond like a human than another news article collected from the internet. Yet that person is missing from the discussion. They are not included when rules about compensation are designed. A feeling carries no copyright, even when it helps build a valuable product.

The fair counterargument

The strongest response to this argument is not that it is wrong. It is that it is incomplete. That view deserves to be taken seriously, not brushed aside before the discussion moves on.

The economic case is straightforward. Raw data, on its own, has very little value. Turning it into a useful AI system requires enormous investment in computing power, research, and engineering. Today, most of that frontier work is still happening outside India. Some economists therefore prefer the term “AI nationalism” instead of digital colonialism. In their view, India is not a passive victim. It is an important participant with growing bargaining power. From this perspective, free Gemini subscriptions through Jio and discounted ChatGPT plans are not examples of extraction. They are ecosystem-building strategies, similar to the below-cost expansion that technology companies have long used when entering major markets. Legally, the arrangement is also transparent. The terms of service are available, users can opt out of certain kinds of data collection, and every one of those 100 million weekly ChatGPT users agreed to those terms before starting a conversation.

These are serious arguments, and they make the issue more complicated. But they leave one important question unanswered. If users are contributing something valuable without realizing it, why is the option to stop contributing buried inside settings that few people ever find? Why is that choice turned on by default? And why does the benefit of that arrangement flow almost entirely in one direction?

What could actually change

None of this leads to a simple solution, and it is important not to pretend otherwise. Banning AI products is neither practical nor desirable. Every week, around 100 million people in India choose to use them for reasons that regulation alone cannot change. But the absence of an easy solution does not mean there are no meaningful ones. At least three ideas are already within reach.

The first builds on a proposal India has already considered. If the country can create a system that returns part of AI revenue to authors, journalists, and other creators whose work helps train AI models, there is no obvious reason the same principle could not apply to ordinary users whose conversations also contribute to those systems. The payment would not need to be large. It could work like a streaming royalty: small, automatic, and tied to participation rather than hidden behind settings that almost nobody changes.

The second is transparency. The same government working paper already proposes that AI companies disclose what copyrighted material they use to train their models before they can operate commercially in India. That requirement could also include interaction data. At the very least, it would reveal how much value companies are creating from the conversations that millions of people have with AI every day.

The third proposal is the simplest. AI companies could make the choice to keep conversations out of model training the default instead of requiring users to search through settings to opt out. That change would cost almost nothing. It is a design decision, not a technical challenge.

None of these ideas requires banning AI products. Each of them asks for something much smaller. They ask us to treat an ordinary person’s data with the same institutional respect that we already give to publishers, creators, and enterprise customers.

The discussion about India’s AI future is often framed as a race to build the next breakthrough model. But there is a more immediate question, and it has received far less attention. Who owns the value created from the hundred million unpaid, deeply personal conversations that Indians have with AI every week?

One fact sits uneasily beside that question. This year’s budget gives foreign cloud and AI companies a corporate tax holiday until 2047 if they operate through data centres built in India. For the next two decades, companies can process data on Indian soil while paying no corporate tax under that policy. At the same time, millions of Indians continue to provide the raw material that helps improve these systems without receiving anything in return.

Late at night, in a flat in Kanpur, a home in Kochi, or a paying-guest room in Gurugram, someone is still typing into a chatbot. The conversation helps the model become a little better at understanding the next person. The exchange continues. The product improves. And the person who made it possible never receives an invoice, or a payment.

For most of my life, biotechnology meant one thing: pharmaceuticals. It was sterile labs, ten-year drug discovery cycles, and a pill you swallow when things go wrong. And for decades, that definition was perfectly fine.

But the times and my knowledge as a student of bioengineering made this definition obsolete.

We are standing on the threshold of something entirely new. A new technological revolution is quietly being built, one that treats biology not just as something to observe, but as something we can engineer, design, and command. This is the world of Synthetic Biology.

And over the next few decades, it isn’t just going to change medicine. It is going to rewrite manufacturing, agriculture, food, and climate technology from the ground up. As Freeman Dyson said “the future will be made of biology.”

What is Synthetic Biology?

To explain what synthetic biology is, let me use a metaphor - If artificial intelligence is how we program machines, synthetic biology is how we program life itself. This sounds like a sci-fi movie but is becoming increasingly real by the day.

For most of human history, biology was purely observational. We watched how diseases spread. We studied how crops grew. We mapped how organisms evolved. We adapted to nature’s constraints because we had no other choice. Synthetic Biology completely changes this equation.

At its core, it combines biology with engineering and computation. Scientists can design and modify cells and microorganisms to perform highly specific tasks, instead of merely studying them. A bacterium can be programmed to produce a chemical, yeast can be engineered to manufacture proteins, and cells can even be designed to detect diseases and respond to them.

Now, take that exact same flexibility - the ability to write genetic instructions and have a cell execute them - and apply it to the plastic in your laptop, the fuel in your car, the fertilizer in your soil, and the protein on your dinner plate.

Just as software code tells a computer what to do, DNA tells that to a cell. It tells the cell what proteins to produce, when to turn certain behaviors on or off, and how to develop.Once scientists started treating DNA as an informational language rather than just a biological blueprint, the workflow started looking exactly like software engineering. You read the genetic code. You edit it. You test the output. You debug the errors. You repeat.

Erskine's super soldier serum was pure fantasy - you can't 'unlock' a genome like a cheat code. But CRISPR is the real, boring, unglamorous version of the same fantasy. CRISPR is a gene-editing technology that works like a pair of molecular scissors, allowing scientists to precisely cut and rewrite specific sections of DNA inside living cells. It can be used to correct disease-causing mutations, add or remove genes, engineer crops and microorganisms, and even create new biological functions.

Thanks to gene-editing technologies like CRISPR, the precision of this process has skyrocketed. According to Jennifer Doudna, co creator of CRISPR, “CRISPR is not just a tool for editing genomes; it’s a platform for rewriting the future of biology.”

At the same time, the cost of DNA sequencing and synthesis has plummeted. What once took years and billions of dollars now takes a few days and a few hundred bucks. And when a technology becomes this cheap and accessible, it stops being a niche laboratory tool. It becomes an industrial platform.

From Medicine to Manufacturing: Biology as an Industrial Platform

Healthcare is one of the clearest examples of synthetic biology’s potential. You remember the rollout of the mRNA vaccines, right? For a few chaotic months, the entire world’s ability to return to normal hinged on a biological code wrapped in a lipid nanoparticle. It felt like a medical miracle. It wasn’t. It was programmable biology, experienced in a pandemic. Cap needed a lab and a war to get his upgrade, the rest of us just needed a lipid nanoparticle and a cold chain.

The stocks of the companies producing the vaccines like Moderna and BioNTech shot up during the pandemic, showing how significantly the world relied on the vaccines produced. That moment changed the perception of biotechnology entirely. Instead of treating biology as slow-moving and experimental, investors and governments began to recognize that biology could become a flexible technological platform, one that is capable of adapting rapidly to new problems. And medicine may only be the beginning.

Consider how we manufacture things today. Almost everything in the modern industrial world - from cosmetics and textiles to industrial detergents - comes from petrochemicals. We dig up crude oil, refine it, and burn massive amounts of energy to turn it into products.

Synthetic biology offers a wildly different approach. What if you just programmed a microbe to produce that exact same specialty chemical? Instead of massive, polluting oil refineries, the factories of the future look like giant breweries. Think of microbes as microscopic factories. Once correct “machines” and “programmes” are set up, they can produce medicines, fuels, chemicals, food ingredients, and even advanced materials.

Reimagining Food, Materials, and the Physical World

Right now, to get a glass of milk or a piece of meat, it requires roughly 9 m² of land, 628 liters of water, and generates about 3.15 kg of CO2-equivalent emissions - and at global scale, dairy cattle alone graze around 830 million hectares, an area about the size of Brazil. And dairy is just one piece of a much bigger imbalance. Livestock - meat and dairy combined - uses 77% of the world’s agricultural land, yet supplies only 18% of global calories and 37% of global protein. As the global population grows, the math on this simply stops working.

Enter precision fermentation. Companies are already engineering yeast and bacteria to produce the exact same proteins found in dairy and meat. This is not a plant-based imitation. It is the literal, molecularly identical protein, just brewed in a vat instead of grown in a cow. Pandora’s avatars were grown in a vat to do a job no human body could. Precision fermentation is doing the same thing to yeast - minus the blue skin and the box office.

Synthetic biology is also beginning to influence materials science. Researchers are exploring biological systems capable of producing spider silk-inspired fibers, biodegradable packaging, self-healing materials, and even living construction materials that can respond to environmental changes.

The broader trend is becoming increasingly clear: biology is evolving into a manufacturing platform. And as that happens, the distinction between biology, software, and industrial production is starting to blur.

This is why, as the field of synthetic biology is maturing, a new breed of companies is emerging. They aren’t trying to build a single drug or a single food product, but rather the underlying infrastructure itself. Companies like Ginkgo Bioworks, the “AWS of biology”, provide the automated labs, the robotics, and the biological design tools so that other industries can build on top of them.

India’s Moment in the Bio Revolution

Just like every other field in the world that deals with data is catching on to the hot topic of AI, so is the field of synthetic biology.

Engineering biology requires testing millions of genetic variations. This is where Artificial Intelligence enters the chat. AI predicts protein structures; synthetic biology builds them. AI processes the digital; synthetic biology manufactures the physical.

And here is the interesting part for India.

India currently sits at an incredibly unique intersection. During the 1990s and 2000s, the country became a global powerhouse because it scaled software engineering talent. Today, it already boasts massive pharmaceutical manufacturing capacity, a deep bench of biotech talent, and millions of software engineers.

As biology becomes increasingly computational, that combination of IT expertise and biotech infrastructure is exactly what is needed to scale bio-manufacturing. And we see start-ups in the country catching on to this.

String Bio, currently valued at ₹1,810Cr, is developing methane-based biological production systems while Sea6 Energy, which raised $18.5 million in Series B led by BASF Venture Capital in 2022, is building seaweed-based biomaterials and sustainable industrial products. In 2024 the Swiss non-profit Global Antibiotic Research & Development Partnership (GARDP) offered up to $20 million in funding and technical support to Bugworks Research Inc. to develop an innovative broad-spectrum antibiotic compound aimed at combating multidrug-resistant bacteria responsible for life-threatening infections. Bengaluru-based biotech firm Pandorum Technologies, which specializes in programmable regenerative medicine, secured $18 million in a Series B round led by Protons Corporate, funding to support its clinical development efforts and drive international expansion. These companies could be the global torchbearers of India’s emerging deep-science based biotechnology innovation community.

The larger opportunity, however, may extend beyond individual startups. The real opportunity is structural. India became a global IT powerhouse because it combined elite engineering talent with scale and cost advantages. Synthetic biology is handing us the exact same playbook. If we invest aggressively in research infrastructure, bio-manufacturing, and computational biology, we don’t just get to participate in this revolution. We get to host it.

The question is no longer whether we can program life. We can. The question is who will build the infrastructure to manufacture the future. Because the economy of the next few decades won’t just run on silicon chips and imported oil. It is going to run on engineered cells. This is the new bio-economy.

Yesterday I watched a 90 minute podcast where Nikhil Kamath talked about The Foundery. I then spent the next 90 minutes wondering why I have so much free time to watch something like that, just to conclude that I could turn it into an article and call the day productive.

So Foundery. What initially feels like three kind and generous people with capitalistic success and ample free time coming together to “remove the struggle” from entrepreneurship. Let’s put them under a microscope, shall we?

Nikhil Kamath, Kishore Biyani and Ronnie Screwvala launched The Foundery in December as a 90 day residential co-founder factory, where they invite people based on the selection criteria and present them with ideas, network, legal help, brand design and everything that they would need to build India’s next unicorn. Sounds fun right?

During the podcast Ronnie was asked whether he thinks the foundery is selfish? Not much to our surprise he replied that it isn’t, because even if 100,000 people want to participate in foundery that would mean they’re doing nearly 100,000 wake up calls before Nikhil changed the topic. What I understand from this is people of India need a wake-up call as they are not aware of the fact that they can start a startup. But what does this wake-up call look like?

T&C Foundery (For some reason the capital texts feel like they are screaming)

The fee for registration for foundery is 5000Rs/-, the cohort 01 had 36K applicants, meaning foundery made 18 Crores before building a single company. It’s always funny how Indian billionaire’s associate negative notion to being selfish. It’s okay to be selfish, everyone is selfish, it’s worse to be dishonest.

And 5000Rs/- registration fee is not even the start.

The contract is containment

Here is where it falls apart. Applicants pay a nonrefundable five thousand rupees plus GST just to be considered. Selected participants get up to 25 percent equity in a company that The Foundery itself designed, meaning the institution retains up to 75 percent of something built almost entirely by the person taking on the personal risk.

Run the actual math and the picture gets worse. A 25 percent stake for four crore rupees prices the entire company at roughly 16 crore, or under two million dollars. Y Combinator takes 7 percent for 500 thousand dollars, which prices the company at over seven million dollars. Techstars takes 5 percent for 220 thousand dollars, pricing the company at 4.4 million dollars. The Foundery writes a bigger check in absolute terms than Techstars and still values the resulting company at less than a third of what Techstars implies for the same stage, an idea and a founder, nothing else.

That gap is not a rounding error. It is a statement. It tells you what The Foundery thinks an Indian founder’s company is worth on day one, before the founder has built anything, before there is any evidence either way. Y Combinator and Techstars are betting on unproven founders and still pricing them at seven figures and up. The Foundery is betting on unproven founders and pricing them like the ceiling is already known and the ceiling is low. Nobody in the accelerator world takes three quarters of the company for a four crore check, and nobody prices a fresh Indian startup at a tenth of what the same stage company would be priced at in San Francisco or Boulder.

The equity split is the least alarming part. The terms and conditions define a fixed cost per participant, reportedly around 28 lakh rupees, that The Foundery can invoice you for if you breach the agreement by leaving early or failing to complete the program. IP generated during and after the program is assigned to their affiliate, not to you. Gag clauses restrict what participants can say about the experience afterward. Rights over a participant’s image and story extend indefinitely. Selection criteria, schedules and the terms themselves can be changed unilaterally at any time. Decisions are described as final and non contestable.

Walking away from The Foundery does not mean losing unvested equity, which is the normal cost of an early exit anywhere else. It means potentially owing money, forfeiting the idea, and being unable to talk about any of it afterward.

What The Foundery actually thinks founders are

When Kamath announced the 75/25 split, Twitter reacted the way you would expect. People said no serious founder would choose this structure. People said it defeats the purpose of building something of your own. The criticism was loud, but it mostly stopped at the number. The number is not the real story. The real story is what the number reveals about how The Foundery sees the people walking through its doors.

Most accelerators are built on a premise, whether they say it out loud or not, that founders are already special, or at least let the founder believe they are special and capable of building something special. YC and Techstars select for people who have shown some evidence of unusual ability, an idea already in motion, a prototype, early traction, something. The accelerator’s job is to accelerate a person who already has the raw material.

The Foundery’s own language says something different. Their website states it plainly: “Most startup ecosystems ask whether someone is ready to be a founder. We believe the right environment forges that readiness.” Read that sentence carefully. It is not saying founders are rare and they find the rare ones. It is saying readiness is manufactured, not discovered. Anyone sufficiently willing to take the risk can be forged into a founder, provided the right environment does the forging.

That is not a description of a talent search. It is a description of a training pipeline. The Foundery does not need someone with a proven idea or a track record. It needs someone with the raw material of a functioning brain, willing to be run through a 90 day process that supplies the idea, the brand, the packaging, the legal structure and the go to market plan. The founder’s job in this model is closer to execution than authorship.

This explains a detail that would otherwise look strange. Why build an entire residential program, fly people to a villa in Alibaug, run a 90 day bootcamp, when the same capital could just be handed to founders who already have a company started? The answer is that The Foundery is not optimizing for the best idea. It is optimizing for a trainable person executing a pre-approved consumer thesis, and the entire apparatus, the vault of ideas, the villa, the 90 days, exists to manufacture that person and the product around them.

Notice also that this model only makes sense for consumer brands. Nobody builds a residential program with brand design and packaging support to launch a B2B SaaS tool. The consumer focus is not incidental. Consumer brands are the only category where this kind of manufactured, story-driven, content-friendly founder actually has something to sell.

The real business model is the marketing

There is a simpler way to say all of this. The villa, the 90 days, the shortlisted cohort, the constant press around Alibaug, none of that is really about building a company. It is a marketing stunt, and the brand is the byproduct.

Think about what a launch actually needs in India’s consumer market. It needs attention before it needs product-market fit. A new soap or snack or skincare brand competing against Hindustan Unilever or ITC does not die because the formulation is bad. It dies because nobody hears about it. The Foundery solves that problem before the brand even exists, by turning the founding process itself into content. Eighty two people flown to a villa, a residential bootcamp, Kamath and Screwvala’s names attached to every headline, is free distribution that a normal first-time founder would spend years and crores trying to buy. By the time a Foundery brand actually launches, it already has a media narrative most startups only get after they succeed.

That is the actual business model hiding underneath the accelerator language. Manufacture attention through the spectacle of the program, let that attention give the resulting brand a running start with consumers, and if the sales numbers come in, sell the company to one of the very goliaths it was positioned against. The irony is that the entire narrative arc is designed to end in acquisition by the incumbent, not in competing with it long term. A billion dollar independent consumer company takes a decade of reinvestment and founder conviction. A quick trade sale on the back of a good first year of sales takes neither, and it is the far more likely outcome given who holds 75 percent and who gets to decide when the offer is good enough.

So the consumer focus is not just about which category is easiest to manufacture a founder for. It is about which category is easiest to manufacture a media moment for, and easiest to flip once that media moment converts into revenue.

The Ronnie Screwvala defense, and why it collapses

When Screwvala was asked about the 75/25 split, his response was that it is not about the value assigned to you, it is about the value you create. If the founder builds a billion dollar company, 25 percent of a billion dollars is still 250 million dollars. He also argued The Foundery is saving founders two to three years of their life, because a founder raising capital through multiple funding rounds typically ends up with somewhere between 25 and 40 percent of the company by the time it has any real value. That part is true. Most founders do end up heavily diluted by the time a company is actually worth something.

But this is exactly where the contradiction shows up. The Foundery’s own pitch sells the idea that a participant can walk out of a 90 day program having built something capable of reaching a billion dollar valuation, and this is why I would never apply to be part of The Foundery. Because this is the lie that they are selling.

Why can The Foundery never actually produce a billion dollar company? There are two reasons.

The first is that a larger competitor might simply copy the model. Before you say there are things like patents and IP protection that stop this, remember the founder does not hold the patent here. The Foundery’s terms assign IP generated during and after the program to their affiliate. So the question becomes: why would anyone building something genuinely patent worthy, something capable of becoming a billion dollar company, sign away rights to that IP for 25 percent of the outcome in the first place. The people with ideas strong enough to be worth copying are exactly the people who have the least reason to give up ownership of them. And if The Foundery’s own idea vault contained something patent worthy, there would be no reason for them to give away 25 percent of it to a random founder either.

The second is that a larger player acquires the brand outright. In a normal startup, the reason acquisition talks are slow and founders retain leverage is that ownership is spread across many parties, the founder, multiple investors across multiple rounds, employees with option pools. Getting everyone to agree to sell takes time, and that friction protects the company from being swallowed cheaply. The Foundery removes that friction entirely. You do not need to convince a founder to give up their dream. You only need to convince The Foundery to take a quick payout, because The Foundery holds 75 percent and can approve a sale on its own terms. The founder is a minority voice in a decision about the thing they built.

I cannot guarantee this will happen. But neither can the founder :)

That is the actual argument here, not that a sale is certain, but that the founder has no say in whether it happens. Run the scenario forward anyway. The brand succeeds and starts winning share from the existing consumer goliaths, the Hindustan Unilevers and the ITCs of the category. At that point the decision about what happens next, sell, hold, expand, belongs to whoever holds control, and that is The Foundery, not the person who built the thing. The founder does not get a vote. They get informed. Whether The Foundery chooses to cash out early, or copying eats into the brand’s edge because the founder never owned the IP that would have stopped it, the outcome is decided somewhere the founder is not in the room. And if the exit comes at a moment the founder disagrees with, walking away is not an option either, since the contract defines leaving as a breach that can trigger the clawback and forfeits any claim to the idea.

Why it should be no other way

So why should the equity structure be any different? The answer loops back to where this argument started. If The Foundery believed founders were something special, individuals selected for their intellect, their conviction, their irreplaceable read on a market, it would have every reason to protect that person’s ownership and give them real leverage over what happens to the thing they built. Companies protect what they consider rare. But The Foundery’s own language tells you it does not see founders that way. It sees them as raw material that the right environment forges into readiness. You do not give controlling ownership, patent rights or a seat in the acquisition conversation to raw material. You give those things to people you consider irreplaceable, and The Foundery has already told you, in its own words, that it does not consider its founders irreplaceable. It considers them trainable.

That is the through-line, the 75/25 split is not a harsh number bolted onto an otherwise founder-friendly program. It is the natural conclusion of a philosophy that starts by denying the founder is special in the first place. Once you accept that premise, the contract stops looking unfair and starts looking consistent. The unfairness was never in the paperwork. It was in the sentence on their homepage, the one that told you exactly how they see the person signing it.

-By Amogh.

Dario Amodei wrote an excellent essay titled Machines of Loving Grace in which he goes on to paint a beautiful world 5-10 years after the powerful AI has been built. He describes a powerful AI as a country of geniuses in a data centre, which is smarter than a Nobel Prize winner across most relevant fields, such as biology, programming, and math. He expects that such an AI will help us speed up research across all the relevant domains by at least 10X, i.e., we can have 100 years of progress in the next 5-10 years after such an AI is built.

Witnessing the pace of the technology being built, we can safely assume that these outcomes have a high probability of occurring.

This is also because the optimism in Machines of Loving Grace is not without historical precedence. The 20th century itself demonstrated how rapidly scientific progress can transform human civilization: global life expectancy rose from nearly 32 years in 1900 to over 70 years today, while diseases like smallpox were eradicated and deaths from infectious diseases collapsed due to vaccines, antibiotics, sanitation, and modern medicine. What makes this remarkable is that these breakthroughs emerged in a world without modern computing or AI. If humanity could achieve such dramatic improvements with comparatively primitive tools, it is plausible that AI systems capable of accelerating research across biology, medicine, mathematics, and engineering could compress decades of progress into just a few years.

However, and this is where the real issue comes, the picture presented in the essay rests comfortably on the pillars of delusion and hope. Which, sadly enough, has little to no precedence in history.

Coordination: A battle we lose repeatedly

A large part of Amodei’s vision implicitly assumes unprecedented cooperation between corporations, governments, and nation states.

Yet history offers very few examples of such long-term global coordination succeeding. The closest modern precedent, the United Nations, began as humanity’s grand institutional promise after the horrors of World War II, only to gradually descend into paralysis, veto politics, and geopolitical theatre. The fundamental problem is not technological capability, but human incentives.

Nevertheless, apart from the meme above, game theory explains this elegantly through the Stag Hunt equilibrium. In the game, two hunters can either cooperate to hunt a stag, yielding enormous rewards for both, or individually chase rabbits for smaller but guaranteed gains. Catching the stag requires mutual trust; chasing rabbits requires only self-interest.

Historically, nation states repeatedly choose the rabbits. Fear, mistrust, and the possibility of betrayal overpower the fragile faith required for long-term cooperation. Even when collective outcomes are obviously superior, actors defect because the cost of being the only cooperator is catastrophic. AI development today increasingly resembles this equilibrium: every nation knows cooperation is safer, yet every nation fears being left behind.

Geography of Mistrust

Ironically, the very country from which this optimism emerges, the United States, has historically been among the greatest destroyers of the trust necessary for such global cooperation. From Cold War interventions and regime change operations to unilateral wars, sanctions, surveillance programs, and technological monopolisation, American power has often signalled to the rest of the world that dependence is dangerous and strategic autonomy is survival. Result? A fragmented world optimising not for collective flourishing, but for geopolitical leverage.

Thus, the unrelenting coordination that Amodei hopes for to reap the rewards of this technology has proven to be extremely hard for us to achieve and hence, shifts some percentage points of probability from a utopian world to a dystopian one.

Who determines outcomes?

More importantly, history shows that the trajectory of civilization has rarely been determined purely by scientific capability. Politics and economics almost always overpower scientific possibility.

Carl Sagan repeatedly emphasises this in his book Cosmos. He describes how the Library of Alexandria was destroyed not because science failed, but because political and civilizational priorities shifted away from preserving inquiry.

Ancient Greek thinkers like Aristarchus had already proposed heliocentrism nearly 1800 years before Copernicus. The idea vanished because it violated the Aristotelian cosmology on which Greek philosophical and religious authority was built.

Hero of Alexandria built primitive steam-powered devices in Roman Egypt, but they remained curiosities rather than sparking an industrial revolution because the Roman economy relied on slave labour, eliminating incentives for mechanisation.

India offers an almost identical pattern. Indian metallurgists were producing crucible “wootz” steel of a quality Europe could not replicate until the Industrial Revolution. The famed Damascus blades of the medieval world were forged from Indian ingots, traded west along the same routes as spices and silk. India held the technological edge in steelmaking for centuries. It did not produce an Indian industrial revolution because the surrounding economic structure, fragmented polities, and no capital markets built around scaling manufacture never converted the knowledge into industry.

Sometimes the suppression is not economic neglect but a direct, deliberate veto. For over two centuries, from 1485 to 1727, the Ottoman Empire banned the printing press for Arabic and Turkish script. The technology was not a mystery to them, as the Jewish and Armenian communities within the empire operated presses freely. The ban existed because the scribal guilds and religious establishment correctly recognised that movable type would dissolve their monopoly over the written word. While Europe’s presses multiplied literacy, scientific publishing, and the dissent that fed the Reformation, the Ottoman world’s intellectual output was throttled for generations by people who understood the technology perfectly well and feared exactly what it would do to their power.

Chinese admiral Zheng He commanded fleets vastly superior to European ships decades before the Age of Exploration, yet imperial political decisions abruptly ended Chinese maritime expansion, altering the balance of global power for centuries. Scientific potential existed in all these civilisations; what determined history was not knowledge alone, but economics, incentives, institutions, and power.

That is the uncomfortable weakness in visions like Machines of Loving Grace. They assume that once humanity possesses sufficiently advanced intelligence, civilization will naturally organize itself around maximizing collective human flourishing. History suggests otherwise. Intelligence has rarely been the bottleneck. Coordination has.

What next?

There are some examples in which humanity came to a consensus. The Declaration of Human Rights is one such idea, and nuclear non-proliferation is another. Arguments can be made about the actual on-ground implementation and adoption of these across the countries; however, we would be better off in a world with them than without them. The real problem is that even these universally accepted declarations came after decades or even centuries of misery and individual events that shook the world.

We are standing in 2026 with all the knowledge and the learnings from these events beneath. What we face is the advent of another technology, which will have asymmetric effects on nearly all aspects of our lives. While we survived poor coordination in the Bronze Age, it can be civilizationally terminal in the age of AGI. Thus, the two questions that stand before us are:

1. Can we make our institutions move at a pace that at least try to match the technological advancements?

2. Can we trust each other like never before?

Nevertheless, we may soon possess the intelligence of gods while retaining the coordination mechanisms of tribes!

December 2022 was the time when work became easy, and ChatGPT first entered our lives. I was in college, and we had just begun exploring the tool. I remembered asking my friend, “Is this available on the Play Store?” (Believe me, it was just an online tool back then).

But soon newspapers started talking about a new invention that interacts with you. Experts described it as something fundamentally different and never seen before. Back then, even having the tool complete simple tasks such as writing a story or presenting structured information felt extraordinary.

Fast forward to today, and we have 2 giants, namely OpenAI and Anthropic, fighting a battle to extend the boundaries of what artificial intelligence can do to unimaginable extents. Besides them, we have xAI’s Grok, Google’s Gemini, Microsoft’s Copilot, Meta AI, Perplexity and DeepSeek, etc. Today, the conversations revolve around the tasks that AI can fully automate, the skills that might become redundant, and how quickly we will reach the singularity, i.e., the point where AI surpasses our intelligence.

How everyday AI updates feel nowadays!

It is a well-established fact that AI is the new buzzword in town, and seldom does a day pass that we do not hear something new and sometimes unsettling about AI. While on the brighter side, we have firms leveraging AI models to assist people in various professions, such as Harvey AI, a platform for lawyers and legal teams and Sierra AI, a customer service agent for enterprises, but on the other side, we have uncomfortable discussions around energy requirements for training these AI agents popping up.

Amidst all this global frenzy, it becomes pertinent to ask where India stands in the race to become AI Vishwaguru.

Disclaimer: AI here is not America-India or Atmanirbhar India; it is Artificial Intelligence!

2. India’s AI Boom

India’s AI startup ecosystem has expanded rapidly since 2022. Bengaluru, Hyderabad, Pune, Gurugram, and Chennai have emerged as major AI hubs. India has around 1.8 lakh startups, and nearly 89% of new startups launched last year used AI in their products or services. Rumours started to spread that investors are asking just 2 things in the pitch: 1. Do you use AI? 2. Any foreign example? And each Yes accounted for $5 million and this look:

“This looked genuine to me”

Recently, companies such as Krutrim, Sarvam AI, Gupshup, Yellow.ai, CoRover, Mad Street Den, Uniphore, and Haptik have gained significant attention. Analysts say AI could add $1.7 trillion to India’s economy by 2035, when asked where? They go quiet.

The picture looks rosy from the top, with government investments flowing in, startups seeking to address challenges unique to the domestic market and large companies rapidly integrating AI in their workflows. However, lately, concerns have started to emerge about how much of its AI stack India actually owns.

3. Inside India’s AI Startups

An AI stack refers to the complete ecosystem powering artificial intelligence, including chips, data, cloud infrastructure, foundational models, tools, and applications built for end users. An analysis of 103 Indian AI startups revealed key findings regarding overall stack ownership and AI sovereignty. For ease of understanding, the complete AI stack can be broken down into 5 vertically integrated layers, as shown in Figure 1.

While nearly 92 of the 103 startups were found to be dependent on foreign cloud, 20 of those 103 are directly reliant on OpenAI’s proprietary architecture via API access.

Besides that, it is a well-known fact that layer 2 is completely dominated by NVIDIA, and even layer 1, which consists of data centres, is dominated by major foreign players such as Japan-based NTT Global Data Centres and Singapore-based STT GDC.

To better understand how India is currently adopting AI, we categorised Indian AI startups across different layers of the AI stack, from simple API wrappers to foundational model and infrastructure companies, revealing where India’s true strengths and dependencies lie.

According to the analysis, almost 88% of the startups were wrappers and fine-tuners, i.e., companies that primarily relied on existing foreign AI models and infrastructure rather than building foundational models or core AI infrastructure from scratch.

The findings suggest that while we have been quick to adopt AI at a large scale and augment it to solve problems that are specific to our needs, the companies building LLMs from scratch and owning their GPUs and Cloud infrastructure remain critically small.

This brings us to our next critical data point, i.e., the percentage of companies dependent on foreign cloud infrastructure.

Now, before looking at the data, it helps to understand what cloud infrastructure actually means here. At its simplest, it is the heavy machinery of the internet. Instead of buying physical servers for your office, you rent computing power and storage over the web. Training modern AI models requires massive computational power. Purchasing the specialised hardware needed to process these workloads is expensive for an early-stage company. Cloud infrastructure eliminates this structural barrier by converting high-performance computing into an on-demand utility. However, here comes the catch: If a country relies entirely on foreign infrastructure, the independence of its software becomes secondary. Someone else ultimately controls the cost margins and dictates the operational terms.

While the figure above breaks down the dependency on various cloud providers, the figure below shows a comparison of ownership at 2 fundamental levels, i.e., the model and infrastructure layers, in which infrastructure is further broken down into cloud dependency and GPU ownership.

Figure 4: Ownership at 3 fundamental levels

In an increasingly geopolitically unstable world, this doesn’t only remain an infrastructure gap; instead, it risks becoming an issue of sovereignty, where a slight change in the modus operandi of these cloud and GPU providers can cause a major dent in our entire AI startup ecosystem.

At last, to sum it up, we calculate the dependency scores of these startups and rank them accordingly, as shown below.

This figure is important for our understanding of the overall picture. The startups having medium to low ownership stand at 22%, which is not a bad figure for a developing country like India. However, when viewed with the other 4 figures above, it also clearly shows and adds to the point that our performance in this global AI race is severely impacted not by the lack of sovereign AI models but by the acute shortage of sovereign AI infrastructure.

4. The Geopolitics of Compute and why ownership matters

Building our own AI infrastructure is significant because artificial intelligence is not merely another technology product anymore; it is gradually becoming foundational infrastructure for economies, governments, militaries, education systems, healthcare, and information flows. Countries that control the AI stack will increasingly control the next phase of economic and strategic power.

For a country like India, this becomes even more important for several reasons.

First, dependence on foreign AI systems creates technological vulnerability. If India relies entirely on models, chips, and cloud infrastructure owned by foreign corporations, critical sectors of the economy may become dependent on decisions taken outside the country. Just as nations once worried about dependence on imported oil or foreign operating systems, dependence on external AI infrastructure can create long-term strategic risks.

Second, AI systems are deeply shaped by the data and societies on which they are trained. Most leading AI models today are primarily optimised for Western languages, cultures, legal systems, and user behaviour. India is linguistically, culturally, and economically very different. Without indigenous AI systems, large sections of Indian realities may remain poorly represented in future technologies. Owning the AI stack allows India to build models suited to our unique characteristics. While startups like Sarvam AI have paved the way, they, too, suffer from the unavailability of quality Indian cloud infrastructure that can support their model.

Third, economic value in technology usually concentrates at the deepest layers of the stack. Countries that only consume foreign technology often capture limited value, while countries that own infrastructure, platforms, and intellectual property capture disproportionate wealth. The United States dominates today’s digital economy largely because it owns critical technological layers such as semiconductors, cloud computing and operating systems.

If India remains confined only to the application or wrapper layer, much of the long-term economic value may continue flowing abroad.

Fourth, AI is increasingly linked to national security. Advanced AI systems are already being integrated into cybersecurity, defence intelligence, autonomous systems and surveillance

No major country would want such systems to remain entirely dependent on foreign entities.

Fifth, ownership of the AI stack influences future employment and industrial competitiveness. AI is likely to reshape industries ranging from software and finance to manufacturing and education. If India merely imports AI tools developed elsewhere, domestic firms may become users rather than creators of the next technological revolution.

Finally, there is a deeper historical dimension. India missed the earlier industrial revolutions in manufacturing and semiconductor ecosystems. The AI revolution presents a rare opportunity where the technological landscape is still evolving. Building indigenous capabilities now may determine whether India becomes a creator of future technologies or merely a large market consuming technologies built elsewhere.

The post-1990s software boom has provided India with a credible workforce that can not only integrate AI into its workflows to increase productivity, but also build foundational AI models if provided the right resources and utilised efficiently. The country’s core AI workforce is on track to cross 1.25 million by 2027. Hence, India stands at a critical juncture again in history. Earlier, we missed the manufacturing train and saw our demographic dividend stuck in agriculture; next, we sat at the driving seat of the software boom, but failed to provide infrastructure for innovation, which resulted in our best talent moving abroad. Now we have a chance again, either to build infrastructure rapidly to absorb the growing talent pool or watch them leave the country again to build for someone else.

The debate around AI is not simply about technology. It is fundamentally about economic power, strategic autonomy, cultural representation, and long-term national competitiveness.

I was two hours into Interstellar last weekend, the wormhole scene, the one where humanity escapes a dying Earth by slipping through a hole in spacetime punched open by beings who have transcended three dimensions, when it hit me. Not the profundity of the film. Something much more uncomfortable.

Bollywood doesn’t make movies like this. And before you reach for the obvious explanations, budget, technical know-how, the right director, let me stop you. We have all of those. S.S. Rajamouli built a mythological epic that grossed over $120 million worldwide on pure spectacle.

So, What is the Real Problem?

The real problem is imagination. Even if someone in Bollywood had a genuinely good sci-fi story, Indian spaceships, first contact, humanity’s survival hinging on a DRDO scientist from Pune, no Indian would buy it. Nobody sitting in the theatre is going to accept, in their gut, that India is out there conquering the cosmos. Our entire cultural ceiling for space is Jadoo from Koi Mil Gaya who glows when he’s happy and teaches a small-town boy to play basketball. We do have Mission Mangal, but it plays closer to an ISRO press release than a science-fiction film.

That is about to change, big time.

So What Exactly Is ISRO Planning?

Look at the timeline. Sounding rockets in the 1960s. SLV and ASLV through the 80s. PSLV by the 90s, the workhorse that still launches most of India’s satellites today. GSLV and Mangalyaan in the 2010s, the Mars orbiter that cost less than a Hollywood VFX budget. Chandrayaan 3’s Moon landing in 2023, the first time any country touched down near the lunar south pole. And now the next fifteen years: Chandrayaan 4 and Gaganyaan’s crewed mission by 2027, the Bharatiya Antariksh Station by 2035, and a crewed Moon landing penciled in for 2040.

This is not a single mission. It is a staircase, and Gaganyaan is the step everyone is currently arguing about, because it is the most expensive one and the hardest to justify on paper.

So let’s argue about it properly.

Gaganyaan’s expanded budget now stands at roughly 20,193 crore rupees, about 2.1 billion dollars, covering eight missions through 2028 including the crewed flight itself. That is an enormous number for a country still building toilets and last mile broadband in half its villages. And the skeptic’s case writes itself, because it is a good case. We already landed on the Moon. Chandrayaan 3 did it in 2023, with a rover, with real instruments, with real science returned, and it did all of that without risking a single human life. Robots don’t need oxygen, food, a return trajectory, or a hero’s welcome home. If the actual goal is research, robots already solved it, cheaper and safer.

So what exactly are we trying to prove? Is twenty thousand crore rupees just pride wearing a flight suit, a chest thumping exercise to say we did it too, in a decade where Elon Musk is literally catching falling rockets out of the sky and NASA is going back to the Moon for keeps? Are we entering a competition nobody actually invited us to, purely so we can say we showed up?

The honest answer is yes, and also no. Yes, there is an element of prestige here that no ISRO press release will ever admit to out loud. But no, prestige is not the actual reason this is worth doing. The actual reason sits one level up, in the bigger picture that the Moon landing headline conveniently distracts you from.

The Bigger Picture

Here is the thing about Mars that doesn’t get said enough. It is closer than we think, and somebody is already building the rockets to get there. Elon Musk has said for years that Starlink and the satellite business are not the point, they are the cash machine that funds the actual point, which is Mars.

At the same time, NASA’s return to the Moon is not a victory lap either. The Artemis programme is explicitly designed to build permanent infrastructure on the lunar surface, with a first crewed Moon landing now targeted for 2028 and missions planned roughly once a year after that, as a staging ground for eventual human missions to Mars.

Here’s the rewritten section, copyable, no em-dashes:

Now. Is Mars actually worth it?

Step back for a moment, because this is the honest question nobody wants to answer out loud. We don’t know. We genuinely cannot pretend we do. For all we know, going to Mars might be the only way to solve India’s population crisis. But for now, let us look at it as a bet. What I hope is that the Mars race plays out like the semiconductor race or the internet race: whoever builds the infrastructure first shapes the economic and geopolitical order that follows for generations. Pride and profit in the same package. But even if this doesn’t turn out like that, it is still a very hedged bet.

What do I mean?

Whatever research ISRO gains from sending people to space will be transferred to the private sector. In June 2020, the Indian government approved private enterprises across all phases of space activities and simultaneously established IN-SPACe to help private players become independent actors rather than merely vendors to ISRO. A memorandum of understanding was signed between the Department of Space and NewSpace India Limited on October 12, 2020, specifically to identify technologies developed by ISRO and transfer them to private entities, academia, and other organisations. Under this agreement, NSIL transferred over 363 technologies for commercialisation within the first year alone. Since 2020, ISRO, IN-SPACe, and NSIL have collectively signed 75 technology transfer agreements with the private sector.

ISRO’s stance is clear: it focuses on advanced research and frontier missions. The private sector takes over operations, manufacturing, and commercial services. Knowledge flows downward. So even if the Mars bet never pays off, even if the crewed Moon landing turns out to be a very expensive flag-planting exercise, the capability and infrastructure built along the way gets handed off to a commercial ecosystem. Like, for instance, building the Indian space economy.

What does the Indian space economy look like?

Well. It looks a lot like Jio.

If you have followed Indian business for more than five minutes, you already know that Reliance has a habit of entering industries it was not invited to, spending aggressively until the economics of the entire sector rearrange themselves, and then looking mildly surprised when competitors complain. Jio did it to telecom in 2016. It is now attempting to do it to retail, to media, to financial services, and apparently, to low Earth orbit.

But before we get to Jio’s satellite ambitions, we need to understand what it is actually trying to compete with. Because the thing Jio looked at and decided it needed a response to is one of the most staggering commercial deployments in the history of any industry.

SpaceX launched its first 60 Starlink satellites in May 2019 on a single Falcon 9. What followed was a deployment campaign that made most industrial rollouts look leisurely. As of June 2026, the Starlink network consists of approximately 10,413 operational satellites in low Earth orbit, accounting for roughly 75 percent of all active maneuverable satellites on Earth. Starlink had more than 12 million subscribers as of June 2026, operates across approximately 160 countries, and by the end of 2025 had become SpaceX’s largest business segment, generating $11.4 billion in revenue. The constellation that analysts called economically unviable in 2018 is now the primary profit engine of a company valued at over a trillion dollars. The FCC has authorised SpaceX to deploy an additional 7,500 second-generation satellites, taking the total authorised number to 15,000.

Starlink has had its eye on India for a while. It applied for an operating licence here, went through multiple rounds of regulatory back and forth, signed a distribution agreement with Jio in March 2025 to resell Starlink services through Jio’s network, and was on the verge of launching commercial operations when India froze final approvals in June 2026. The reason for the freeze, officially, was security agency concerns linked to reported use of Starlink terminals during the Iran conflict without proper authorisation. The unstated reason, which anyone paying attention can read between the lines, is that New Delhi was increasingly uncomfortable with the idea of a foreign-owned private satellite network providing connectivity to Indian citizens during geopolitical flashpoints, with no domestic alternative in sight.

Enter Jio, which had been quietly watching all of this and apparently decided that the correct response was to build its own constellation.

At the Reliance Industries Annual General Meeting this year, Akash Ambani announced that Jio is evaluating a sovereign LEO satellite constellation of over 1,600 satellites at approximately 650 kilometres altitude, providing broadband and direct-to-device connectivity across India. A formal proposal has been submitted to IN-SPACe. The government is reportedly assisting with ITU orbital slot filings. Akash Ambani’s framing at the AGM was characteristically understated: Jio connected India on the ground, and now it would connect India from the skies.

Now here is where Skyroot comes back into the picture, and why the entire chain of logic connecting Gaganyaan to Jio to orbital data centres actually holds together.

Jio’s constellation needs to be launched. 1,600 satellites at 650 kilometres altitude. If India does not have domestic launch capacity capable of handling that volume at competitive cost, every single one of those satellites either rides on a SpaceX Falcon 9, an Arianespace vehicle, or a Blue Origin rocket. That money, the launch contracts, the manufacturing fees, the ground station integration costs, leaves India entirely. Estimates for building and deploying a constellation of this size run to $10 to $15 billion. The launch component alone, at international rates, would represent a significant outflow of capital for services that could, in principle, be provided domestically.

This is the straightforward case for Skyroot. Skyroot Aerospace, the Hyderabad-based startup founded by former ISRO engineers, became India’s first space unicorn after raising sixty million dollars in May 2026 from GIC and Sherpalo Ventures, with structured debt from BlackRock, pushing its valuation past one billion dollars. Its Vikram-1 rocket, built with 95 percent indigenous components and 3D printed engines that cut production time by 80 percent, is weeks away from India’s first private orbital launch. Vikram-2, with a cryogenic upper stage for heavier payloads, is in development for 2027.

Skyroot exists because ISRO’s infrastructure exists. It was seeded by the same 2020 policy reforms, built in the same ecosystem, and is launching from the same facilities at Sriharikota. Every bit of engineering rigour that Gaganyaan forces ISRO to develop, crew module hardening, escape systems, ground operations stress-testing, flows into the same facilities Skyroot uses for its own launches. And if Skyroot can eventually scale to the launch cadence that a Jio constellation requires, the capital that would otherwise have gone to Blue Origin stays in India.

And beyond satellites, the frontier that comes after connectivity is already visible.

In November 2025, a startup called Starcloud placed the first NVIDIA H100 GPU in space aboard a 60-kilogram satellite, then trained the first large language model in orbit in December 2025, running it on the complete works of Shakespeare. Three days before SpaceX completed its merger with xAI in February 2026, SpaceX filed an FCC application for up to one million orbital data centre satellites at altitudes between 500 and 2,000 kilometres, projecting that launching one million tonnes of satellites annually could generate 100 gigawatts of AI compute capacity. Google has announced Project Suncatcher, aiming to deploy its TPU chips into orbit by 2027, in partnership with Planet, a US-based space technology company. The European Space Agency is funding a 300 million euro programme studying orbital data centres for energy independence.

The reason this makes sense is not complicated. Data centres on Earth require enormous energy purely for cooling, and AI-related electricity consumption is projected to grow 50 percent annually through 2030. In orbit, the vacuum of space handles cooling passively, solar power is near-constant and free, and the only real constraint is the orbital slot you occupy. If the AI compute boom continues at its current pace, orbital data centres stop being a thought experiment sometime before 2035.

India is not watching from the sidelines on this one, which is either encouraging or a sign that everyone has collectively lost their minds, depending on your prior.

In May 2026, Pixxel, the Bengaluru-based planetary intelligence company, announced a partnership with Sarvam AI to develop and build India’s first orbital data centre satellite. Under the partnership, Pixxel will design, build, launch, and operate a satellite called Pathfinder, while Sarvam will provide the AI backbone, handling both training and inference directly in orbit, with full-stack language models running on board. The Pathfinder is a 200 kg-class satellite scheduled to reach orbit as early as Q4 2026. Unlike conventional satellites that use low-power processors optimised for survivability, Pathfinder will carry data centre-class GPUs, the same generation of hardware used in terrestrial data centres for AI workloads. The satellite will also include Pixxel’s hyperspectral imaging camera, allowing it to capture high-resolution spectral data and analyse it in orbit using AI models.

Both training and inference will happen directly in orbit, without any dependence on foreign cloud or ground infrastructure, creating a fully sovereign pipeline from observation to insight, end to end. Sarvam’s CEO Pratyush Kumar put it plainly: “AI infrastructure is not just a software question. It is a sovereignty question.”

The practical applications are not abstract. Sarvam says the satellite will be able to flag a wildfire, track a crop disease, or monitor a pipeline leak in the same pass it observes them. NVIDIA says orbital inference could reduce response time from hours to minutes. Most satellites today collect data, beam it to Earth, and wait for ground systems to process it. That gap between capture and insight is one of the biggest bottlenecks in Earth observation, and Pathfinder is specifically designed to close it.

Note what this actually is: an Indian AI company built on Indian capital, running Indian-built models, on an Indian-built satellite, processing data in orbit without routing anything through an American cloud. The entire stack, from the GPU to the model to the imaging sensor to the satellite bus, is domestic. That is not a press release talking point. That is a meaningful statement about what India’s space and AI ecosystems are now capable of building together, quietly, without waiting for permission.

India building cheap rockets, sharing infrastructure with private players, and keeping a human spaceflight programme alive is not about Mars in 2040. It is about not watching the entire orbit-as-infrastructure economy being assembled by people who started earlier, while India writes the cheques. Pathfinder is scheduled for Q4 2026. The cheque, in this case, stays home.

-By Amoghsiddhu.

Yesterday I was trying to stalk my ex’s Insta account. I looked at the recent posts - mostly individual and some with her female friends, whom I already knew. I looked at the account of one of her best friends to get a clue from there. Still, nothing was found. After half an hour, I gave up and went back to my doomscrolling session. I liked a few reels that reverberated with my broken heart and casually gave Zucky my emotional data. While scrolling, I found a few reels liked by her, which said, “Will he ever come back?” I jumped on my bed. This was the sign I was looking for, and the algorithm God provided me with the right data to give her a call in the evening.

Though later in the evening, she said, “I found it funny, so I liked it. Instagram does not give a laughing emoji reaction!” and then she cut the call. Both my phone and I went completely silent.

Can you connect the dots?

Now scale that Instagram intelligence operation: my deeply embarrassing half-hour of cross-referencing posts, mutual friends, and liked reels, by about ten million. Add satellite feeds, financial transactions, drone footage, phone intercepts, dark web chatter, and troop movements across multiple borders. Give all of that to a government trying to determine whether a hostile army is mobilising 4,000 kilometres away or whether a sleeper cell just went active inside a city of 20 million people.

The instinct is exactly the same:
Something doesn’t add up. I need more data. Let me connect the dots.

The difference is that I had Instagram’s algorithm and a ridiculous reason to use it. An intelligence analyst has eleven disconnected agency dashboards, a threat that cannot wait till morning, and a superior demanding a briefing in six hours.

The problem is never the lack of data. It is that the data exists everywhere, owned by everyone, but organised by no one. By the time it is reconciled across agencies, formats, and classification systems, the moment is already gone.

This is the problem Palantir was built to solve.

How One Company Productized God-Mode Intel

Founded in 2003 by Peter Thiel and Alex Karp, Palantir began with a simple observation after 9/11: the failure was not merely intelligence collection, but intelligence coordination. Different agencies possessed fragments of the picture, but the systems never spoke to each other fast enough.

Its first investor was the CIA, through In-Q-Tel, it’s VC arm.

The company took its name from Tolkien’s Palantiri, ancient seeing stones capable of observing events unfolding across vast distances in real time. Tolkien imagined them as instruments of power and perception. Thiel and Karp turned the idea into a defence company.

What Palantir built across platforms like Gotham, Foundry, and AIP is often lazily described as “surveillance software.” It is more accurately an operational intelligence layer: a system that ingests radically different streams of information: satellite imagery, SIGINT intercepts, financial records, sensor feeds, field reports, and maps relationships across all of it, and turns fragmented noise into something decision-makers can act on in minutes instead of days.

The US Army now has contracts with Palantir worth billions. Its TITAN battlefield systems delivers AI-processed satellite intelligence directly to soldiers in the field in real time. The company is worth over $200 billion with fewer than 5,000 employees.

Individually, none of them possess superhuman abilities.

Collectively, they have built the closest thing modern warfare has to an all-seeing stone.

And India, which exports many of the engineers capable of building such systems, is still struggling to make its own intelligence databases effectively speak to each other.

Remember Operation Sindoor?

India got a glimpse of what a Palantir-like operational system could look like during Operation Sindoor in May 2025.

For the first time, multiple intelligence and defence agencies were pulled into something resembling a unified real-time operational picture. RAW, NTRO, DIA, and IB were synchronised through integrated threat dashboards under the National Security Council. AI systems fused satellite imagery, drone feeds, electronic intercepts, battlefield surveillance, and historical military data into live operational intelligence.

Lieutenant General Rajiv Kumar Sahni later revealed that 23 AI-enabled applications were used during the operation for multi-source battlefield fusion. One indigenous system, ECAS (Electronic Intelligence Collation and Analysis System), reportedly achieved around 94% accuracy in identifying enemy radars, missile systems, and electronic signatures using historical pattern modelling.

Operational commanders received live geospatial feeds directly during strike planning, something Indian defence systems historically struggled to coordinate at speed.

In simple terms: India briefly built a wartime fusion centre.

And that is precisely the point.

Operation Sindoor proved that India possesses the engineering talent, AI capability, and intelligence infrastructure required to build Palantir-like systems during a crisis.

And that raises a far more important question:

If India can assemble something resembling a Palantir during a crisis, what exactly is stopping it from building one permanently?

The Problem

We face a problem similar to what US faced during 9/11. The bottleneck is not intelligence, it is coordination.

The country already possesses satellites, cyber units, surveillance infrastructure, electronic intercept systems, military intelligence networks, and some of the world’s strongest engineering talent. The problem is that much of this capability exists across dozens of separate organisations operating through fragmented systems, overlapping mandates, and disconnected databases.

RAW, IB, NTRO, DIA, MAC, CERT-In, NCIIPC, Defence Cyber Agency, state intelligence units, each sees part of the picture.

Very few can see all of it in real time.

This fragmentation has surfaced repeatedly across crises.

Kargil in 1999 exposed gaps in military intelligence coordination. Mumbai in 2008 revealed failures in information sharing between agencies. During the Galwan tensions, commercial satellite imagery had already indicated unusual PLA activity in the region for months before the crisis escalated publicly.

The issue was not the absence of signals. The signals existed. The challenge was turning scattered information into a unified operational understanding quickly enough for decision-makers to act on it.

And in modern warfare, speed increasingly determines advantage.

What Are We Building?

To be fair, India is not starting from zero.

Somebody, somewhere inside the system, clearly looked at Palantir and thought, “We should probably have one of those.”

The closest contender today is probably Innefu Labs, a Delhi-based company founded by Tarun Wig and Abhishek Sharma. Their platform, Prophecy Guardian, already ingests satellite feeds, telecom signals, OSINT, social media activity, financial intelligence, and surveillance inputs into unified intelligence dashboards for agencies like the Indian Army, NIA, paramilitary forces, and police departments.

In other words: the instinct is correct.

Connect the dots faster than the threat can move.

Tarun Wig once claimed they could match Palantir feature-for-feature at a fraction of the cost. Which, to be fair, sounds exactly like something every Indian founder says right before either building a global company or discovering the slow pace of enterprise procurement cycles.

And honestly, Innefu is credible. Probably the most credible player India currently has in this category.

But scale matters.

Palantir spent nearly two decades embedding itself into the operational nervous system of the American state. Innefu has raised roughly $33 million till date.

Though, around them, an ecosystem is slowly emerging. Staqu is turning CCTV footage into searchable intelligence. Videonetics has built large-scale urban surveillance platforms deployed across airports and smart cities. DRDO’s CAIR division has developed dozens of AI systems for defence applications, including ECAS, one of the key systems reportedly used during Operation Sindoor.

But when you look at the broader pattern, the bottleneck still remains the same. India is great at building the individual pieces like sensors, drones, and AI algorithms, but we are still struggling to build the orchestration layer that ties everything together into a live operational picture. The future of national security does not belong to the country with the biggest databases, but to the one that can connect the dots faster than the enemy.

Without that unified dashboard, our analysts are left doing exactly what I did during my embarrassing half-hour of scrolling: drowning in a sea of disconnected noise, trying to find a pattern that is not there, and getting a massive reality check the second you make the call.

India imports everything it cannot live without, from countries it cannot entirely trust, through sea lanes it does not control. Here is what happens next.

You remember the LPG crisis, right? For a couple of weeks, LPG cylinders became the most valuable things in the county, every restaurant was using everything except gas to cook meals and memes of LPG shortage flooded instagram.

That is not bad luck. That is structural dependency, experienced in a kitchen.

Now multiply that feeling by 1.4 billion people, replace the LPG cylinder with oil, semiconductors, antibiotics, and the minerals inside every EV battery and missile guidance system the country is trying to build, and you have a fairly accurate picture of where India’s economy sits in 2026.

Welcome to the problem. It is a big one. But, and this is the part worth reading, it is also, genuinely, one of the largest untapped business opportunities in the country’s history with the first part I like to call: With friends like these…

In early 2025, the US announced sweeping tariffs on imports from across the world. India got hit with 26%. Stocks of major companies hit an all time low, people lost jobs in these companies and finally, India and the US negotiated a partial deal and moved on. Standard procedure.

China did not move on.

Beijing looked at the tariff list, looked at its periodic table, and decided to respond in a way no trade negotiator had a playbook for. It did not raise tariffs back. It restricted exports of the materials the entire world’s industrial economy quietly depends on.

Gallium. Germanium. Graphite. Antimony. Seven rare earth elements. If you cannot immediately locate these on a periodic table, you are in excellent company, alongside every defence ministry, every EV manufacturer, and every semiconductor fab that suddenly discovered they had a serious procurement problem with no obvious fix.

Here is why these specific materials matter more than any tariff ever could.

Gallium goes into every high-frequency semiconductor on earth, the chips inside defence radars, satellite communications, 5G towers, and EV power systems. There is no practical substitute. Germanium sits inside fibre optic cables, infrared sensors, and night vision equipment, the kind militaries tend to consider non-negotiable. Graphite is the anode in every single lithium-ion battery ever made. The average EV needs 70 to 80 kilograms of it. Antimony goes into ammunition, flame retardants, and solar panels.

Now here is China’s position in each of these. It produces 80% of the world’s gallium. 60% of its germanium. 77% of its natural graphite, and refines over 90% of the synthetic graphite that actually goes into batteries. These are not competitive market shares. These are monopolies that took two decades of patient industrial policy to build, while everyone else was focused on the finished products China was selling them.

The result: monthly gallium exports fell 66% the moment licensing controls kicked in. Prices jumped 68% in European markets within weeks. By December 2024 the US ban was total. By April 2025 it had expanded to seven rare earth elements, terbium, dysprosium, and five others that live inside the permanent magnets powering wind turbines, EV motors, and missile guidance systems.

Source: Trading Economics.
Gallium price hike in 2024 Oct-Dec after China licensed its export.

No invasion. No ultimatum. Just a licensing memo, and large parts of the world’s industrial supply chain suddenly discovered how dependent they had become.

India watched from a particularly uncomfortable vantage point. Its own graphite imports were already subject to Beijing’s export-licensing regime. A country planning tens of gigawatt-hours of battery manufacturing capacity and positioning itself as an EV production hub was forced to confront a simple reality: a critical material underpinning that ambition sat behind a permission slip issued by a government with which it shares a disputed border.

The US–India tariff disputes were trade disagreements. Critical-mineral controls were something different altogether. One can be negotiated. The other must be engineered around.

Then, in June 2025, reports emerged that CATL, China’s battery giant, had instructed engineers to withdraw from a Foxconn-linked facility near Chennai. There was no press conference and no diplomatic standoff. Yet a manufacturing operation suddenly faced the consequences of dependence on foreign expertise embedded deep within its supply chain.

This is the new grammar of geopolitical pressure. It does not arrive with declarations of war. It appears as a delayed shipment, a licensing approval that never arrives, a missing component, or an engineer quietly told not to return. The weapon is the supply chain. The vulnerability is dependence.

Here is the particular irony of India’s situation.

The clean-energy transition was supposed to reduce dependence. The idea was simple: replace imported fossil fuels with solar panels, batteries and electric vehicles powered by domestically generated energy.

Instead, India risks exchanging one dependency for another.

The technologies that power the green transition rely on materials such as lithium, cobalt, nickel, graphite and rare earth elements. India imports most of them, while much of the global processing and refining capacity remains concentrated in China. The result is that a future built on batteries and renewable energy still depends on supply chains that originate far beyond India’s borders.

Governments have recognised the problem. India has accelerated critical-mineral exploration and auctions. International partnerships such as the Quad Critical Minerals Initiative aim to create alternative supply chains for mining, refining and manufacturing. These efforts matter. They will likely become an important part of India’s long-term strategy.

But they all share the same limitation.

They are trying to secure materials that India has not yet acquired.

But there is another way to think about critical minerals, one that focuses not on securing new sources of supply, but on making better use of the resources already in circulation. This is the central idea behind the circular economy.

What if the materials India needs are already here?

Because the reality is that many of the minerals India worries about importing have already entered the country once. They are sitting inside discarded smartphones, dead laptop batteries, scrapped electronics, old electric motors, retired solar equipment and end-of-life EV battery packs.

The lithium in a spent battery does not stop being lithium. The cobalt remains cobalt. The copper, nickel, graphite and rare earth elements do not disappear when a product reaches the end of its life. They simply become waste, or more accurately, they become resources waiting for a system capable of recovering them.

This is where the conversation shifts.

Most discussions about critical minerals focus on mines. Circular economies focus on urban mines.

A traditional mine extracts materials from the ground. An urban mine extracts them from products that have already been manufactured, transported, consumed and discarded. In many cases, the concentration of valuable materials is surprisingly high. A tonne of electronic waste can contain significantly higher concentrations of precious metals than a tonne of mined ore.

The challenge is not geology. It is infrastructure.

India generates millions of tonnes of e-waste each year, but much of it is handled through informal channels that recover limited value. The challenge is not a lack of waste, but a lack of collection, recycling and processing infrastructure.

That is also the opportunity.

The circular economy is often framed as an environmental solution. More importantly, it is an industrial one.

Every battery, kilogram of copper, cobalt, nickel or graphite recovered domestically reduces import dependence and strengthens supply-chain resilience. Recycling turns discarded products into a domestic resource base.

In that sense, the circular economy is not just a waste-management strategy. It is a supply-chain, trade and industrial strategy.

And increasingly, it is a national-security strategy.

This is why governments and investors around the world have begun paying attention. The question is no longer whether valuable materials can be recovered. The technology already exists. The question is who will build the infrastructure required to recover them at scale.

Because the country that controls the recovery of critical materials controls something increasingly valuable: a supply chain that does not need permission from anyone else.

If all of this still sounds a little abstract, consider what investors are already betting on.

Companies like LOHUM and Attero have built businesses around a simple idea: yesterday’s waste is tomorrow’s raw material. LOHUM, which recovers critical minerals from used batteries, raised $54 million in a funding round led by Singularity Growth Opportunities Fund and Baring Private Equity India, valuing the company at more than $450 million. Attero, one of India’s largest e-waste recyclers, has raised over $50 million from investors and says it processes more than 144,000 tonnes of e-waste and lithium-ion batteries annually. Their growth reflects a broader belief among investors that waste recovery is becoming an important part of India’s resource economy.

That might sound niche. It is not.

Every EV sold today is a future source of critical minerals. Every discarded phone, laptop and battery contains materials that India currently spends billions importing. What looks like waste is increasingly being treated as inventory.

This is why the circular economy is attracting serious attention. India’s e-waste stream is growing rapidly, EV adoption is accelerating, and demand for critical minerals is only expected to rise. The more electrified the economy becomes, the more valuable recovery infrastructure becomes.

The opportunity is not just environmental. It is economical.

Every battery recycled domestically is one less battery material that needs to be imported. Every kilogram of recovered lithium, cobalt or copper reduces exposure to volatile global supply chains. Companies that can recover these materials efficiently are not merely managing waste; they are becoming part of India’s resource base.

The most interesting part is that the raw material is already here.

Unlike a new mine, which can take years to discover and develop, India’s urban mines are sitting in drawers, warehouses, scrapyards and landfills across the country. The challenge is not finding the resource. It is building the systems to recover it.

And that is exactly where some of the biggest opportunities of the circular economy lie.