Capital Can Buy a Warehouse. It Cannot Buy Your WhatsApp Group.

On 10 January 2021, Kavin Bharti Mittal announced the end of Hike’s messaging app. Hike the company would pursue other businesses. It had Bharti, SoftBank and Tencent’s money, 100 million registrations, a $1.4 billion valuation and a product built expressly for Indians. His verdict fit in a tweet: “India won’t have its own messenger.” What it amounted to was brutal and accurate. You can build an Indian messaging interface in a year. You cannot build an Indian messaging network without extraordinary protection, because the network already exists and it belongs to someone else.

In February 2025, Uber made the opposite admission. It stopped taking a share of every auto-rickshaw fare in India and moved its drivers to a flat subscription, the model a Bengaluru auto union and a payments company had launched two years earlier. Its spokesperson explained the change in one clause: “so as not to be at a competitive disadvantage.”

One is an Indian founder conceding a market to an American network. The other is an American giant rewriting its contract with Indian drivers because local competitors had rewritten theirs first. Four years apart, they are the same story, and it is not the story we usually tell.

The usual version runs like this. China blocked Google, Facebook, Twitter and YouTube, so Baidu, Weibo and Youku grew up without competition. India let the Americans in, so Indian clones of American apps died, and India got WhatsApp. It is a clean story and half right. The half it leaves out explains why India has a Flipkart but not a Twitter, a PhonePe but not a Spotify, a Zomato but not a Zoom, and why Uber now runs an Indian business model.

Start with the premise, because it is wrong as stated. “Indian copies of American apps fail” has a long list of exceptions: Flipkart against Amazon, Myntra against Amazon Fashion, PhonePe against Google Pay, Zomato and Swiggy against Uber Eats, Hotstar against Netflix, Ola against Uber for a decade, Meesho against everyone. The outcomes are just as mixed elsewhere: Hike and Koo closed their original social products, while Sandes, Mitron, Chingari, JioChat, JioMeet, Gaana and Snapdeal had very different trajectories and levels of adoption.

It is tempting to sort these into networks that lost and operations that won, and that sorting is roughly right. But e-commerce and ride-hailing have network effects too, music and video calling are not locked the way a family WhatsApp group is, and ShareChat and Moj show that an Indian social product can build a large audience without displacing a foreign one. So the better test has three questions.

First: how hard is it for a customer to switch, and who bears that cost? If the thing that makes the product useful is other people, the cost is paid by everyone the customer knows, and it is almost never paid. Second: can the newcomer build an advantage the incumbent cannot import, a warehouse, a merchant on every street, a cricket right, a better contract with the person doing the work? Third: can it finance the years between launching and the advantage mattering? Timing and regulation bend all three, and can change the odds dramatically.

That is the difference between copying a product and competing for the system that makes the product useful. The rest of this essay walks through the four systems Indian companies have fought over: the social graph, physical density, the public payments rail, and the economic contract with the supplier. Then it asks what China really did, who paid for the losing years, what engineering can remove from that bill, and why the next round is beginning to run the other way.

The Social Graph Lock-in

Hike is usually told as a product story: too many features, a WeChat-style super app Indians did not want. Its founder has said as much. The timeline says something simpler.

Hike launched on 12 December 2012, when WhatsApp already had 100 million monthly users worldwide and was the default on the BlackBerrys and Nokias Indians were upgrading from. Eight months later WhatsApp disclosed 20 million monthly actives in India; Hike had 5 million registrations. By November 2014 WhatsApp had 70 million Indian actives and India was its largest market. Hike reached 100 million registrations in 2016 and raised $175 million from Tencent and Foxconn at a $1.4 billion valuation. By December 2019 it had 2 million monthly actives. The gap between registered and active reveals the problem: people installed it, but their networks remained elsewhere. It was not under-funded, under-connected or under-localised. It had $261 million, the heir to Bharti Airtel as founder, stickers in ten Indian languages and free SMS to people who had not installed it. None of that touched the only variable that mattered, which is who else is already there.

Koo ran the same experiment in microblogging and got the answer faster. It launched in March 2020, caught a political tailwind in 2021 when ministers quit Twitter over a content dispute, peaked at 9.4 million monthly users in July 2022 and had 3.1 million by April 2023. It had raised $65 million from Tiger Global and Accel and shut on 3 July 2024 after a sale to Dailyhunt fell through. The sentiment that brought people in did not become the habit that keeps them.

India measured the switching cost directly in January 2021. WhatsApp announced a privacy policy that would share data with Facebook, and the reaction was the strongest reason Indians have ever had to leave an American network. A CMR survey found 79% of users reconsidering WhatsApp and 28% planning to leave. Signal was downloaded 24.6 million times in India in four days. The government wrote to WhatsApp’s CEO. Then LocalCircles asked 17,000 people across 232 districts what they had actually done: 55% had downloaded an alternative, 21% were using one, 5% had deleted WhatsApp. The sample was self-selected and the number should be read as a direction, not a census. The direction is unambiguous. WhatsApp has more than 500 million Indian users today. LocalCircles put the mechanism in one line: you cannot fully switch until roughly three-quarters of your groups and contacts have switched. The switching cost in messaging is not yours. It is everyone else’s.

India multiplies that cost in a way China never had to deal with. It has the largest diaspora on earth: 35.4 million overseas Indians as of November 2024, 5.4 million of them in the United States and 3.6 million in the UAE, sending home $129 billion a year, the largest remittance flow of any country. Every one of those households is a WhatsApp group with members an Indian-only network cannot reach. Hike proved the point without meaning to: in 2014 over 90% of its users were in India, at the moment WhatsApp had 600 million worldwide. A messenger that only Indians use cannot hold an Indian family.

There is a hardware reason the lock holds as well, and it is specific to India. Hundreds of millions of Indians run their digital lives on entry-level Android phones where storage is a managed commodity. A 2017 survey reported widespread storage pressure on smartphones in India. Google developed Files Go to make managing limited storage easier; circulating WhatsApp media was one of the common sources of clutter. The app that fills the phone is the one you are not allowed to delete, because your family is on it. When space runs out, the app that goes is the alternative, however good it is. The social graph holds the storage, and the storage holds the graph.

Short video gave us the control case. On 29 June 2020 India banned TikTok, which had 167 million Indian users. TikTok was suddenly gone, although Instagram still had a vast installed audience. Moj launched on 1 July, Josh on 3 July; Instagram began testing Reels in India on 8 July with 100 million Indians already inside the app. By December 2020 Redseer found the Indian apps had taken 40% of TikTok’s former share, and by early 2022 Moj claimed 160 million monthly users and Josh 150 million. Four years on, Nikhil Pahwa’s verdict to the Washington Post was that Reels had won outright, and Moj and Josh had moved into live video and paid engagement. Nepal ran the counterfactual: it banned TikTok, Reels took over, it lifted the ban in August 2024, and TikTok came straight back.

Two things stop this from becoming a blanket rule. ShareChat and Moj did not die; they built a different graph, in Indian languages and formats, that did not require anyone to leave Instagram. A newcomer can fail to replace a social graph and still succeed at building one beside it. And not everything with users is a graph. Nobody needs their cousins to switch music apps with them; Spotify’s barriers in India were catalogue, licensing cost and habit, which is why a late entrant could take the lead. Zoom’s position in Indian workplaces rests on interoperability and corporate purchasing, not on friendship. Calling all of these “network effects” is like calling every kind of traffic a traffic jam. The thing that matters is whether the value of the product depends on an entire existing group moving together.

People like to explain the WhatsApp lock with prestige: Indians, the argument goes, see an American app as a status marker and an Indian one as second-rate. I have not found a study that shows it, and I would not print it without one. What the data shows is a gap between what people say and what they do. A month after the TikTok ban, 65% of urban Indians told YouGov they were likely to switch to Indian or non-Chinese video apps; in the same survey 63% wanted the ban revoked. In January 2021, 79% said they were reconsidering WhatsApp and 5% left. Call it the social default rather than prestige. The American app is where the group already is, and “vocal for local” is something people say in surveys and do not do with their contact list.

The Physical Density Moat

The Indian companies that beat American originals did not copy the product. They copied the category and rebuilt the layer the original could not ship from Seattle.

Flipkart copied Amazon’s warehouses, not its website. Sachin and Binny Bansal left Amazon’s India office in 2007 after Amazon decided against launching e-commerce here. Early investors passed because they believed India had no roads, no warehouses and no trained delivery staff. Flipkart built that layer itself: its own logistics arm, cash on delivery in 2010, a 30-day replacement policy in 2011. Those were not features. They were the reason an Indian who had never bought anything online would try it once. Flipkart also got six years before Amazon was allowed to sell anything: India’s foreign investment rules permitted foreign ownership only of a marketplace, not an inventory-led retailer, which is why Amazon spent 2012 running a price-comparison site called Junglee and arrived in June 2013 with 100 sellers and two categories.

Snapdeal is the comparison that matters, because it was also an Indian copy, raised about $1.4 billion in 2014–15 alone, and lost. It copied the other American model, eBay’s asset-light marketplace with outsourced logistics, in a country with no logistics layer to outsource to. Kunal Bahl’s February 2017 letter to staff is the most honest document in Indian startup history: they grew before the economic model worked, started new projects before perfecting the first, and built a team for a business far larger than they had. FreeCharge, bought for about ₹2,400 crore, went to Axis Bank for ₹385 crore; a $950 million sale to Flipkart collapsed because the investors could not agree on who got paid. Snapdeal did not die. It retreated to value commerce in smaller towns, and its parent AceVector listed in October 2026 at about ₹1,750 crore, a fraction of what had gone in. Model alone did not decide this; capital allocation, merchandising and management did too. But same country, same years, same nationality, opposite outcome is hard to explain any other way.

Flipkart nearly lost anyway. In 2015 it cut in-house logistics and went app-only, and by mid-2016 Amazon had briefly overtaken it. Investors pushed the founders aside, brought in Kalyan Krishnamurthy, and the October 2016 Big Billion Days went all-in on smartphone exclusives and no-cost EMI: 50% of festive sales to Amazon’s 32%. Walmart paid $16 billion for 77% of it in 2018. In FY25, by CLSA and ICICI Securities’ estimates, Flipkart holds 50 to 60% of Indian e-commerce by value and Amazon 25 to 30%. In the market-share estimates cited here, Flipkart remained ahead of Amazon India; the leader is Indian-built and American-owned.

Zomato and Swiggy bought the orders and built the fleet. Food delivery is where the difference between a social copy and an execution copy becomes visible. A social copy is measured in monthly actives, and monthly actives cannot be bought; Koo bought downloads, not habit. An execution copy is measured in transactions, and transactions can be bought. That is why Indian execution copies look like winners during the burn, and why the ones still standing when the discounts stop are the real ones.

Swiggy was founded in 2014 with its own fleet because Zomato’s restaurants could not deliver reliably. Zomato spent 2018 catching up: 5,000 riders in January, 74,000 by October, 90% of orders self-delivered. Between them they were spending $30 to 40 million a month on discounts and rider incentives; Info Edge, Zomato’s earliest investor, sat out a round and described the war as raising money and setting it on fire. Zomato lost $294 million in FY19 and Swiggy $330 million. They kept going because India was the whole company.

Uber Eats launched in India in May 2017 and was a line item. By 2019 it was 3% of Uber Eats’ global bookings and more than a quarter of its losses, around $500 million a year by the Financial Times’ account, in the year Uber had promised Wall Street profitability by 2021. In January 2020 it sold the India business to Zomato for a 9.99% stake and the riders moved to Zomato’s fleet the same day. Foodpanda, the Rocket Internet clone, never built a fleet at all; Delivery Hero sold it to Ola for $31.7 million in 2017 and Ola suspended it in May 2019 at about 5,000 orders a day. Amazon Food lasted from 2020 to 2022. By Q1 FY25 Zomato held 58% of the market and Swiggy 42%.

None of this proves foreign companies cannot build local operations; Amazon and Walmart have. The point is narrower. Indian average order values are $5 to 7 against $30 in the United States, so the only route to margin is density per rider, per van, per pin code. Density is local operating knowledge: which incentive keeps a rider through the dinner rush, which kirana will put your sticker on the counter, which neighbourhoods will accept cash on delivery. It cannot be managed from a headquarters that sees India as 3% of bookings, and it cannot be bought faster than it can be built. A rider cannot deliver tomorrow’s dinner on today’s venture round.

The Interoperable Public Rail

The received version of PhonePe is that an Indian app beat Google Pay. The data says something more useful, and it turns on a piece of public infrastructure.

PhonePe shipped the week UPI went live in August 2016, thirteen months before Google launched Tez. Google still took the lead within three months: Tez accounted for 52% of UPI transactions in December 2017, on a slicker product, cashback referrals and the best Android distribution on earth, and it led the peer-to-peer phase for most of the next three years.

What made the contest winnable at all was UPI itself. The rail is interoperable: a customer on Google Pay can pay a merchant’s PhonePe QR code, and the merchant never has to assemble an exclusive universe of customers first. That is almost the exact opposite of the WhatsApp problem. Nobody’s family had to move. The fight shifted to the physical layer, which is where PhonePe went: 3 lakh merchant outlets in 2018, 90 lakh offline shops by February 2020, 37 million merchants by 2023, 40 million by 2025, with zero fees on UPI, instant settlement and a QR code any app could pay into. Person-to-merchant payments crossed 40% of UPI in late 2020. In December 2020 PhonePe overtook Google Pay on volume, 40.4% to 38.0%, and has led every month since. In August 2025, NPCI data put PhonePe at 9.15 billion transactions, 45.7% of volume and 48.3% of value, and Google Pay at 7.06 billion, 35.3% and 35.6%.

Three qualifications keep this honest. PhonePe is part of the Walmart group, and separated from Flipkart in 2022 partly to look domestic for an IPO. WhatsApp Pay did not compete on equal terms: NPCI held it to 20 million users, then 40 million, then 100 million, from November 2020 until the cap was lifted at the end of 2024, and it still holds about half a percent of UPI. And NPCI’s proposed 30% cap on any single app’s share has been deferred three times, now to December 2026, largely because enforcing it would mean dismantling the Indian leader.

The rail is the overlooked common factor in two of the Indian wins in this essay. UPI separated the payment rail from the payment app. Beckn and ONDC are attempts to do the same for mobility and commerce: let buyers and sellers discover and transact across participating apps rather than inside one company’s closed universe. Namma Yatri, in Section 7, was built on Beckn and joined ONDC in 2023. Neither PhonePe nor Namma Yatri had to own the whole network to compete on it.

Openness does not guarantee competition. PhonePe and Google Pay together hold over 80% of UPI. An interoperable rail lowers the cost of entering a market without removing the advantage of being the first app people remember to open; in 2026 smaller payment companies warned that one-click checkout defaults could make a customer’s first choice permanent. And the economics of the rail itself are unsettled: in October 2026 NPCI deferred a planned 0.4% merchant fee on UPI payments above ₹2,000 until after Diwali, having spent a decade telling merchants it was free. Public infrastructure is not free of commercial trade-offs.

But it changes the starting point. A founder on an open rail no longer has to persuade a nation to join a proprietary network before building something useful. What remains to be built is trust, distribution, software and operating discipline, which are exactly the things Section 2 says Indian companies have been good at.

What China Actually Did

The wall is real, so state it precisely. YouTube was blocked in March 2009, Facebook and Twitter in July 2009 after the Xinjiang riots. Google pulled its search engine out in March 2010, though Baidu already led 56% to 43% before it left. Instagram followed in 2014, WhatsApp was throttled in 2017 and Skype pulled from app stores the same year. Behind the blocks sits a rule book: an internet content licence only a Chinese-controlled entity can hold, a 50% cap on foreign ownership of value-added telecom services dating from WTO accession, and a 2017 cybersecurity law that keeps data at home. Asked in December 2017 why Google and Facebook were absent, the regulator said they were welcome if they would abide by Chinese law.

That wall explains Baidu, Weibo, Youku and WeChat. It does not explain Taobao, JD, Didi or Tujia, and those are the ones India should study, because the American original was never blocked.

eBay entered China in 2003 and held 79% of the market. Taobao launched the same year, made listings free, put a chat window between buyer and seller and, in October 2003, launched Alipay to hold the buyer’s money in escrow until the goods arrived. By 2005 Taobao had 59% and eBay 36%; eBay shut its Chinese site in 2006. Market research at the time found free listings were not among the top five reasons buyers chose Taobao. Trust was. eBay’s China executives told interviewers that headquarters did not listen to them. Uber spent two years and $2 billion in China, was doing 150 million rides a month, and sold the business to Didi in August 2016 for a 17.7% economic interest because its own investors would not fund the fight. Amazon, Airbnb and LinkedIn each exited China without ever being blocked.

Notice what those losses share with Flipkart and PhonePe’s wins. Taobao’s escrow is Flipkart’s cash on delivery. Didi’s driver density is PhonePe’s merchant density. In each case the local company rebuilt the trust and operations layer for local conditions while headquarters ran the global playbook, and in each case the American company had more money. China also took the foreign money without the foreign product: SoftBank and Yahoo funded Alibaba, Naspers funded Tencent, through structures that kept control Chinese. India took the money too; the difference is that in India the control went with it, to Walmart, and to Amazon’s own subsidiary.

India also created sector-specific restrictions, without adopting China’s sweeping internet controls. Foreign investment rules kept Amazon to a marketplace model in 2013, capped its biggest seller in 2016 and banned its exclusives in 2019. The RBI forced payments data onshore in 2018. NPCI kept WhatsApp Pay on a leash for four years. The TikTok ban removed the only foreign short-video incumbent in 2020, and Press Note 3 the same year put Chinese capital behind a government approval. The resulting outcomes are mixed: domestic firms became leaders in e-commerce and payments, but Instagram Reels shows that removing a foreign incumbent did not guarantee a domestic winner in short video. Regulation shaped the field; execution still mattered.

The honest counter-argument is that the wall also kept Chinese consumers off the global internet, and most Indians would not trade WhatsApp for a domestic clone in order to have a domestic clone. The YouGov number makes it concrete: 63% of urban Indians wanted the TikTok ban revoked. Nor is a domestic champion automatically the best outcome for consumers; a market can gain from global scale and interoperability even when the leading product is foreign-owned. India chose openness in the categories where openness costs it a champion. That was a trade, not an accident, and the right question is not why India did not close its internet but which businesses can build durable advantages while it stays open.

Who Pays for the Losing Years

There is a second reason Indian copies die that has nothing to do with graphs, and it starts with what Indians pay.

India downloaded 25.5 billion apps in 2025, more than any country, roughly double the United States. It is not in the top twenty by in-app spending; Sensor Tower puts Indian in-app revenue at about $1 billion for the year against $52 billion in the United States and $25 billion in China. Downloads and in-app revenue are different measures and should not be treated as spending per user; read together, they nevertheless demonstrate the challenge of monetizing vast mobile usage. Netflix earns about $7 a month from an Asia-Pacific member and $17 from an American one, and India sits below the Asia-Pacific average; in December 2021 it cut Indian prices by up to 60%, to ₹149 for mobile, when Media Partners Asia estimated it had 5 million Indian subscribers to Hotstar’s 46 million. India’s entire music streaming market was $372 million in 2025, with about 8% of listeners paying, and Spotify sells a ₹7 mini plan.

In a market like that, somebody other than the customer pays for the formative years of any digital product. The question is who, and for how long.

An American entrant books India as an option on a global P&L. Amazon’s marketplace arm lost ₹3,470 crore in FY24, cut that to ₹374 crore in FY25, and nobody in Seattle asked whether the experiment should continue; Amazon has committed another $48 billion to India by 2030. Netflix entered in 2016, cut prices by 60% in 2021, and its Indian entity now books modest profits, ₹20 crore in FY22 rising to ₹85 crore on ₹3,769 crore of revenue in FY25, while the content bill sits with the parent and India remains a small line in a global business. Nobody in Los Gatos ever had to decide whether India was worth it. Uber India’s net ride-hailing revenue fell 89% in FY25 as it spent against Rapido, and its parent put in ₹2,921 crore to keep fighting.

An Indian company’s India P&L is the whole P&L. When capital tightens, in 2017 or in 2022–24, the standalone Indian network businesses die and the American ones keep spending. Snapdeal’s founders went to zero salary in 2017 and spent the next nine years shrinking the company to a size it could afford. Hike shut with $261 million behind it. Koo shut with $65 million, citing the funding winter. Jabong, which once matched Myntra in sales and turned down Amazon at a $1.2 billion valuation, sold for $70 million. Gaana, which claimed 160 million users, lost the market to a Swedish company that arrived nine years after it. Even Flipkart’s founders were removed from control in 2016 when Amazon pulled level, and the company was sold two years later; its CEO’s explanation was that startups live from funding round to funding round.

Now look at who the Indian winners in low-margin categories actually were. Hotstar did not beat Netflix as a startup. It was Star India’s streaming arm, launched on 11 February 2015 through a wholly owned subsidiary, timed to the Cricket World Cup, with Star’s television library at marginal cost, ₹302 crore of IPL digital rights already bought and a marketing blitz on Star’s own channels, eleven months before Netflix arrived into a field of Eros Now, Sony LIV and YouTube. Its first-year revenue was ₹186 crore; Star carried the gap, then bid ₹16,348 crore for consolidated IPL rights in 2017. JioCinema was marginal until Reliance paid ₹23,758 crore for the 2023–27 IPL digital rights and streamed them free, picking up 5 crore downloads in one weekend. PhonePe was bought by Flipkart for under $20 million four months before UPI launched and funded by Flipkart and then Walmart through years of merchant acquisition. Myntra survived its own app-only mistake in 2015 because Flipkart’s balance sheet absorbed it.

Hotstar’s moat was not just that the parent company could afford to outbid everyone for IPL digital rights. It was a mobile-first streaming architecture engineered specifically to survive tens of millions of concurrent viewers on entry-level Android devices and volatile cellular networks. Netflix shipped a premium global product; Hotstar shipped a localised engineering solution tied to a uniquely Indian cultural obsession, priced for a market that expects to pay almost nothing. The public numbers show what that meant: 18.6 million concurrent viewers for the 2019 IPL final, 25.3 million for that year’s World Cup semi-final, 55 million during IPL 2025. The capital bought the rights. The engineering locked in the scale.

So the first half of the money story is this: Indian copies die of capital, not of copying, and the ones that lived in low-ARPU categories had a balance sheet behind them, which is exactly how Amazon funds Amazon India. The asymmetry is not that Indian investors want 100x and American ones accept 1x; Tiger, SoftBank and Walmart funded Flipkart through billions in losses. It is patient balance sheet versus an approaching funding deadline, and the passport on the balance sheet is irrelevant.

If the story ended there, the lesson would be that money is the answer, and Reliance ran the experiment that proves it is not.

Jio is the controlled test the thesis needs: one parent, one balance sheet, 476 million active telecom subscribers, apps preloaded on Jio phones and bundled with Jio SIMs, Meta and Google as shareholders. In July 2020 it launched JioMeet, a screen-for-screen copy of Zoom with free 24-hour calls; Mukesh Ambani announced 5 million downloads within days, and no active-user figure has been published since. The same month it redesigned JioChat to mirror WhatsApp; it has 50 million downloads and no measurable dent in WhatsApp. In 2018, nine months before Spotify arrived, Reliance merged JioMusic with Saavn at a combined value over $1 billion and gave every Jio SIM 90 days of free Pro; by FY23 Spotify led Indian audio streaming with 26% share. JioMart is not in the top three of quick commerce. Reliance’s actual messaging strategy, in the end, was to sell Meta 9.99% of Jio Platforms for $5.7 billion and build JioMart inside WhatsApp.

Where Jio won, money bought something scarce that nobody else could get: spectrum, IPL rights, and eventually Disney’s India business, merged with JioCinema into JioHotstar in February 2025. That is the same shape as Flipkart buying warehouses and PhonePe buying merchant feet on the street. Money is not useless against networks; it can subsidise creators, seed marketplaces and buy companies. What it cannot do is force an entrenched relationship network to move at once. If money alone decided these fights, JioChat would be WhatsApp.

The Architectural Escape Velocity

There is one more kind of capital advantage, and it is the only one in this essay a founder can build alone. It sits below the business model, in the architecture of the software.

The default startup pattern for fifteen years has been the global template: put every user’s data in a centralised cloud backend, run the application’s logic on rented infrastructure, and pay for storage, compute, databases, bandwidth and an operations team as usage grows. Sometimes that is unavoidable. Messaging, marketplaces, collaboration tools and live video need shared infrastructure; a cricket stream for 55 million people cannot run on anyone’s phone alone. But founders routinely copy the architecture of a platform when they are building a tool one person uses by themselves, and in a market where consumer spending is comparatively modest, that recurring AWS bill can depend on outside capital. To be precise about the history: Hike’s $261 million went on marketing and payroll, not servers. The point here is about the next generation, not the last.

The escape velocity for that generation lies in local-first engineering. Move suitable processing on-device and offer optional sync through the user’s own iCloud or Google Drive account, and the application can significantly reduce backend spending per user. It also works offline, on the congested networks Section 5 describes, which in India is a product advantage before it is a cost advantage. This does not make the bill zero. Sync APIs have limits, Apple’s and Google’s ecosystems carry lock-in, and account recovery, sharing, security, support, backups and privacy obligations, health data above all, reintroduce costs; so can app-store fees, depending on the platform and commercial model. What architecture changes is the slope of the cost curve. A company whose infrastructure spend barely moves when a solo user opens the app can run on a different funding horizon from one whose cloud bill rises with every account. That is escape velocity in the literal sense: not immunity from failure, but freedom to reach useful scale without financing centralisation the product does not need.

It does not solve the WhatsApp graph problem; no architecture makes your family switch. It does answer the capital question for products whose value is personal rather than collective: solo utilities, writing tools, health trackers, offline reference. For those products, a founder may reach meaningful scale with less outside funding, competing more on utility and less on balance-sheet endurance. Some will not need a deep-pocketed parent. The founder’s advantage may be a warehouse, a QR code, a better supplier contract, or a deliberately small server bill.

That gives the money story its second half, and the two together are the thesis of this essay: thin capital can end an operations challenge before it matures, while even abundant capital cannot guarantee that an entrenched relationship graph will move together. Capital works when it converts into something specific, a fulfilment system, a denser supply base, a unique catalogue, a cheaper transaction, a trusted counter. Without that conversion, spending buys attention and then has to buy it again.

The Supplier Economic Contract

Everything so far has been about Indians copying Americans. Ride hailing is where it started to run the other way, and it is worth telling in order, because the order is the point.

Ola was the Indian copy of Uber. It launched in 2010, three years before Uber reached India, built the auto-rickshaw category Uber had not thought about, and held the lead for a decade. By FY24 it still had more than 40% of the cab market. Both companies ran the same contract: match passenger and driver, take 20 to 40% of the fare. Drivers understood the weakness in that arrangement before most analysts did. At the bottom of urban transport the commission came out of a very small fare, and the driver supplied the vehicle, the fuel and the labour.

On 1 November 2022 an app called Namma Yatri launched in Bengaluru. It was not built by a startup in the usual sense. Juspay, a payments company, built it with the Beckn Foundation and the Auto Rickshaw Drivers’ Union on the open Beckn protocol (it joined ONDC in March 2023), and its executives drove autos themselves to understand what they were designing for. It charged drivers no commission; from mid-2023, a subscription of ₹25 a day or ₹3.50 a trip. It launched the week Karnataka was threatening to ban Uber, Ola and Rapido for overcharging on autos. Within fifteen months it had 213,000 drivers, 28 million trips and $51.75 million paid out to them, and had cut its cost per ride from ₹219.5 to ₹8.6.

Rapido, a bike-taxi company since 2015, copied the thesis and added price. It launched cabs on 5 December 2023 with a lowest-price guarantee and the same zero-commission subscription; the launch coverage named Namma Yatri as the template. In February 2024 it moved autos to lifetime zero commission. It had the two-wheeler density Ola and Uber had never built and a take rate of 0 to 5% against their 18 to 22%.

Then the incumbents moved, in the order you would predict. In April 2024 Ola and Uber piloted subscriptions for auto drivers in a handful of cities. On 18 February 2025, the month Chennai’s auto and cab unions boycotted both over commissions, Uber rolled out zero commission for auto drivers across India. Its explanation deserves to be quoted whole: “Given the industry’s shift towards a subscription-based model for drivers, we have decided to align our approach accordingly so as not to be at a competitive disadvantage.” Ola extended zero commission to cabs in June 2025, the last to move.

The scoreboard this year: Rapido carries about half of all app-hailed rides when you count bikes and autos, and more than 20% of cabs. Uber holds about 45% of cabs. Ola has fallen to 25 to 30%. Asked on a podcast what Rapido was getting right, Uber’s CEO pointed to the driver subscription model, and said Ola had got distracted while its founder built an electric vehicle company.

The supplier is the last-mile person in ride hailing, and the companies that read the supplier forced the incumbent to change its contract. The passenger did not need a new map. The driver needed a different bargain, and Indian competitors made that bargain impossible to ignore. Uber had run subscription arrangements in other markets before, so this is not an invention claim. It is something more useful than one: an American company adopting an Indian business model in India, in public, with a press statement.

Two caveats, so nobody mistakes this for a victory lap. Namma Yatri has not reached break-even and Uber’s CEO says Rapido is a long way from profit; the model won share, and whether it makes money is open. And tax may yet undo it. Karnataka’s advance ruling authority exempted Namma Yatri from GST in September 2023 because it was software rather than a transport aggregator, then ruled in 2024 that Rapido and Uber were liable under near-identical models, which the platforms called a distortion of competitive parity. In October 2026 the GST Council proposed clarifying the law so that every platform pays 5% on the full fare regardless of commercial model. If that passes, the economics that spread the subscription model change for everyone at once. Market share is not a moat, and a moat is not a profit.

The Foil: Software With No Graph and No Street

One contrast sharpens all of the above. The Indian software companies that have won globally, rather than defended India, are the ones that never had to fight a social graph or build a delivery fleet. Zoho has been bootstrapped and profitable for three decades selling business software to the world from Chennai. Freshworks listed on Nasdaq in 2021. Postman is the default API tool for developers who have never heard of Bengaluru. None of them needed a customer’s cousins to switch, a rider on a scooter, a cricket right or a regulator’s leash on a foreign rival. They compete on product utility, operating efficiency and price, in categories where contracts, integrations and data migration may create switching costs, but coordinated family-wide migration is not required. That is the same three questions from the opening answered in a different way, and it is why the Indian company most likely to be a global brand sells to businesses rather than to the family group chat.

So How Does a Startup Win Here?

It would be tidy to end with a playbook. There isn’t one, and the cases above are the proof. For every Flipkart there was a Snapdeal that raised comparable money in the same years and ran a defensible version of the same idea. For every Zomato there was a Foodpanda with a German parent and a head start. The pattern helps identify structurally difficult fights and explains what the survivors had in common. It does not tell you how to win them, because a large part of that was never in anyone’s control.

Start with luck, because founders undersell it and investors pretend it away. Flipkart got six years before Amazon was allowed to sell anything in India, not because of anything the Bansals did but because of investment rules written for retail. PhonePe shipped the week UPI went live and got demonetisation as a launch campaign three months later. Hotstar launched eleven months before Netflix. Namma Yatri launched the week Karnataka threatened to ban its three competitors. Zomato and Swiggy fought their war in 2018 and 2019, when capital was cheap and Uber had an IPO to protect. Luck cuts the other way as cleanly: Hike launched into a WhatsApp that already had 100 million users, Koo rode a political wave and drowned when Musk bought the platform it was rebelling against, and Jabong held out for a better price in the one year prices were still going up.

Then grit, which is real and not sufficient. Snapdeal’s founders took their salaries to zero in 2017 and it saved nothing, because the model underneath was wrong. Zomato spent six years clawing from 39% to 58% against a rival that was better capitalised for most of them, and it worked because the model underneath was right. Grit decides outcomes between companies that have already chosen the right fight. It does not rescue a company that has chosen the wrong one.

What was actually in the founder’s hands, across every case here, comes down to four questions.

What exactly must customers abandon to use it? If the product requires their whole group to move, acquisition cost is not measured in downloads, and nobody in this essay has won that fight. If it can coexist with the incumbent, or ride an interoperable rail, entry is a different problem.

What scarce local advantage can the company actually build? Not a Hindi interface or a cheaper copy, but delivery density, merchant relationships, content rights, trust, regulatory knowledge, engineering built for the market’s phones and networks, or a better contract with the person who does the work. Flipkart and Snapdeal were both Indian copies in the same market in the same years; the one that built the logistics layer lived.

Why is this the right time? A rule change, a new rail, a neglected segment, an incumbent’s retreat or distraction. None of these is under a founder’s control, and ignoring them does not make them go away.

Who pays until the advantage is self-sustaining, and what can engineering remove from that bill? A parent, a profitable adjacent business, committed financing, or a category in which bought transactions become a moat before the money runs out. For products whose value is personal rather than collective, an architecture that does not need a backend at all (Section 6). The aim is not to avoid spending. It is to make sure the money spent becomes durable capability instead of temporary attention.

And there is a list of things no founder controls that decided several of these outcomes outright: a regulator’s view of a foreign competitor, a headquarters decision made for an earnings call, a rival’s founder getting distracted by electric vehicles, a government banning the incumbent, a billionaire buying the platform you were built to oppose, a tax ruling that treats two identical business models differently. Anyone who tells you they planned for those is telling you a story after the fact.

So the honest answer to “how does an Indian startup win against an American original” is this. Think carefully before confronting an installed relationship graph head-on; the burden of persuading groups to move together is exceptionally high. In every other category, choose the model that builds the local layer rather than rents it, understand the person at the last mile better than anyone in a head office can, arrange your capital and your architecture so that a bad year does not end you, and then work harder than is reasonable while you wait to find out whether the timing was on your side. That is not a formula. It is a description of the survivors, and the survivors are not the whole population.

Two Admissions

Go back to the two quotes.

Mittal understood the immediate challenge. Displacing India’s established messenger, or a large public-conversation network, is difficult because the graph was already here, it reaches the cousin in New Jersey and the uncle in Dubai, it occupies the last free gigabyte on a ₹7,000 phone, and nothing you can buy in Bengaluru, not stickers, not Tencent’s money, not 476 million preloaded handsets, changes who else is on it. A new chat app can copy the features. It cannot download your friendships. That does not mean India can never produce a social platform; ShareChat shows it can build one beside the incumbent. It means a challenger needs a reason for people to gather that is stronger than its passport.

Uber was also right, and its concession tells you where the next decade goes. The Indian wins have never come from copying American products into American-shaped categories. They came from the places where India’s constraints produce a different answer: a ₹30 scooter delivery that only works at density, a kirana counter that takes a QR code before it takes a card, a 20% commission an auto driver will not pay once someone offers him ₹25 a day, a public rail that nobody owns. Flipkart copied Amazon’s warehouse. PhonePe copied nothing and built forty million merchants. Namma Yatri made Uber copy its contract.

The two admissions describe the two limits of copying. In one market the useful thing was already connected to everybody else. In the other it had to be built and negotiated street by street. For the next Indian founder the question is no longer how to build the Indian version of an American app. It is which Indian constraint the American app cannot survive.