India Is a Dealer's Choice Market

There is no settled version of India's rules for you to buy. Anyone selling you one is selling you a snapshot and calling it a map.

Every foreign client I've worked with says the same thing about this market, usually about two years in, usually when the business case is slipping: the rules keep changing. The observation is correct. The grievance is misplaced. Changing rules aren't a defect here. They're the specification, and you can build against a specification.

I got the words for it at a card table last weekend, in a game where the dealer redefined what winning meant before every hand. Never in secret — always out loud, in advance, in front of everyone. You could hate the condition and you still had to play it. More on that game shortly, because the useful part is what it taught me about the people who lose at it.

First, what it looks like with real money.

In March 2024 India published the terms in plain language: commit at least $500 million, build locally inside three years, source a quarter of your components here, and you may import 8,000 cars a year at 15% duty instead of 70 to 100%. Every one of those terms was dated and public, which is another way of saying the condition was announced before the cards went out.

Mercedes-Benz, Volkswagen-Skoda, Hyundai and Kia all engaged with it. Tesla declined. I remember reading that list when it came out and thinking the absence was the whole story — four foreign carmakers with far more sunk in India than Tesla had, every one of them willing to take a bet Tesla wouldn't.

Musk had spent years saying publicly that our import duties were among the highest in the world. He was right. It also wasn't a strategy. It was arguing with the dealer. In April 2024 he cancelled a planned trip to meet the Prime Minister and went to China instead. By that July his executives had stopped calling Indian officials at all.

Tesla opened an experience centre in Bandra Kurla Complex in July 2025 and shipped cars in from Shanghai at roughly 70% duty. Between September 2025 and June 2026 it retailed 450 cars in India, under fifty a month. Over the same months BMW sold 3,433 and Mercedes-Benz 1,116, both foreign carmakers in the same market under the same duty regime. Tesla has since cut the Model Y from ₹59.89 lakh to ₹50.89 lakh, which is what discounting into a position you never wanted looks like. On 19 May 2026 the Heavy Industries Minister confirmed Tesla had told the government it would not build here at all.

Now the honest complication, because somebody will raise it. Tesla's 2026 refusal wasn't really about India. The company had unsold inventory. It built 408,386 cars in the first quarter of that year and delivered 358,023, fifty thousand more made than sold, and committing half a billion dollars to a new plant anywhere was hard to defend.

That is exactly the point. Waiting didn't preserve Tesla's options. It handed the decision to a later version of the company operating under different pressures, and that version said no. The company that could have justified a plant here was the capacity-constrained, still-growing Tesla of 2021 — precisely the one doing the complaining.

The cost of waiting isn't the delay. It's that the market gets allocated while you deliberate. India's electric car sales grew 84% in FY2026 to nearly 200,000 units. Tata took 78,811 of them, MG 53,089, Mahindra 42,721 after a fivefold jump in a single year. Even BYD, whose billion-dollar plant proposal India rejected outright, sold 5,361. Every one of those companies was operating under conditions at least as awkward as Tesla's. One of them had been actively turned away and still outsold Tesla ten to one.

Nothing here was hidden. Tesla knew the terms and four of its competitors took them. What failed was the decision to treat a published condition as something to negotiate away rather than something to build against. I've now sat through some version of that decision more times than I can count. Across the 240-odd India engagements my firm has run, it's the closest thing I have to a reliable early warning.

Where the frame came from

Last weekend I lost four hours at cards to six people in their twenties: my second cousin, his new wife, two more couples from their college years, and me, the only person in the room over forty.

The game began as three-card poker and stopped resembling it almost immediately. Whoever won a hand dealt the next one, and the dealer decided how many cards each player got and what winning meant that round. Highest total, or closest to fifteen, or low card wins, or whatever else they invented on the spot and announced with total confidence. Every hand was a different game, but the condition was always stated out loud, before the cards went out, by somebody everybody could name.

There's an obvious objection here and it's the right one. At that table the deal rotated among equals. In India it doesn't. The state deals every hand and never passes the pack. True, and it doesn't matter, because the rotation was never the point. The point is that the condition gets declared before play rather than discovered afterwards. A permanent dealer who announces the rule in advance is something you can plan around.

A permanent dealer who changes it after you've bet is something else, and India has done that too. So one limit, stated plainly: re-declaration is survivable, retrospection isn't. When India amended the Finance Act in 2012 to tax a transaction the Supreme Court had already ruled on, that was a rule reaching backwards into a hand already played. Vodafone spent the better part of a decade unwinding it. Nothing in this piece applies to that and I won't pretend otherwise.

Somebody translated

I was terrible at the game. Comprehensively terrible. The whole room found me funny, and I got helped through round after round by people half my age.

Then one of them sat with me for twenty minutes and translated. He knew I played Texas Hold'em, so he didn't explain the new game. He couldn't have, it changed every four minutes. He mapped one onto the other instead. Hand strength still matters. Position is meaningless, there's no betting round. Bluffing is worthless when the condition is closest to fifteen. Your instinct to hold the good cards loses you every low-card hand.

After that I won about one hand in five, and on the rest I stopped losing badly. He hadn't given me the rules. The rules expired at the end of each hand. He'd given me the mapping: which of the things I already knew still transferred, and which would sink me if I trusted them.

This is the part where I'm supposed to tell you that translation is what my firm sells. I won't, because it would be the wrong lesson to take from that table. He wasn't selling anything. He noticed I was drowning and gave up twenty minutes of a game he was winning.

Three tests you can run this week

What that twenty minutes actually gave me was a short list of things to check before playing a hand. I've been carrying the business version of that list around for years without writing it down, and the card table is what made me finally do it.

Here they are. If your India operation is missing its business case, they take a day between them.

Measure your absorption clock. Both of them. Take the last three policy changes that hit your India business and time two separate gaps.

The first is the warning clock: notification to operational detail. On the EV scheme that ran fifteen months, notified March 2024, detailed guidelines June 2025. Governments are slow at this and it's the part everybody notices.

The second clock is the one that decides outcomes. Guidelines landed 2 June 2025. The application portal opened on the 24th and shut on 21 October. One hundred and twenty days to commit ₹4,150 crore. Mercedes, Volkswagen-Skoda, Hyundai and Kia all got inside it.

The working threshold is this. You get roughly four quarters of warning and roughly one quarter to act. If your organisation needs longer than a single quarter to turn a specific published condition into a signed commitment, the warning period is worth nothing to you. Tesla had both clocks running for two years and used neither.

I've measured this properly in automotive. I'd expect it to hold anywhere the state runs a formal application window with a closing date, which is most incentive schemes in this country. I haven't tested it in pharma or software and I'm not going to claim I have.

Find your local decision ceiling. When a condition changes, who inside your India entity can commit money alone, without a call to another continent? Name the person and name the number.

The honest answer, for most foreign-owned India entities I've worked with, is under ₹25 lakh. Where it moves, it moves with tenure — the entities that can sign meaningfully more are almost always the ones that have been operating here a long time. It is rarely granted at entry, which is precisely when it matters most.

Set that against the paragraph above, but carefully, because the obvious reading is unfair. Nobody expects an India MD to commit ₹4,150 crore alone. A number that size belongs at group board level in any company in any market, and it should.

The 120 days was the filing window, not the decision window. The decision window was the fifteen months of warning, and the companies that made the deadline used it to get board-ready. Mercedes didn't decide in four months. Mercedes decided during the warning period and filed in the four.

The ceiling isn't measuring signing authority. It's measuring whether the India entity has standing in group planning — whether it is a business with a seat at the table or a sales office that reports numbers upward. That's what determines whether anybody was preparing while the first clock ran. An entity trusted with ₹25 lakh is not an entity whose head of market gets asked what the group should do about a policy announcement fifteen months out.

The ceiling isn't a governance detail. It sets your absorption clock. Every rupee of authority you withhold from the India entity is time you add to the second clock, and the second clock is the one that closes.

Audit your transferred instincts. List the operating assumptions your India plan inherited from the last market you entered, and mark the ones you've actually tested here rather than carried in.

Four break more often than everything else combined: what people will pay, who the buyer actually is, how long the decision takes, and whether your differentiator is still a differentiator here. Pricing is the one companies expect to get wrong, so they flex on it. The other three are the expensive ones because nobody thinks to check them.

The fourth is the one that frustrates me most, and analysts miss it constantly. Differentiation is treated as a property of the product. It is a property of the market you are standing in.

I watched this play out with a global no-code software platform preparing to enter India. They arrived with a specific and reasonable conviction: India's hyperlocal economy was throwing up thousands of small, particular problems, so there had to be enormous demand for a GIS-native app builder that non-technical people could use to solve them themselves.

We put it to the people who'd actually have to adopt it. Web developers, freelancers, small IT service firms, training institutes, students. Two things came back. The Indian no-code market was already crowded. And the GIS capability that anchored the entire pitch — the thing that made the product special everywhere else — was reproducible here with a Google Maps API call by anyone who wanted to bother.

Which left no durable advantage, and therefore no defensible return on the development and maintenance the roadmap assumed. The product wasn't weak. The moat simply didn't travel.

That's the shape of it, and it's invisible from head office, because the assumption carries the best credential there is. It worked last time.

The point

Companies that struggle in India are rarely the ones that don't understand the rules. They're the ones waiting for the rules to hold still before they sit down.

They aren't going to hold still. Every condition in this market gets declared by somebody, on a date, in public, and revised faster than your planning cycle was built to move. That is the specification. The question isn't whether it's fair. The question is what your absorption clock reads, what your India head can sign alone, and which of your imported convictions is about to cost you a year.

Tesla had two years of warning, four competitors demonstrating the answer, and the largest EV market growth story in the world in front of it. It spent the time arguing about the conditions instead of playing the hand.

I lost most of the hands I played that evening. I'd go back tomorrow.