I Tried to Value Zepto Using a DCF. Here's Why the Model Broke.

What is a company that has doubled its revenue two years running, and still lost almost six thousand crore rupees this year, actually worth?

That was the question I sat down to answer a few weeks before Zepto's IPO, armed with nothing but a discounted cash flow model and the numbers in its own prospectus. I thought it would take an afternoon. It took a week, and at the end of it, I didn't have an answer. I had a range so wide it barely qualified as one: a floor under $4 billion, a ceiling somewhere in the mid teens of billions, and no non-arbitrary way to close that gap.

This is a post about why the model broke, and about why breaking was actually the most useful thing it did. A quick note before I start, because someone always raises it: yes, a DCF is a tool built for mature, listed companies with predictable cash flow, and Zepto is a private, pre-IPO business burning cash by design. Using it here isn't a mistake I need to apologise for. That's the entire point. I'm not trying to prove I can run a clean, uneventful DCF. I'm trying to show where the standard tool breaks, and why the thing it can't price is exactly the thing the IPO is trying to price.

Start with what's actually real.

Module 1 of Wharton's Fundamentals of Quantitative Modeling course opens with a distinction that sounds trivial and isn't: a model is only as honest as its separation of data, meaning what's given, disclosed, and sourced, from judgment, meaning what's assumed, projected, and chosen. Skip that split and you don't have a model, you have a guess wearing a spreadsheet. So before a single DCF number appears, here's the clean divide, straight from Zepto's updated draft red herring prospectus, filed with SEBI in June 2026.

Revenue from operations climbed from ₹4,454.52 crore in FY24 to ₹11,109.94 crore in FY25 to ₹22,623.58 crore in FY26, a doubling in the most recent year alone, with order volumes compounding at roughly 119.5% over the two-year stretch. Losses grew too, from ₹1,214.79 crore to ₹4,699.71 crore to ₹5,905.19 crore, but slower than revenue in the latest year, and that matters. By March 2026 the network had grown to 1,139 dark stores across 66 cities, serving 4.79 crore annual transacting users, up 25% year on year, and processing more than 1.75 million orders a day. Unit economics are quietly improving as well: the adjusted EBITDA loss per order fell from roughly ₹136 in FY25 to about ₹79 in FY26. And advertising, the highest margin rupee in the whole business, brought in ₹1,636 crore in FY26, more than thirty times its FY24 level, and now makes up close to 7.8% of net receivables value, up from just 1.1% two years earlier.

What isn't data is every one of those trajectories extended forward by even a single year. How long does revenue keep doubling? When does the per-order loss cross zero? Does advertising scale into a real profit engine or plateau? None of that is disclosed, because none of it is known, not to me, not to the lead bankers, arguably not to Zepto itself. That's the judgment column, and a DCF is almost entirely built out of it.

A discount rate with no data behind it.

A standard DCF discounts future cash flows at a weighted average cost of capital, and the cost of equity piece of that comes from CAPM: the risk free rate, plus beta times the equity risk premium. Wharton's Business and Financial Modeling Capstone walks through exactly this build, and it's also where the first crack shows up. CAPM needs a beta, a measure of how a stock moves against the market, and Zepto doesn't have one, because Zepto has never traded. So you proxy it from listed comparables with quick commerce exposure, and the two obvious ones are Eternal, which owns Zomato and Blinkit, and Swiggy, which owns Instamart.

Here's the problem. Their measured betas are all over the place depending on the data provider and the lookback window. Pick a low-beta read on Eternal and you land on a cost of equity around 11 to 12%. Blend in Swiggy, or reach for a re-levered sector beta above 1, and you're pushed toward 15 to 16%. Same company, same cash flows, and the discount rate swings four or five points purely on which comparable you trust. Checking against Aswath Damodaran's published India and sector betas is the sane move, since it stops you from anchoring on one broker's number, but the sanity check confirms the discomfort rather than resolving it. The “right” beta for a pre-profit quick commerce business is a judgment call dressed up as a lookup, and on a company whose cash flows sit mostly in the terminal value, a four-point WACC swing isn't a rounding error. It moves the valuation by billions. That swing, on its own, is a finding.

Where the terminal value actually breaks.

This is the heart of it. A textbook two-stage DCF forecasts explicitly for a handful of years, then attaches a terminal value that assumes the business settles into a stable, perpetual growth rate, usually pegged near long-run nominal GDP, so 5 to 7% for India. The terminal value typically accounts for the majority of the total valuation. For a mature company that's fine, because the “settling down” assumption is roughly true. For Zepto it isn't roughly true, it's the whole question. Zepto's valuation only works if revenue keeps growing far above GDP for several more years before it can plausibly mature into a single-digit-growth, grocery-logistics business, and the year you assume that transition happens isn't a footnote. It's the dominant driver of the entire output.

Wharton's growth models module, exponential versus logistic “S-curve” growth, is the right tool here: instead of a hard cutover from 100% growth to 6% growth, you fit a curve that decelerates smoothly. But that just relocates the judgment, because now you're choosing the curve's inflection point, which is the same “when does it mature” question wearing a different costume. To see how badly this bites, I ran the model as a two-variable sensitivity check, varying the discount rate against the year in which high growth gives way to terminal growth. The dollar figures in each cell depend on intermediate assumptions I held fixed elsewhere, so don't read too much into any single number, but the shape of the grid is the real finding:

Look at the two corners. The “high discount rate, quick maturity” corner values Zepto around $3 to 4 billion. The “low discount rate, long runway” corner values it north of $18 billion. That's close to a 5x spread between two entirely defensible sets of assumptions, and I cannot tell you which one is correct, because the difference between them is a bet on how the Indian quick commerce war ends, a question the DCF takes as an input when it's actually the thing I want the model to tell me.

This is the mechanical version of the argument Aswath Damodaran makes in Narrative and Numbers: for young, high-growth, cash-burning companies, a single-point DCF output is close to meaningless without the story that justifies the growth assumptions behind it. The number isn't the analysis. The narrative you had to assume to get the number is the analysis.

What the market already knows that the model doesn't.

So the intrinsic model gives me a range, not a value. The natural next move is to ask what the market actually pays for businesses like this, and unlike a year ago, we now have real Indian comparables, because the sector has finally started listing. Meesho went public in December 2025 at a 46% pop on debut, and as of this writing trades at a market capitalisation of roughly ₹87,600 crore against FY26 revenue of about ₹12,600 crore, close to a 7x revenue multiple. Nykaa trades richer still, near 9x sales on FY26 revenue just over ₹10,000 crore, but Nykaa is actually profitable, with a net profit of roughly ₹204 crore for the year, so it's arguably being priced on a quality premium rather than a growth premium. FirstCry sits at the other end, closer to 1.3x sales, dragged down by slower growth and a business that only recently turned free cash flow positive.

So the market's revenue multiple for an Indian consumer-internet platform that's still proving profitability spans roughly 1x to 9x. Anchor Zepto somewhere in the Meesho-ish middle, say 5 to 7x on ₹22,600 crore of FY26 revenue, and you land around $13 to 19 billion, above almost every cell in my DCF grid, which tells you the public market is pricing in optimism the intrinsic model struggles to justify on cash flows alone.

But the sharpest data point isn't the listed comps, it's Zepto pricing itself. Follow the trajectory. $7 billion at its Series H, a roughly $450 million round led by CalPERS in October 2025. Then, as the IPO neared, that number started sliding: reports in mid-July 2026 put the anchor book being finalised at around $5.1 billion, with foreign institutions reportedly signalling interest closer to $4.5 billion pre-money, and domestic institutions said to peg it lower still, in the $3 to 3.5 billion range. That isn't a market disagreeing with my DCF. That's a market disagreeing with itself, sophisticated, well-resourced investors landing anywhere from roughly $3.5 billion to $7 billion on the identical asset within months. When the smart money's own range is that wide, a single-point DCF was never going to be narrower.

Why it broke, and what that actually tells you.

The DCF didn't fail because of a formula error. It failed because a DCF assumes the hard part is already decided: what kind of company this becomes, and when. Terminal value is that assumption, wearing a Gordon growth formula. For a mature utility, that's fine. For a pre-IPO, hyper-growth business fighting a cash-burn war against two better-capitalised rivals, the “stable state” isn't a modelling input. It's the open question the entire IPO exists to price.

Which points at the tool I actually needed and didn't use: real options. A company like Zepto isn't really one deterministic cash flow stream to be discounted. It's a portfolio of bets, winning quick commerce outright, scaling advertising into a genuine profit engine, hitting breakeven on some particular timeline, and equity is a claim on the payoff if those options land. That optionality is precisely what a two-stage DCF cannot capture, and precisely what the market prices anyway, which is why the market's number tends to sit above whatever the model alone can justify.

I'm not going to run an options-pricing model on a grocery delivery company in a blog post. But naming that the framework exists, and understanding why it's the natural next step, is the difference between having run a DCF and actually understanding what a DCF is and isn't for. That, honestly, was the whole exercise. Growth models, sensitivity analysis, and CAPM didn't help me nail down what Zepto is worth. They helped me understand, precisely, why it can't be nailed down with these tools alone, and that turned out to be the more useful thing to learn.

As of this writing, Zepto's anchor round is still being finalised and the IPO remains on track for the July to September 2026 window, at a valuation well below its $7 billion peak. When the book finally builds and a price actually prints, the gap between wherever it lands and the range floating around beforehand will be worth a follow-up post of its own. If the market can't agree on the number either, that's not the DCF's failure. That's the point.