The Insight Labs Daily.

Rs 325 crore of the fresh issue goes towards supporting disbursals under default loss guarantee arrangements — the structure where a digital lender promises a partner bank or NBFC that it will absorb the first slice of losses. Another Rs 250 crore goes into Whizdm Finance, its own lending subsidiary. Both are capital for carrying credit risk, not for growth marketing.

Accel holds 21.89% going in and Tiger Global 13.79%, and both are selling into the OFS alongside the founders. A smaller fresh issue with a trimmed OFS usually means the book was tested and the price was the sticking point. Moneyview chose the valuation over the size.

The ratio is the interesting part. A pipeline two-thirds the size of the operating base means the growth is not being funded off IHCL's own balance sheet — it is management and franchise contracts, where somebody else builds the hotel and IHCL supplies the brand, the booking system and the operating standard.

That is how a hotel company stops behaving like a real-estate company. The number of doors carrying the flag starts to matter more than the capital tied up per room, and the risk of a soft demand year sits with the owner rather than the operator. The cost is control: 255 signed hotels is 255 owners who each have to be held to the standard the Taj name implies.

The distribution tells you where frozen food is actually being bought. Protein Pantry sells direct-to-consumer in Delhi, Mumbai, Bengaluru and Jaipur, and through Blinkit, FirstClub and Flipkart Minutes in four cities. Pune, Kolkata, Lucknow, Chandigarh, Ludhiana and Chennai are meant to follow on quick commerce by the end of the year — not through general trade.

India's protein market has been built around supplements and snack bars, which are shelf-stable and travel easily. A frozen ready-to-cook product is neither. It needs cold chain at every step, which is why the factory matters more than the marketing budget here, and why a ten-minute delivery network is currently the only shelf that can carry it at any real scale.

Every one of those 350 doors sits inside somebody else's commercial arrangement — a franchise partner, a department store, a marketplace listing. The brand gets the sale and a margin. It does not get the shopper's email, their size history, or the reason they walked out without buying.

A first-party site changes what the brand knows before it changes what it sells. The measure a year from now is not the revenue share of in.tommy.com, which will be small, but whether the full-price mix improves — because owning the storefront is largely a way to stop discounting on channels where you do not set the price.

The framing is narrower than most AI assistant pitches, and that is the point. Instead of asking a user to describe a task, it treats the sentence 'I'll send that over by Thursday' as the instruction. The input already exists; nobody has to learn a new interface to produce it.

The risk sits in the same place as the idea. An assistant acting on what it infers from your email is useful only if its reading of intent is right most of the time, and safe only if the human review step is real rather than a checkbox. The economics also assume inference cost falls faster than usage grows, which is why the round names model processing as a line item at all.