"Returns can be increased by extension of the investment horizon rather than by extension of risk."
"The compensation structure of the investment management industry is broken down into artificial time horizons... a future event that will have a decidedly positive impact on the price of a certain security, undisputable though it may be, has limited utility to the professional manager if that event is expected to take place beyond his or her artificial time horizon."
"Fund managers' reluctance to purchase such securities is expressed in the discount rate."
Once you see "the discount rate is just someone's annual review," you can't unsee it.
Reluctance becomes a price. Psychology converts directly into yield.
The asset and the liability have different durations, and if you let the short-duration liability (a one-off order for two hundred units) dictate how you run the long-duration asset (a fab optimized for million-unit runs), you don't get a slightly worse fab. You get a fab that can't compete with the fab next door that didn't make that mistake.
Every market with a long tail has the same secret: someone is being paid not for what they make, but for refusing to optimize the way everyone bigger than them was forced to.
A fab that makes chips at scale and a fab that makes chips in batches of two hundred are not different sizes of the same business... they're different businesses that happen to share a word. "A production run for one chip in a large fab can last up to three years," and retooling for something new costs you "downtime and lower yields that hurt profit."
That's not a complaint about the semiconductor industry, it's a description of how the industry's capital structure works: you build a $20 billion fab, and the only way to amortize $20 billion is to run the same recipe, at the same yield, for as long as physically possible. The fixed cost doesn't care what's fashionable. It cares about utilization. So when Texas Instruments or NXP discontinues a chip, they're not making a product decision. They're making a balance-sheet decision dressed up as a product decision. The chip didn't get worse. The demand for it didn't disappear. What happened is that the demand fell below the threshold where it's rational to dedicate fab capacity to it, and "rational" here is doing a lot of quiet work... it means rational given the asset they already built, not rational in some abstract sense. A semiconductor company with a $20 billion fab and a chip that ten customers still want isn't choosing between making money and not making money. It's choosing between a small amount of money on this chip and a larger amount of money on the next one, using the same machine. The opportunity cost is the whole decision.
This is the inside-out part, and it's worth sitting in for a second, because the easy read of "OEMs orphan their own customers" makes the OEMs sound careless. They're not careless. Hatcher, who spent six years inside Samsung's fabs, says it plainly: "No one's going to pick up the phone for a couple hundred chips." "The fundamental mismatch is that semiconductor companies go out of business if they have high mix, low volume."
https://t.co/dFJPbRVEim
Exchanges are the closest thing capitalism has to a perpetual motion machine. The product line is a tollbooth that prints money on every transaction it doesn't have to manufacture. A million orders cost a computer roughly what a thousand do... the marginal trade is nearly free, but the marginal fee isn't. That gap between cost curve and revenue curve is the whole business model, and it's why an exchange with a two-decade track record beats its home index almost everywhere you look — New York, Tokyo, Hong Kong, London, Singapore. Different regulators, different currencies, same physics.
Pricing power you can't write a contract around is just a head start. Management always agrees the moat is real right up until the moment they're asked to underwrite it themselves.
The exchange is the only business that gets paid for being indifferent to its own outcome... a derivative on trading itself, settled in cash, struck at zero.
India's share of the global aerospace supply chain is moving from roughly 2% toward 10% directionally.
This isn't China+1 in the consumer-electronics sense (cost arbitrage, tariff avoidance). Aerospace supply chain diversification is driven by a different and arguably more durable logic... single-source risk concentration. Boeing and Airbus learned during COVID and subsequent supply shocks (Spirit AeroSystems' 2024 quality crisis being the most visible recent example) that having too much of the supply base concentrated in too few qualified suppliers is itself a tail risk to production rate. The OEMs are actively seeking to qualify new geographic suppliers not because India is cheaper (though it helps), but because supply chain single-points-of-failure became existential during 2020-2024. This is a structurally different and stickier kind of tailwind than cost arbitrage, because it doesn't reverse when Indian wages rise... it's about resilience, not just price.
The 5-10 year qualification cycle is the other side of this coin... it means this reshoring wave, once a supplier is qualified, is sticky almost by construction. OEMs don't re-shop qualified aerospace suppliers the way they'd re-shop a commodity component vendor, the switching cost itself becomes the moat. So the macro trend here isn't really "India is getting cheaper aerospace work" it's "India is becoming permanently embedded in aerospace supply chains because the cost of re-qualifying elsewhere exceeds any savings from doing so."
Airbus + Boeing forecasting ~42,000-48,000 new aircraft over 20 years, passenger traffic doubling by 2045-50, airline industry crossing $1 trillion in 2026. A meaningful chunk of that 42,430-aircraft demand figure is fleet replacement, not net fleet growth, which means it's substantially insulated from any single economic cycle. Replacement demand for aging aircraft doesn't go away in a recession the way discretionary new-route expansion does.
Take-rate is sustainable right up until the supply side does the math on what building their own distribution would have cost... Nykaa's owned brands are that math, run by the house, against the house's own tenants.
Liquidity is a marketplace's only real asset, and the fastest way to spend it is to compete with the supply that built it.
So: is House of Nykaa a moat, the way the case for the stock needs it to be? For the next five years, probably yes, the data advantage is real, it’s compounding, and nobody else in Indian beauty has 45 million customers’ search history to mine. Past five years, the answer depends entirely on whether any outside brand does what Dot & Key is doing by accident: builds enough demand outside the app that the landlord needs the tenant as much as the tenant needs the landlord. Most won’t. The date palm grows tall either way. The open question is how many of the travelers underneath it ever plant a tree of their own before they need the shade.
Nykaa is a beauty company, in the sense that it sells beauty products, the way a casino is a hotel company in the sense that it has rooms.
Bada hua to kya hua, jaise ped khajoor… panthi ko chhaya nahi, phal lage atidoor. Kabir, five centuries early: what does it matter that you grew so tall, like the date palm, the traveler gets no shade from you, and your fruit hangs too far for him to reach. Keep that couplet in your pocket. By the end, it will have stopped being about trees.
Here is the anomaly. Nykaa spends an increasing share of its own revenue paying Instagram, Google, and 170,000 creators to find customers, and then turns around and charges 4,200 beauty brands for the privilege of being found by those same customers, inside its own app, using data those brands generated by selling on the platform in the first place. Money goes out one door to rent attention. A larger amount comes back through a different door, from people who have no door of their own. A business that can charge twice for the same audience, once coming in and once going out, is not running a beauty company. It is running a toll road that happens to sell lipstick.
Here is the steelman, given its full weight before it’s dismantled: every retailer in history has done some version of this. Walmart’s private label, D-Mart’s house brands, Big Bazaar before it… own the shelf, learn what sells, build your own version, nobody calls this predatory, they call it retail. The third-party brands chose Nykaa over building alone because Nykaa, even taking a cut, gets them to revenue faster than the alternative. That’s not exploitation. That’s a rational trade, made with eyes open, by founders who ran the math and picked the faster path. The objection only works if you assume the brands had a realistic alternative that was actually better, and for most of them, at the stage they joined the platform, they didn’t.
The steelman is right about the choice and wrong about the information. Walmart’s private label doesn’t know, before it launches, exactly which competitor SKU is converting at what price in which pin code, because Walmart’s checkout data doesn’t carry the search query that preceded the purchase. Nykaa’s does. The asymmetry isn’t that the landlord builds its own brand… every landlord eventually does that. It’s that this landlord builds with a map of exactly where the tenant’s foot traffic was heading before the tenant noticed it was being watched.
Free infrastructure creates competitive applications. The entity that sits between them, making the connection work reliably, becomes the most durable business in the ecosystem.
In the 1990s, as the internet was being built, the battle everyone watched was between infrastructure players and application players.
The infrastructure companies were critical and modestly valued. The application companies were exciting and wildly overvalued.
What few people noticed was the emergence of a third category… companies that weren’t building the internet and weren’t building on the internet, but were building the tools that made the internet commercially usable. The payment processors that connected the banking system to e-commerce. The security companies that made consumer trust in online transactions possible. The logistics companies that connected digital demand to physical fulfillment.
These companies — early PayPal, Verisign, FedEx in its transformation, were not glamorous. They were solving operationally complex, commercially invisible problems in the gap between infrastructure and application. Over the next two decades, many infrastructure companies became commodities. Many application companies failed or were acquired. The middleware companies, the ones that had built genuinely irreplaceable connective tissue between the two layers, quietly became some of the most durable and profitable businesses in the digital economy.