Process, Valuation & What Survives Every Market Cycle | Netra Anniversary Edition

ELI5/TLDR

DSP’s Sahil Kapoor walks through how to value five different asset classes, each with borrowed frameworks from twenty practitioners. The unifying idea: a valuation is not a forecast, it is a structured disagreement with the price the market is quoting. Every asset class needs its own tool — you cannot run a DCF on a stock, a bond, gold, and a private loan and expect the same machine to work. The one thing that endures across cycles is not any number but the process for arriving at it.

The Full Story

Why an anniversary edition has no charts

DSP’s Netra is normally a charts-and-data letter. The anniversary edition is deliberately the opposite: no market data, no forecasts, just the “unchanging parts of investing.” The framing device is the Ise Grand Shrine in Japan — 1,300 years old, yet the physical structure standing today was built within the last twenty years.

This structure is rebuilt every 20 years… the physical objects can change but the process to build this structure it survives.

An apprentice learns the build in his twenties, masters it by sixty, and hands it down at seventy. The shrine is not engineered for durable materials; it is engineered for durable knowhow. Kapoor calls this anti-fragility in practice and uses it as the thesis: markets, rates, earnings, and valuations all decay, but the process of knowing how to value must survive.

His definition of investing is narrow on purpose: buy an asset below intrinsic value, wait for it to revert, collect the gap. Trading, momentum, speculation are not condemned — they are simply ruled out of scope. And if that is the only way to invest, then the ability to value assets is the whole game.

The universal lens: does it generate cash flow?

The opening framework comes from Rajeev Thakkar (PPFAS). The first question is not whether an asset is popular or scarce or fashionable — it is does it generate cash flow. If yes (stocks, bonds), valuation reduces to estimation or calculation. The math is just compound interest: know three of {future value, time, required return, present value} and the fourth falls out.

The split that matters is bonds versus stocks. A bond’s future is defined — maturity, coupon, cash flows are all known, so you are mostly calculating. A stock is, in Kapoor’s phrase, “a bond with an unknown coupon and no maturity date,” so you are estimating. Gold and art generate no cash flow at all — there you are forecasting behaviour, which needs an entirely different lens.

You can’t use a DCF on stocks and bonds and gold and credits and come up with an answer… a bond is not a stock. A stock is not gold.

Thakkar’s two filters: anchor estimates in reality (a utility’s cash flows are more predictable than a biotech’s, so low predictability demands a bigger margin of safety), and trees do not grow to the sky (no supernormal margin deserves a high multiple in perpetuity — not every movie is a Baahubali). And no formula protects you against bad people, so an integrity filter sits underneath everything.

Equities: price, quality, expectations

The recurring sin in equities is starting with a multiple — low P/E good, high P/E bad. Kapoor’s rebuttal: a multiple is not a valuation, it is a compressed story. The right question is what must be true for that multiple to make sense. Three frameworks attack this.

Abhishek Singh (DSP) — valuation bands. Three pillars: valuations matter, nobody knows, structure beats activity. Since the exact value is unknowable, you build a band — the lower edge implies conservative assumptions (low ROIC, modest growth) and is where you buy; the upper edge implies peak-everything extrapolation and is where you trim. The middle is the hard part: it tells you to do nothing.

Most investors… don’t fail because they do not know enough. They fail because they cannot sit still.

Inactivity is reframed as discipline, with a John Milton line (“they also serve who only stand and wait”) doing the heavy lifting. The point of valuation is not constant activity — it is knowing when activity is justified, which is only at the extremes.

Harish Krishnan (Aditya Birla Sun Life) — good to great. A Pareto split: ~5–10% of companies are truly great, ~25–30% are good, two-thirds are mediocre. Greatness is built on inherent business advantages (customer obsession, first-principles thinking, attracting talent, hard-to-replicate scarcity) — not Excel-backtested factors. The moat, he says, lies “between knowing and doing.” But the sharp insight is in the returns data: greatness is usually already priced in. The asymmetry comes from transitions. From a 2001–26 study, good-to-great transitions produced ~18% excess return over three years, bad-to-good ~17%, and downgrades roughly −10%.

Market rewards improvement more than the labels.

So the question is not “is this great?” but “what is changing — is ROIC rising, is cash generation more reliable, is capital allocation improving?”

Ramnik Kundra (DSP Pension) — reverse DCF. His line is the thesis statement of the whole letter:

I do not start a valuation to find an answer. I started to find a disagreement with the price.

A target price creates false precision. Better to ask what the current price is already assuming. An expensive-looking stock can be a buy if embedded expectations are too low; a cheap-looking one can be a trap if the price correctly discounts falling profitability. His worked example is Nvidia: at ~$4.5T market cap, ~$97B FCF, ~$216B revenue, the reverse DCF spits out an asking rate of ~18% FCF growth for ten years (to ~$500B). The 18% is an output, not an assumption. Context for scale: total global data-center capex is ~$600B/yr, so the price asks Nvidia to eventually out-earn the entire current data-center economy. Apple compounded FCF at ~23% — but from a ~$10B base, versus Nvidia’s ~$100B. The scenario table: bull case (18% for 10 yrs) values it ~$5.2T — still below the then-price; bear case (8%) ~$2.4T — a ~55% downside; with a 30% margin of safety you pay ~$2.4T.

Buy at any price is an illogical illusion.

Debt: which anchor is alive?

Bonds are mathematically cleaner — defined cash flows, the dominant variable being the interest rate. But the anchor keeps shifting. Suresh’s framework (DSP) says bonds do not follow one standardised model. Sometimes term spreads dominate, sometimes real yields, sometimes the India–US differential, sometimes BoP/current account, sometimes RBI liquidity and government-borrowing supply.

Every macro framework has a shelf life.

The skill is not reading every data series but identifying which one is currently driving the market, and constantly re-asking whether that framework is still alive. The practical read offered: when real yields are high, inflation controlled, and the rupee not under BoP stress, duration becomes attractive — but flows, an oil spike, or currency pressure can flip the same duration trade. Current GSec driver: global flows plus RBI’s ability to manage liquidity. The point is knowing what the yield is compensating you for.

Commodities: separate the cash flow from the cycle

Two sub-frameworks. Commodity businesses (Ankita, on memory semiconductors): the starting point is never P/E or DCF — those are outputs. The input is the supply–demand balance. Memory (DRAM, NAND, HBM) is a commodity — standardised product, few interchangeable players (SK Hynix, Micron, Samsung), limited pricing power. The deck tracks inventory days (peaked in the 2023 glut, leaner by 2025 → supply discipline), HBM decoupling from legacy memory, and gross margins troughing in FY23 and recovering through FY26. From FY24–27, demand exceeds supply across all segments. The checklist: is demand > supply, can supply not catch up, are inventories falling, prices rising, margins improving, capital discipline intact? With only ~7 listed names, security selection on financials and inventory days becomes the analyst’s job.

Do not value a commodity business like a compounder… a cheap stock can be expensive at the wrong point in cycle and an average business can be a good investment at the right point in cycle.

Gold and silver are valued as money, not cash flow. The DSP framework treats gold as the ultimate monetary base: total money supply divided by all gold ever mined (~~7.1bn troy oz). It uses US M2 (~~$22.7T) as the base — dollar is the primary reserve currency; M2 not M3 because M3 is contaminated by credit and they want the monetary base, not credit multipliers. It excludes China/Russia/India (currencies not universally accepted — and converting into rupees later lets inflation account for their debasement), and includes half of Eurozone reserves as a conservative proxy for the offshore Eurodollar market. Output: a fair value around $4,300–$4,500. Silver gets no monetary anchor (disconnected for 60–70 years) — it is valued off the gold-silver ratio, base 60, but margin of safety only appears when the ratio hits 80–85. Kapil Gupta (Nuvama) has a separate USD-balance-sheet framework that sees gold as even more undervalued.

Currency: a dashboard, not a number

Madhavi Arora’s (Emkay) framework opens with an admission: no single FX model works consistently. The toolkit — PPP, interest-rate parity, REER/BEER/FEER, balance of payments, terms of trade, RBI’s reaction function, capital flows — each useful in a different regime. PPP is a long-term anchor (6% India vs 2% US inflation → ~4% INR depreciation/yr) but useless for short-term timing. REER is a valuation gauge; India’s is near decadal lows (~90), suggesting the rupee is no longer expensive — but the caveat is that a REER move driven by lower food inflation does not improve competitiveness. The surprise of the last 18 months: four-decade-low inflation, a benign current-account deficit, and yet major rupee weakness.

FX models… are guidance tools. They are not turning point machines… It’s a dashboard of pressures.

The practical conclusion: with REER near lows and BoP stress manageable, if oil stays contained and flows improve, rupee assets (large caps, long-duration GSec, India allocation) become attractive. Currency is “the balance-sheet expression of a country.”

Credit: the job is to avoid losers

Kept for last because it is the most unforgiving. A Kapoor (True North) frames credit as asymmetric: upside capped, downside permanent. There is no multibagger in private credit to cover a bad loan, so the job is not picking winners — it is avoiding losers. Three truths: collateral is not a thesis, it is a recovery tool and a behavioural deterrent (if the deal only works because of collateral from day one, underwriting failed); enforcement in India is slow and costly, so collateral aligns but does not protect; and investors experience portfolios, not deals.

The Plan A/B/C structure: A = cash flows (lend to high-quality companies, short tenors, regular coupons, underwrite cash flow not asset cover); B = collateral (recovery tool only); C = portfolio construction (concentration limits, sector spread, tenor discipline, vintage mix, more deals not fewer). The data point: international credit funds run ~2–4% single-deal exposure; India still commonly runs 10–15%.

12% yield does not protect you from a 15% position going bad.

In India, Plan C is not optional — it is the strategy. Credit valuation is ultimately about survival.

The single message

Do not worship price, interrogate price.

Do not outsource judgment to a multiple, a thesis, or a narrative — understand what it is hiding. The frameworks differ by asset class, but the discipline is the same: figure out what the price already assumes, then decide whether you disagree.

Key Takeaways

A valuation is a disciplined disagreement with price, not a forecast or a target.
First question for any asset: does it generate cash flow? If yes → estimate/calculate (DCF-able). If no (gold, art) → forecast behaviour, different lens entirely.
A stock is “a bond with an unknown coupon and no maturity date” — same math, far higher estimation error.
A multiple is a compressed story, not a valuation. Ask what must be true for it to make sense.
Valuation bands beat point estimates: buy at the conservative edge, trim at the aggressive edge, do nothing in the middle. The middle is the discipline.
The market prices in greatness already — excess returns come from transitions (good→great ~18%, bad→good ~17% over 3 yrs; downgrades ~−10%). Ask “what is changing,” not “is this great.”
Reverse DCF turns the asking rate into an output: Nvidia at ~$4.5T implies ~18% FCF growth for 10 years; the bear case (8%) implies ~55% downside.
“Buy at any price” is incoherent — a great company can be a bad investment once the price has consumed the future.
Every macro/debt framework has a shelf life; the skill is identifying which anchor (real yields, BoP, India-US spread, RBI liquidity) is currently driving the market.
Commodity businesses are valued off supply-demand balance, not multiples; never value a cyclical like a compounder. Timing dominates.
Gold is valued as money (US M2 ÷ all gold mined, + half Eurozone reserves as a Eurodollar proxy) → ~$4,300–4,500 fair value. Silver only via the gold-silver ratio, with margin of safety at 80-85.
FX has no single working model — it is a dashboard of pressures (PPP, REER, BoP, flows), each alive in a different regime.
Credit is asymmetric: avoid losers, don’t pick winners. Cash flow first, collateral as recovery tool, portfolio construction always — in India, Plan C is the whole strategy.

Claude’s Take

This is a genuinely useful letter, and the framing is the point. The Ise Shrine metaphor could have been gimmicky but it earns its keep: the argument that process is the only durable asset is the right thing to say in a foundation edition, and it gives the otherwise-disparate frameworks a spine.

The strongest section is equities, specifically the Kundra reverse-DCF. Turning the discount rate into an output and asking “what is the price already promising” is the single most clarifying move in valuation, and the Nvidia worked example is concrete enough to be portable. Harish Krishnan’s transitions-over-labels data is the other genuinely non-obvious takeaway — that the market already pays for greatness, so the alpha is in the upgrade path, is a useful correction to quality-investing dogma.

The gold-as-money framework is the most contestable bit. M2-divided-by-gold models have been “proving” gold is worth several thousand dollars for a long time; the choice of M2 vs M3, which countries to include, and “half of Eurozone reserves” as a Eurodollar proxy are all judgment calls that move the answer materially. It is a reasonable lens, not a fair-value oracle, and the letter is at least honest that gold needs a different tool — but treat the $4,500 number as one assumption stack among many.

The honest caveat throughout — these are dashboards and bands, not turning-point machines — is what saves it from being a sales document. It is a DSP letter, so the contributors are a mix of DSP and friendly fund managers, and there is no challenge to any framework on offer; you are getting the practitioners’ best case, not a debate. But the meta-message — interrogate price, don’t worship it — is sound, and the per-asset-class discipline is a clean mental filing system. An 8: dense, framework-rich, low on fluff, light on adversarial pressure-testing.

Process, Valuation & What Survives Every Market Cycle | Netra Anniversary Edition | DSP Mutual Fund