Investment Approach: Hold Period of 5 Years
Five years is long enough for business change to become per-share value, and short enough that milestones, return decomposition, and variant perception have to be written down.
The question
The five-year lane asks a narrower question than the 25-year experiment: which public companies have a plausible five-year path to materially better per-share value than the market currently discounts? It is the practical research lane — long enough for real business change to show up in the numbers, short enough that the thesis has to name its milestones in advance.
This is not trading, and it is not a shorter version of the 25-year scorecard. A five-year winner does not need to survive a quarter century or defend a moat against every future entrant. It needs a specific, observable change in the business — an inflection in demand, margins, or capital intensity — to play out inside the window and to be underwritten before it is consensus.
What makes a 5-year winner
The recurring patterns are concrete. S-curve acceleration, where adoption of a product or platform crosses from early to mainstream inside the window. A technology platform shift that re-rates the unit economics of the businesses exposed to it. AI-stack exposure large enough to matter — not a press release, but a measurable share of revenue or capacity tied to a finite-window build-out. A value-chain bottleneck or capacity shortage that the market is slow to price. Operating leverage or a free-cash-flow inflection as fixed-cost investment turns into margin. Underappreciated earnings power. Or temporary controversy obscuring durable economics.
None of these is a thesis on its own. Each is a hypothesis that has to survive return decomposition and the return gate below.
The return gate
The base case must clear a 10% annualized return hurdle without relying on terminal multiple expansion. A thesis that only works if the market re-rates the multiple is a sentiment bet, not a five-year underwriting case.
Return decomposition is required. The expected return has to be separated into its sources — revenue growth, margin change, share-count change, dividend, and any multiple change — and the multiple is held flat or faded in the base case. If the base case is already the consensus expectation, the candidate is not a five-year winner: there is no variant perception to be paid for.
Technology and S-curves
Technology exposure is treated as a finite-window research problem, not a theme. The relevant question is where in the stack the durable economics sit and for how long: the distinction between compute, memory and high-bandwidth memory, networking, power and cooling, software incumbents, and applications matters because each has a different S-curve and a different competitive structure.
Broad AI exposure is not proof of anything. The discipline is to locate the specific bottleneck or adoption curve a company is levered to, estimate the height and slope of that curve, and ask whether the company's position is defensible for the duration of the window — not forever, but for five years.
False inflections
Most five-year theses fail on a false inflection — a change that looks like a durable turn and is not. The recurring traps: the TAM mirage, where a large addressable market is mistaken for a capturable one; pull-forward demand mistaken for sustained adoption; unit economics hidden by growth; platform language without evidence of increasing returns; a margin story without pricing power; and the great-product, bad-stock case, where the product is real but the price already embeds the outcome.
Naming the false-inflection risk explicitly, before entry, is part of the underwriting — the same discipline the 25-year lane applies to thesis-break conditions.
Pilot discipline
The five-year framework is still being pressure-tested before it produces named entries. The first pilot scorecards are NVDA, VRT, and MU, with APP fourth — names chosen to stress-test the framework across different parts of the technology stack, not because they are recommendations.
The GOOG position recorded on the five-year ledger is an audit trail of holding-period decisions — not a recommendation.