Reading the AI Rally: Which Tech Stocks Earn Their Multiples

Every market rally writes the same exam: which prices are earnings arriving early, and which are hope arriving all at once? The artificial intelligence rally, the defining trade of the past two years, is now old enough to grade properly. The analytical framework that separates durable re-ratings from bubbles in rehearsal comes down to three things: cash flow, structural moats, and who is actually paying for the underlying infrastructure.

Start with cash flow

• Cash flow is where the winners and the mere storytellers of the AI boom divide.
• The core of this rally, the specialized chip designers, the big cloud computing platforms and the precision equipment makers, is converting surging demand into revenue and free cash flow at a pace with few precedents in market history. Their multiples may look elevated by historical standards, but they’re attached to real, growing denominators.
• The periphery of the rally is different. Companies whose AI exposure amounts to a paragraph in the annual report trade on the same lofty multiples with only a fraction of the evidence.
• History’s verdict on such gaps is never kind, though the market rarely hurries to deliver it.

Follow the spending

• The central debate is about customer spending and capital expenditure.
• The unprecedented data center build-out is being funded by a handful of mega-cap firms whose own investors increasingly ask when concrete returns will arrive.
• Analysts at major institutions regularly publish two cases: the build-out as the defining infrastructure project of the century with returns to match, or a reckless capital arms race whose aggressive depreciation schedules will eventually punish latecomers and stretched balance sheets.
• Company filings, which can be audited through the SEC’s EDGAR database, let any diligent investor inspect actual capital expenditure every quarter.

What past manias teach

This environment calls for a reliable framework rather than a rigid forecast. The nineteenth-century railway mania built the railways that transformed commerce, just as the dot-com bubble of the late nineties laid the fiber-optic cable that powered the modern internet. The recurring pattern is almost never that the technology was fake. The error is that equity markets make indiscriminate claims on it, pricing every participant as the eventual winner of a market that will crown only a few.

1. First, a technology can be genuinely revolutionary while the stock of its leading standard-bearer is a bad buy at prevailing prices.
2. Second, the second owner of a great asset bought at a silly price almost always does worse over a long horizon than the first owner who bought a good asset at a fair price.
3. Finally, time in the market, rather than timing its peaks and valleys, is what converts being right about technology into actually getting paid for it.

Bottom line

None of this is an argument to stand entirely aside from the defining technology of our era. It’s an argument to hold it the way its own economic history instructs: concentrated among proven cash generators, diversified against the pure storytellers, and sized so that a sharp repricing is a buying opportunity rather than a personal crisis. The rally will create substantial fortunes and unmake others, sometimes reversing the same fortune twice in a short span. A rigorous valuation framework doesn’t need to predict exactly which companies survive. It just ensures that question never becomes an existential threat to your portfolio.

The sensible stance toward the rally is neither blind worship nor outright dismissal. Own the reliable cash flows if you want exposure to the theme, size positions to allow for speculative story stocks repricing violently downward, and remember that in every major technological wave, the infrastructure sellers were paid first and most reliably. The AI rally will eventually end the way its predecessors did, with the technology triumphant and turned into an everyday utility while many early, speculative stock prices look indefensible in hindsight. Telling those two realities apart is the whole job of equity analysis.