Reading a P/E ratio (trailing vs forward)

Rule of thumb: forward P/E is the more decision-relevant number (you own a stock for its future earnings), which is why the Bubble dumbbell draws it in green — but it's an estimate, so treat it as a forecast, not a fact.

When a stock's forward P/E is far below its trailing P/E, it means analysts expect earnings to grow sharply over the next year. Because P/E is a fraction, the growth shows up in the denominator: if the price barely moves but earnings jump, the ratio collapses.

DDOG, same price, different earnings base:

  trailing  =  price ÷ $0.38  ≈  680×   ← tiny past profit
  forward   =  price ÷ $2.42  ≈   67×   ← projected profit

  earnings grow ~6× → the P/E "compresses" ~10×

So a 680× trailing multiple on Datadog is a denominator artifact, not evidence of pure mania — the company spent heavily to grow, so past GAAP profit was thin. By contrast NVDA shows almost no gap: its earnings already caught up to the price, so trailing and forward sit close together. The size of the amber gap is itself the signal — it's how much future growth the market has already paid for.

Forward estimates aren't guesses pulled from the air. For many companies they're surprisingly well-anchored:

• Management guidance. Companies publish official outlooks (e.g. Datadog guided full-year 2026 non-GAAP EPS to roughly $2.36–$2.44).
• Recurring revenue. Subscription (SaaS) businesses can be modelled from existing contracts and net revenue retention (often >120%), so next year's sales are largely already booked.
• Operating leverage. As revenue grows, fixed costs don't grow as fast, so a rising share of each new dollar drops to profit, and earnings grow faster than sales.
• Analyst consensus. Dozens of analysts track the name and continuously revise a consensus estimate; "forward P/E" usually divides by that consensus.

The flip side: estimates can be wrong, and they get cut. If growth disappoints, the forward P/E you were paying quietly turns out to have been much higher than it looked.

A high P/E isn't automatically "expensive" or "bad" — it means the market is paying a premium today for expected future growth. The danger is that a rich multiple prices in perfection: it only works if the company keeps growing fast for years, so a single disappointing quarter can re-rate the stock violently. That's the core lesson of the dot-com giants — Cisco was a great business at 150–200× and still fell ~89% when its growth normalised. Being right about the company is not the same as being right about the stock at that price.

If a company has no positive GAAP earnings, the denominator is zero or negative and the ratio is meaningless — shown as n/m ("not meaningful"). On the Bubble page, CRWD and 2000-era Amazon and Yahoo! land here: they trade on revenue, ARR or price-to-sales instead. That's the closest analogue to the 2000 profitless-IPO froth — a valuation built on a metric that excludes real costs.

1. Compare within a sector, not across. A 30× software name and a 30× bank are not "the same price" — growth, margins and capital intensity differ wildly.
2. Watch the forward-vs-trailing gap. A big gap tells you how much growth is already priced in — and therefore how much has to actually happen for the stock to just tread water.
3. Cyclicals invert it. Miners, automakers and chip fabs often look cheapest (low P/E) at the top of their cycle, when earnings peak — and dearest at the bottom.
4. Pair it with growth (PEG) and profitability. A 60× P/E growing 40% a year is a different animal from a 60× P/E growing 8%.

See the Earnings Bubble view to watch all of this live — today's high-P/E cohort's forward vs trailing spread, side by side with the 2000 dot-com giants.