Both companies' reported earnings mislead — Google's are inflated by a reversible $99B accounting gain; Amazon's are hidden by enormous depreciation. This strips the noise with owner earnings to ask the only question that matters: how much cash could an owner actually take out?
Neither is cheap. Both trade at a ~2–3.5% "owner-earnings yield" — the whole price is a bet that their massive AI spending earns a return. Google is the higher-confidence business with the messier headline; Amazon is optically cheaper but only under some assumptions, and its earnings quality is not as clean as it first looks.
Google just reported a 294% jump in profit. Almost none of it was real.
Alphabet's Q2-2026 earnings per share came in at $9.11, up 294%. Strip out one item and the real number is about $2.61. The difference is a $99 billion gain on investments — and 99% of it is unrealized (an accounting mark-up, no cash changed hands):
| Mark-up on a private company (a funding round re-priced Google's stake) | $77.4 |
| Mark-up on public/other stakes (includes a SpaceX position) | $21.4 |
| Actually sold for cash | $0.3 |
| Total | $99.0 |
It's 71% of pre-tax income, and it's reversible — a market drop or a down-round unwinds it (exactly what happened to Amazon's Rivian stake in 2022, which pushed Amazon to a net loss). So the headline P/E of ~16 is a mirage. Clean it up and Google trades near ~31–32× earnings.
The −6.5% drop made it less expensive. It did not make it cheap.
| Metric | Reported | Clean / real |
|---|---|---|
| Trailing P/E | 16× (mirage) | ~31–32× |
| Forward P/E (2026) | — | ~22× |
| Its own 5-year P/E range | 19–29× | now at / above the top |
On genuinely clean earnings, Google sits at or above the top of its own 5-year valuation range. It is a fully-to-richly valued, high-quality business — not a bargain hiding behind a low headline P/E.
Four terms do all the work in this study. Here they are in plain English.
Amazon's higher P/E doesn't mean the market values it more. It's a depreciation artifact.
Amazon carries about 3× the depreciation of Google ($66B vs $21B) on a much lower-margin base. Depreciation crushes reported profit (the bottom of the P/E) but not cash. So Amazon's P/E looks high — but on EV/EBITDA, which adds depreciation back, the premium reverses:
| GOOGL | AMZN | |
|---|---|---|
| P/E (headline) | 16× (mirage) | 29× |
| EV/EBITDA | ~25× | ~18× |
| Operating margin | ~32% | ~11% |
| Depreciation (2025) | $21B | $66B |
| Stock last 12 months | +66% | +3% |
| Own 5-year P/E range | at the HIGH | near the LOW |
Caveat: EV/EBITDA flatters Amazon — that depreciation is real money spent replacing servers and warehouses. It's a directional equalizer, not proof Amazon is cheaper. The owner-earnings test (§5) settles it.
What can an owner actually pull out? And how much do the AI-spending assumptions matter?
| Convention | GOOGL | AMZN |
|---|---|---|
| Standard (SBC already counted) | ~$85–100 | ~$78 |
| Strict (extra charge for SBC dilution) | ~$87 | ~$58 |
| Owner-earnings yield — standard | ~2.4–3.0% | ~2.9–3.0% |
| Owner-earnings yield — strict (SBC) | ~2.3% | ~2.2% |
On the standard convention, Amazon yields a bit more — cheaper. But charge properly for stock-based comp (the strict view) and the two converge to ~2.2–2.3% — Amazon's edge disappears. So "Amazon is cheaper" is convention-dependent, not robust. Either way, both are low-single-digit yields — neither is cheap.
Free cash flow subtracts all capital spending as if every dollar were just replacement. That's false while these businesses are building AI capacity. Google's owner earnings (~$90B) are ~1.2× its FCF ($73B). Amazon's owner earnings (~$78B) are ~7× its reported FCF ($11.2B) — its cash is being consumed by growth spending. Valuing Amazon off its FCF is a category error.
Accurate valuation means knowing which numbers to trust. Here's the honest map.
Google's profit is quietly propped up by a 2023 accounting change that lengthened server lives (+$3.0B net income, still in force) — and a depreciation "catch-up" is coming that will pressure margins in 2027–29. Amazon is mixed, not pristine: it shortened some server lives (conservative) but also lengthened heavy-equipment lives (a flatter). Net, Amazon's accounting is modestly cleaner than Google's — not the one-sided win it first appeared.
This study argues against "historically a good moment to buy." On clean numbers Google is fully-to-richly valued, so a long-dated call option is leverage on a growth bet with no valuation cushion — not a discount. If you want a hyperscaler on owner economics: Google for confidence, Amazon for cleaner accounting and more torque — but neither is a bargain, and Amazon's apparent cheapness depends on how you treat stock comp.
noindex).