Business

Know the Business

Alphabet is a two-engine monopoly attached to a moonshot factory: the engine is a global search-and-video ad network that prints $129B of operating income on a $403B revenue base, and an infrastructure-for-AI business (Google Cloud) that finally turned profitable at scale in 2024 and more than doubled op income in 2025. The bottleneck for the next five years is not demand — it is how cheaply Alphabet can buy, power, and amortize the roughly $175–$185B of 2026 data-center capex it has committed to, and whether AI-native search monetizes at advertising-grade economics. The market still prices this as a mature ad franchise; the under-appreciated question is whether TPU-powered Cloud + Waymo is worth a multi-hundred-billion re-rating, or whether AI capex ends up as the first permanent hit to 30%+ operating margins in fifteen years.

1. How This Business Actually Works

One auction pays for everything else. ~73% of revenue is ads, and the highest-margin slice inside that is Google Search & other ($224.5B in 2025, +13% YoY). Search is a real-time auction where advertisers bid for keyword intent; Alphabet books revenue when a user clicks. Each incremental query costs Alphabet almost nothing in variable cost (serving compute + TAC paid to partners like Apple/Samsung for default placement), so incremental dollars fall to operating income at 60%+ conversion rates. Everything else — YouTube, Android, Chrome, Maps, Gmail, Play, Cloud — exists to feed this auction more intent, more often, from more devices.

The economic engine has four stacked layers, and a weakness at any layer leaks margin:

The bargaining power is asymmetric on both sides. Advertisers have no real substitute for high-intent query traffic at Google's scale — Meta sells social attention, Amazon sells shopping intent on-platform, but neither captures the moment a user asks a direct question. On the supply side, content creators (publishers, YouTubers) have to be on Google because Google owns the discovery layer. TAC to Apple (~$20B/yr) is the one payment where Alphabet is the weaker counterparty — Apple controls the iPhone default. That contract is also the target of the DOJ remedies phase.

What drives incremental profit in 2026 is no longer query growth (it's already trillions) — it's (a) Cloud margin scaling from 24% toward 35%+, (b) YouTube subscription mix (YouTube crossed $60B ads+subs in 2025), and (c) whether AI Overviews / AI Mode in Search monetize per-query at parity with ten blue links. So far, cost-per-click is up 7% in 2025 on 6% paid-click growth, which says monetization is holding.

2. The Playing Field

Among the mega-cap platforms, Alphabet sits in a specific slot: lower operating margin than Microsoft, lower FCF margin than Apple, but higher revenue growth than Apple and comparable growth to Microsoft — while spending the most on R&D of any company in the world in absolute dollars ($61B in 2025).

What the peer set reveals. Microsoft runs the best hyperscaler economics in the world — 46% operating margin on a $282B base — because Azure rides on pre-existing enterprise licensing relationships. Meta is the most profitable on gross margin (82%) because it owns its distribution (no Apple tax) and has zero TAC, but is spending 35% of revenue on capex to catch up in AI. Amazon is the low-margin volume machine — retail drags the consolidated number, but AWS is still the pace-setter in cloud. Apple is in a different business (hardware + services) and shows the outlier 152% ROE because of aggressive buybacks.

Alphabet's problem isn't absolute margin — it's that its moat cash cow (Search ads) pays for three other things the peer group doesn't have to fund simultaneously: (1) the largest R&D budget in corporate history, (2) capex now matching Meta's ratio, and (3) Other Bets losses of $7.5B. Strip those out and core Google Services runs at ~40% op margin (Services op income $139B / Services revenue $343B). The "best peer" comp for each slice: Microsoft for Cloud, Meta for core advertising, Apple for hardware discipline. Alphabet is competitive on each but best-in-class on none — except the one that matters most: query volume.

3. Is This Business Cyclical?

Less cyclical than people think, but no longer zero-cyclical. Digital advertising is GDP-plus at the cycle level (advertisers pull budgets in recessions but shift share toward performance/digital), and Search ads are especially defensive because advertisers monetize the bottom of the funnel — closer to a purchase decision. The 2008, 2015, and 2020 downturns all showed Alphabet losing ~2–5 points of operating margin and then rebuilding, not collapsing.

Three cyclical exposures worth naming:

1. Brand advertising (YouTube, Google Network) is cyclical. Performance ads are sticky; brand spend evaporates first. YouTube ads held up in 2022–23 but Google Network revenue has actually shrunk ($30.4B → $29.8B from 2024 to 2025).
2. Other Bets losses scale with valuations. The 2025 Waymo $2.1B comp charge shows that moonshot operating losses are partially a function of stock marks — when late-stage tech valuations inflate, SBC-linked moonshot costs inflate too.
3. Capex commitments are now a permanent fixed cost. $91B in 2025 going to $175–185B in 2026, plus $52.7B of future data-center leases not yet commenced. A revenue slowdown no longer deleverages — depreciation keeps growing for 4–6 years after spend.

The new cyclical risk is self-inflicted: capacity ahead of demand. If AI Cloud utilization disappoints, the 2026–2028 period could see the first multi-year operating-margin compression at Alphabet since 2013–2014.

4. The Metrics That Actually Matter

Forget the P/E. Five numbers actually tell you whether this business is creating value.

5. What I'd Tell a Young Analyst

Don't debate whether Search has a moat. Debate whether the AI capex bill is a new, permanent cost of goods sold or a one-time build-out. That is the entire bull/bear thesis in one sentence.

• What to watch. Google Cloud operating margin (anchor: 24% in 2025, needs to cross 30% to justify capex), Search cost-per-click stability under AI Overviews, depreciation as a % of revenue (6.6% in 2025, heading to ~10% by 2027), and whether Waymo ride-revenue crosses $5B run-rate (currently buried in Other Bets' $1.5B).
• What the market is missing. Services op margin ex-Alphabet activities is already 40%+ — Cloud scaling from 24% to 30% would add ~$5B to op income. The AI narrative is so loud that the terminal-value-destruction case (users just ask Gemini instead of Googling, with lower monetization) is probably over-discounted if you look at 2025 CPC + click trends.
• What would change the thesis. A DOJ remedy that forces Chrome or Android divestiture, a structural drop in Search CPC lasting >3 quarters, or Cloud growth decelerating below 25% YoY while capex stays elevated. Any one of those breaks the compounding story.
• The dumb mistake. Assuming the $45B annual buyback + $10B dividend means this is a returns-of-capital story. It isn't. Alphabet raised $37B of debt in 2025 and is about to acquire Wiz ($32B) and Intersect ($4.8B). This is a reinvestment story pretending to be a cash-return story — read the capex line, not the buyback line.