Alphabet Inc. GOOG

Key data

The business

Alphabet sells two things. Google Services, about four-fifths of revenue, is Search, YouTube, Android, Play and subscriptions: advertisers pay to sit next to a query or a video, and the price they pay per click is set in a live auction billions of times a day. Google Cloud, the other fifth, rents computing power and, increasingly, Google's own Gemini AI models to companies. The United States is now 51% of sales. The moat is the search index and ad auction feeding a machine no rival can match on scale, wrapped around owned data centers and custom chips that let Google run AI cheaper than anyone renting someone else's hardware. The thing a customer actually touches is mundane: a text box, a video feed, a bill for cloud compute. That mundanity is the durability.

The numbers

Revenue has climbed almost every quarter, the operating engine is accelerating, and the headline profit line has come unmoored from it.

Revenue grew 24% in Q2 and operating income grew 30% to $40.8B, a 34% margin, the honest measure of the business. Reported net income of $112.2B and EPS of $9.11 are not: a non-operating investment gain of roughly $98B before tax lifted pretax profit from $34B a year ago to $139B. The analyst "beats" of the last four quarters (9.11 against a 2.87 estimate) are those gains landing, not the operation outrunning expectations. Strip the windfall and the operation grew 30%, which is the number that matters.

Revenue compounded about 17% a year for five years to $403B (2020 to 2025), and EPS about 18% over four years (2021 to 2025), helped by years of heavy buybacks shrinking the share count. The 1H 2026 profit already tops all of last year purely because of the investment gain, so treat that row as revenue-true and profit-inflated.

The cloud arm is where the surprise sits.

Cloud revenue nearly doubled, and the backlog behind it is $519.5B of committed contracts, $513.9B of it Cloud, with just over half expected to convert to revenue inside 24 months. What this memo believes and the screen does not: the sub-17-times trailing multiple is not a hidden re-rate waiting to spring, it is that one-time gain flattering the "E." Normalize it and you own a low-twenties grower at about 19 times forward earnings, fairly priced for the quality. The cloud backlog is the one thing that could still move the story, and the print that settles it is Cloud's revenue and segment margin over the next two quarters.

Management

The insider tape is noise: no buys, $3.4M of small sales over twelve months, the largest $841K by legal chief John Kent Walker on June 29, plan status not disclosed. The real management signal is capital allocation, and it just flipped. Buybacks ran $45.7B in 2025 and then fell to zero in Q1 2026. At the same time long-term debt roughly doubled from $46.5B to $98.2B between December 2025 and June 2026, the company issued 6.25% mandatory convertible preferred stock, and it privately placed about 28.6M Class A and C shares in June. Alphabet, historically a cash-returning fortress, is now raising debt, preferred and equity to fund a capital-spending surge that hit $35.7B in Q1 2026 alone against $91.4B for all of last year. That is a different company than the one that bought back $60B a year.

How it fails or surprises you

Cloud runs away with it (right tail). Cloud grew 82% to $24.8B with a $513.9B backlog behind it, over half converting inside two years. If revenue holds this pace and segment margins scale the way scaled compute does, Cloud alone re-rates the whole company. The market underpays for it today because the capex feeding it crushes reported free cash flow. Watch Cloud revenue growth and segment operating margin.

The buildout doesn't pay. Capex is heading toward a $140B annual run-rate, free-cash-flow yield is down to 1.3%, buybacks stopped, and debt doubled. If AI and cloud demand slows before the data centers earn their keep, returns compress and the funding pivot looks premature. Watch capex-to-revenue against Cloud's margin trend.

Chatbots eat the search box. Services still grew 14%, so the feared cannibalization has not shown up in the numbers yet. That is the fact this read explains least: a durable shift of queries to AI answers that carry no ad auction would hollow out four-fifths of the profit slowly, invisibly at first. Watch Services revenue growth quarter to quarter; a deceleration toward high single digits would signal the turn.

Closing thoughts

The screen shows the cheapest trailing multiple in twelve years, but that is an artifact of a $98B investment gain, and stripping it leaves a 30%-operating-growth business at about 19 times forward, fairly valued rather than mispriced. The fatter tail is upside, and it is Cloud: a $514B backlog is a large, disclosed thing the market is discounting because the capex to serve it is masking cash flow today. The left tail is real but slow, a gradual bleed of Search queries to AI answers, and it would announce itself in the Services line long before it broke the model. What is actually at risk if the buildout disappoints is the multiple, not the franchise.

The bet is still that Google's Search and YouTube remain the dominant ways people find information and watch videos online, while Cloud grows into the second engine, and that AI chatbots that answer questions directly don't drain searches, which generate most of the advertising profit. It breaks if Services growth rolls toward zero while capex keeps climbing. The one pair of numbers that tells you first: Services revenue growth and Cloud's segment margin, read side by side each quarter.