◆ What the Market Isn't Pricing In

The most useful fact about Google's valuation is where it was fourteen months ago. In March 2025 the market, gripped by the conviction that chatbots would make search obsolete, marked the stock down to about seventeen times earnings¹ — a disruption discount, the multiple of a company thought to be facing its executioner. By the middle of 2026 it had recovered to roughly twenty-seven times, a touch below its ten-year average². In between, the thing the market feared simply failed to happen: in the first quarter of 2026 Search revenue grew nineteen percent, queries hit an all-time high³, and management credited the very AI features that were supposed to kill search with drawing people back to it. The panic was wrong, and the price has largely un-panicked. What follows is a note on what, even after that recovery, the market may still be undercounting — and it is not a price forecast.

Here is the heart of it. For twenty years Google has been, to the market, an advertising company — a magnificent one, but a single-engine business, and it is still valued essentially as such: a market multiple laid over a river of ad profit, with everything else treated as a rounding error or a cost. That framing was fair for most of its history. It grows less fair by the quarter, because three other engines are now running inside the company, and a blended ad multiple captures none of them well.

The first is Google Cloud, and it has crossed the line that changes how a business should be valued. Cloud revenue grew sixty-three percent to twenty billion dollars in the quarter⁴, but the more important number sits underneath: its operating profit more than tripled, and its margin climbed toward the low thirties from the high teens a year earlier⁵. For most of its life Cloud was a money-losing land-grab the market was right to discount; it is now a large, fast-growing, genuinely profitable business with a backlog of contracted work that nearly doubled in three months to more than four hundred and sixty billion dollars⁶. The multi-year arc is what makes this concrete: as recently as 2022 Google Cloud lost about $1.9 billion; it turned a $1.7 billion profit in 2023, roughly $6 billion in 2024⁷, and by early 2026 was earning more in a single quarter than in all of 2024. A business does not travel from red ink to that by accident — it crosses from land-grab to franchise. A profit engine of that size and slope, still buried inside an ad multiple, is the plainest thing the price undercounts.

The second is subtler and may matter more. Almost alone among the giants, Google owns the entire AI stack — the chips, the models, the data, the distribution, and the products — end to end. It designs its own TPUs, and in 2026 it began selling them into other companies' data centers⁸, which makes it not merely a user of AI silicon but, quietly, a merchant rival to Nvidia in the one market everyone wants exposure to. It has a frontier model in Gemini, the distribution to put it in front of billions by default, and the rare luxury of running its own AI on its own chips at a cost it controls — the cost of answering fell about a third in a year⁹. The market pays readily for Google's ads; it has not obviously paid for the optionality of being the only fully integrated AI supplier on earth.

The third engine barely registers in the price at all: Waymo now gives more than half a million autonomous rides a week, roughly doubling in under a year across eleven cities¹⁰ — an option on an enormous market carried at nearly nothing inside an ad multiple. I do not want to make this sound one-sided, because it is not, and two cautions belong in plain view. The reported earnings flatter the picture: the quarter's eighty-one percent jump in net income¹¹ leaned heavily on a one-time, unrealized gain on investments worth tens of billions, so the operating business grew far less than the headline, and a price-to-earnings ratio built on that number looks cheaper than the operations warrant. And the antitrust threat is real and structural — courts have already found the default-search deals unlawful¹² and the ad-tech business monopolistic, and although the sale of the ad exchange the government sought was refused in September 2026¹³, the conduct rules that remain could trim the very engine that funds all the others. A conglomerate discount is sometimes deserved.

So why argue anything is underpriced when the multiple has already climbed back to its long-run average? Because that average was set on the old identity — Google the ad company — and the company has quietly become something else while the label stayed the same. The market corrected its 2025 mistake of pricing Google as AI's victim; it has not yet grappled with the possibility that Google is one of AI's principal beneficiaries, with a profitable cloud, a merchant-silicon business, a frontier model, and a robotaxi fleet all compounding inside a valuation that still mostly counts the ads. The antitrust overhang is genuine and the headline earnings are flattered — those are the honest reasons for a discount. But the gap worth watching is between a market still appraising a single-engine advertiser and a business that, quarter by quarter, is turning into a full-stack AI conglomerate whose parts the blended multiple was never built to see. That, not the settled question of whether AI kills search, is where a patient owner should be looking.