Google Cloud Doubled — and Alphabet Still Fell: The TPU Quarter

中文版 Chinese version

Figures are Alphabet’s reported Q3 2026 results (quarter ended September 30, reported October 28, 2026) and consensus comparisons as reported by Reuters, CNBC, and LSEG-based coverage; where after-hours moves differ across outlets (roughly 8–10%), the range is given rather than a single print.

The quarter in one table

Operating income still rose 30% to $40.6 billion. This was not a bad quarter in absolute terms. It was a bad quarter relative to the price of the story Alphabet is telling — and the market said so immediately, with shares down roughly 8–10% in after-hours trading, a move Reuters-linked coverage put near $400 billion of market value. Microsoft, reporting the same day with positive free cash flow, rose more than 5%. The contrast was the trade.

The real news: Google became a chip vendor

Buried inside the cloud beat is the structural shift. Google Cloud chief Thomas Kurian made his first appearance on an Alphabet earnings call to say what the numbers already implied: sales of Google’s Tensor Processing Units (TPUs) — the custom AI chips it once built only for itself — made a “meaningful” contribution to cloud growth. Alphabet now sells TPU systems: hardware, networking, storage, and software together, to nearly 100 customers, with recent partnerships to deliver systems with Crux AI and Oracle, Apple scaling foundation-model pretraining on the latest TPUs, and joint work with Meta on PyTorch support.

That makes Alphabet a direct Nvidia challenger in the merchant silicon business, not just a Nvidia customer. It is also why the CFO’s inventory comment matters. Finance chief Anat Ashkenazi said TPU systems not yet sold now show up in inventory — hardware built and paid for before customers pay. One analysis of the quarter put the swing starkly: Alphabet carried $2.4 billion of inventory at the end of 2025 and $13.3 billion nine months later, with the buildup attributed to TPU systems assembled for signed customers awaiting delivery. Treat that inventory figure as reported analysis, not an Alphabet headline disclosure — but the mechanism Ashkenazi described is not in dispute, and it changes Alphabet’s cash profile in a way search never did. Search ads are sold and collected almost as they run. Chips are built, stocked, and delivered on someone else’s schedule.

Why a doubling cloud could not save the stock

The arithmetic is unforgiving. Cloud is now a $30.7 billion quarterly business; advertising is an $82 billion one. Even doubling cannot cover an ad miss when the ad business is what funds a $51.4 billion quarterly capex run-rate. And the ad miss has a specific, uncomfortable cause: chief business officer Philipp Schindler attributed a larger-than-expected deceleration in Search growth to greater use of AI Mode, the chatbot interface whose ads Google is still experimenting with. Query growth, Alphabet said, is the highest in five years — but queries that do not convert into revenue at the old rate are exactly the AI-in-search question the market has been asking for two years, and this quarter did not settle it.

So the bear case and the bull case are the same numbers, read in opposite directions. Bull: a $30.7 billion cloud business growing 103%, with a merchant chip line that diversifies Alphabet away from both Nvidia supply and ad cyclicality. Bear: that business is being built with $51.4 billion quarters, negative free cash flow, rising debt and equity issuance, an ad engine decelerating partly because of Alphabet’s own AI product, and a 2027 capex increase still unquantified — one investor put the fear plainly, that a $300 billion capex number “could potentially scare the market.”

The judgment

Do not read this quarter as “AI spending punished.” Microsoft’s rise the same day kills that reading. Read it as a funding question: Alphabet’s AI buildout is now large enough that its financing — inventory-heavy chip sales, debt, dilution, and an ad business that must keep growing while its own AI Mode changes how search monetizes — is itself a business risk, not just a cost line. The TPU pivot is genuinely good news for competition: a second merchant AI-chip supplier at scale is the one thing the whole industry has been waiting for. But a vendor’s cash cycle is harsher than a search engine’s, and October 28 was the first quarter the market priced Alphabet like it might have to live with that difference.

Sources: Alphabet Q3 2026 results as reported October 28, 2026 (Reuters via syndication; CNBC via earnings coverage; LSEG consensus figures); Google Cloud and TPU commentary from the earnings call (Kurian, Ashkenazi, Schindler); inventory analysis of the TPU buildup as reported in post-earnings analysis.