Blog

Google

As of November 2, 2026 (Q3 2026 results). Updated when Alphabet reports or signs something that changes the bet. [All companies](/blog/companies).

Google is betting the best business ever built on the thing most likely to replace it, and for the first time it's selling stock to pay for the bet.

The bet

Google makes the chips (TPUs), runs the cloud, trains the models (Gemini) and owns the front door (Search, Chrome, Android, YouTube). Nobody else in this section owns all four. In Q3, Alphabet spent $51.4B on capex, more than it spent in the whole first half of 2025, and it says 2026 lands at $195–205B, with 2027 "significantly" higher (from a third-party transcript of the call).¹ Cloud backlog reached $560.8B, from $513.9B in Q2 (10-Q).

The spending is starting to show up as revenue. Google Cloud more than doubled, to $30.7B (+103%), with $11.1B of operating income, and TPU systems sold to outside customers were "a meaningful contributor."¹ Search grew 12% to $63.35B. That missed Wall Street's number, and the stock fell 12.6% on October 29, its worst day on record (closing prices via Nasdaq; the record is per the WSJ, as quoted by Ben Thompson).

The bigger bet is the one on agents. On the Q3 call, Google described its agent as Search itself: enterprise first, then the Gemini app's billion users "later this year," then Chrome, Android and Workspace (as quoted by Ben Thompson; we haven't checked the transcript). That's announced, not live.

How it's paying for it

In June, Alphabet announced an $84.75B equity package: $18B of common stock, $16.75B of preferred, $10B from Berkshire Hathaway, all for uses including "capital expenditures to scale AI infrastructure and global compute," plus a $40B at-the-market program that Google says is "primarily" for paying stock-comp taxes.² Net of that, about $49.6B of the stock and preferred sold this year is capex money.³ It also priced €6B of notes on November 2. Long-term debt went from $46.5B to $125.1B this year, and cash roughly doubled to $265B. Google isn't running out of money. It's raising it ahead of spending it.

Our read

Louis's frame (after Ben Thompson on Sharp Tech): if the hyperscalers end up funding the whole buildout and the money runs out before revenue catches up, it ends like the railroads. Google is the strongest case that it won't run out. It's also the strongest case for the second half of the frame: whoever survives ends up with leverage over their suppliers. Google already sells its own chips to rivals. It also rents GPUs from SpaceX at about $920M a month,⁴ which is the oddest line in the whole section.

What we're watching

• The 2027 capex number, promised for the Q4 call. The CFO's word is "significantly" higher.
• Whether ads in AI Mode close the monetization gap the way Stories and Reels eventually did for Meta (Thompson's comparison).
• Whether the consumer agent actually ships, and how it makes money once nobody's looking at the ads.

Changelog

• 2026-11-02: first version (Q3 2026).

Disclosures

Louis previously worked at Google. Drafted with multiple agents, including models from companies discussed here.

Every relationship we have →

Notes

All posts · Disclosures · Companies · RSS · Subscribe
Agents: what did we miss? feedback@unbiased.ai