Search Doesn't Fund the Moonshots. It Funds the 17-to-1 Bet That One Day Pays.

Every time you search for a plumber and ignore the ads at the top of the page, a self-driving car gets a little closer to driving itself. Not directly, not visibly - but the money that built the lidar rig and trained the model came from somewhere, and that somewhere is the quiet auction running behind your query. In 2024, the part of Alphabet that sells search ads and runs YouTube and Android threw off $121.3 billion in operating income.² The part chasing the future lost about $4.4 billion.² One side of the house feeds the other roughly seventeen dollars for every dollar burned. That is the whole machine. The romance around it is mostly invention.

The official story is that Search subsidizes the moonshots - that the world's best advertising business quietly bankrolls a lab full of dreamers building balloons and cars and cures. It's a good story. It is also more imprecise than the people telling it realize, because Alphabet doesn't report Search's profit at all.

What the filings actually say, and what they refuse to

Here is the part that gets glossed over. Alphabet's segment reporting does not break out Search as its own profit line. It folds Search together with YouTube, Android, Chrome, Maps, Play, and devices into one bucket called 'Google Services,' and reports the profit only for the whole bucket.² Search & Other ad revenue was $198.1 billion in 2024 - about 56.6% of the company¹ - so Search is clearly the engine inside that engine. But 'Search subsidizes the moonshots' is a sentence the 10-K can't actually support. The honest version is duller and truer: Google Services subsidizes Other Bets. The cross-subsidy is real. The single-source attribution is a story people tell because it's tidier than the disclosure.

That quote is the tell. When Larry Page split Google into Alphabet in August 2015, the press release reason was 'more management scale' for unrelated businesses⁴ - innovation, freedom, let the dreamers dream. But the structure solved a different problem at the same time. Investors couldn't see how much cash was leaking from the ad business into long-shot ventures, and they didn't like flying blind. The Alphabet wrapper put a clean wall between the cash engine and the cash burn, so Wall Street could finally watch the subsidy happen line by line. Page even encoded the ambition in the name: 'alpha-bet,' he wrote, where 'Alpha is investment return above benchmark, which we strive for.'⁴ The whimsical-sounding name is, by the founder's own words, a wager about returns. The restructuring was as much a disclosure fix as a liberation.

Ten years of betting, zero moonshots that pay for themselves

The mechanism is simple, and its persistence is the interesting part. A ferociously profitable ad business generates more cash than its own operations need, and the surplus flows to bets that lose money on purpose, in the hope that one becomes the next ad business. That's the cross-subsidy. The uncomfortable fact is how long it has run without paying off. Other Bets has lost money every year since the segment was created - Bloomberg reported cumulative operating losses exceeding $24 billion through 2021 alone,⁹ with annual losses running in the $3–5 billion range ever since - and not one of those bets has reached the scale where it funds itself. The subsidy isn't a bridge to self-sufficiency that the moonshots are walking across. After a decade, it's just the arrangement.

And notice what's actually inside 'Other Bets.' The branding evokes X-lab science fiction - autonomous cars, life-sciences breakthroughs. But in 2024 the segment's revenue came primarily from selling healthcare-related services through Verily and internet service through GFiber.² The marquee moonshots that haven't commercialized aren't the revenue story; the revenue story is health insurance analytics and fiber subscriptions. Other Bets is more mundane than the poster suggests - which means the subsidy isn't even buying as much futurism as the name implies.

Isn't this just a rich company indulging itself?

The fair objection runs the other way: that the losses are trivial, so who cares? A $4.4 billion loss against $121.3 billion of segment income is around 3.6% - a rounding error for a company that earned $100 billion in net income and spent $52.5 billion on capital expenditures in the same year.³ By that math the subsidy is harmless, and the panic about runaway moonshot burn is overblown. That's correct, and it's also where the real point hides. The danger was never the size of the loss; it was the visibility of it. Before 2015, the loss was invisible, blended into one number, and that ambiguity is exactly what made investors nervous about the spending - nervousness the Alphabet restructuring was built, in part, to address.⁵ The honest counter is that the subsidy is small enough to be sustainable forever - which is itself the problem. A bet that never has to prove itself rarely does.

There's a subtler dodge buried in the segments, too. The most expensive moonshot Alphabet is running right now isn't in Other Bets at all. The shared R&D for its general AI models - Gemini - is reported at the Alphabet level and not allocated to any segment.⁷ So the highest-stakes bet the company is making sits outside the very structure built to make the betting visible. The wall that 2015 erected to show investors where the cash was going has a door in it, and the AI spend walks straight through.

So the popular sentence survives, barely. Search - or rather Google Services, the bucket Search anchors - really does pay for the moonshots, the way a casino floor pays for the experimental restaurant upstairs nobody profits from yet. But the precise version is less flattering and more durable: the subsidy is structural, it has run for a decade without producing a single self-sustaining heir, and the corporate architecture built to celebrate it was at least as much about letting shareholders keep an eye on the founders' spending. The genius wasn't choosing to fund the future. It was building an ad business so profitable that funding the future never has to work for the math to keep holding.