The PwC AI Forecast Doesn't Explain Google's Stock Drop

PwC released a new forecast on Tuesday predicting $31.6 trillion in global AI infrastructure spending through 2050. Google's stock was already down 11% over the prior three weeks.

The headline connects two things they don't actually touch. The PwC number describes what hyperscalers have already been doing — building data centers at a pace that dwarfs the last internet boom. The stock move reflects what investors are questioning: whether the company can earn back what it's spending. These are two different conversations, and confusing them is the most expensive mistake a Google investor can make right now.

Here's the actual sequence that matters.

Alphabet announced in February that it planned to spend $180–$190 billion on capital expenditures in 2026 — roughly double what it spent in 2025. Investors were already uneasy. Then in July, after second-quarter earnings, the company raised that guidance to $195–$205 billion and said it expects capex to grow "significantly" in 2027 as well. The stock fell more than 6% in one day and hasn't recovered.

What drove the spending is clear. Google spent $44.9 billion on property and equipment in the second quarter alone — up 100% from the year-ago period. It was the first quarter in the company's 22-year history that capex exceeded revenue, pushing free cash flow negative by $5.9 billion. The company needed so much capital to keep building that it raised $49.6 billion through a stock offering and another $20.3 billion in debt, just in the second quarter.

This isn't theoretical. The cash is leaving.

Now look at what's supposed to come back. Google Cloud revenue grew 82% year-over-year in the second quarter to $24.8 billion, and cloud operating income more than tripled to $8.81 billion. The remaining performance obligations — contracted work the company has signed but not yet delivered — hit $514 billion. That's roughly five years of cloud revenue at the current annual run rate.

The backlog is the single most important number for a Google investor to sit with. It represents revenue the company has already contracted for, much of it in multi-year deals with the biggest enterprise customers in the world. Management said over half of that backlog will be recognized as revenue over the next 24 months. If those contracts hold — and there's no reason to doubt they won't, given they're enterprise commitments with cancellation terms — the spending has a revenue floor.

But revenue is not return on investment.

Here's where the economics get harder. Management acknowledged that supply constraints are forcing the company to use third-party compute capacity as a "bridging strategy," which will create "modest margin pressure in the near term." In plain language: Google can't build its own infrastructure fast enough to meet customer demand, so it's renting compute from other providers at a markup. The CFO called these deals "highly ROI positive over the lifetime," but the near-term margin compression is exactly what happens when you're paying someone else's capital cost on top of your own.

The PwC forecast is useful for understanding one structural detail: the vast majority of this $31.6 trillion goes into equipment — chips, servers, networking — not buildings. By 2050, equipment is projected to represent 93% of data center capex, up from 70% today. AI chips need replacing every four to six years. This isn't a front-loaded investment like railways or the original internet rollout; it's a recurring burn, and the cost never goes away.

Google is locked into that same cycle. The $195–$205 billion in capex for 2026 is only this year's installment, and management expects the number to grow in 2027. The question isn't whether the AI infrastructure buildout is real — it is, and Google is one of the few companies positioned to profit from it. The question is whether the returns on each dollar of capex are visible and growing fast enough to justify the commitment.

Right now, the numbers offer a split verdict. On one side, the company has $514 billion in contracted cloud work, cloud margins are expanding from roughly 21% to 36% year-over-year, and AI models are processing 22 billion tokens per minute through Google's APIs — up from 16 billion recently. Nearly 90% of the Fortune 100 use Gemini Enterprise. The demand is demonstrably there.

On the other side, free cash flow turned negative for the first time, the company diluted shareholders to fund the buildout, and the path to positive cash flow depends on that $514 billion backlog converting at attractive margins over the next several years — while capex keeps accelerating. Google trades at a trailing P/E of roughly 17, which looks cheap until you realize that multiple includes a $99 billion one-time gain on equity investments (Anthropic, SpaceX stakes). Strip that out and the non-GAAP valuation runs closer to 33 times forward earnings.

The stock is down 11% from its five-week high because investors are asking a perfectly valid question: when does this spending cycle start paying back enough to generate the free cash flow that justified the valuation in the first place?

That's the right question. But the answer doesn't come from a $31.6 trillion forecast about global infrastructure. It comes from Google's own backlog conversion rate, its cloud margin trajectory, and whether the company can reduce its reliance on third-party compute as its own capacity comes online.

Watch the next two earnings reports for three things. Whether cloud revenue growth sustains above 60% as the company begins "lapping" prior-year acceleration. Whether the third-party bridging strategy shrinks as internal capacity scales. And whether free cash flow returns to positive territory on a quarterly basis — because that's the moment the market will stop pricing Google as a company burning cash on an AI bet and start pricing it as one collecting on it.

Until then, the gap between the spending and the proven return is the gap investors are sitting in.