Google projects full-year capex will be $195B to $205B in 2026, after saying in April that it will spend as much as $190B; its Q2 capex hit $44.92B, up 100% YoY
Julia Love /Bloomberg:NEW
Context & Ripple Effects
Alphabet entered 2026 forecasting $175B-$185B of capex after reporting $91.4B in 2025, then lifted the range to $180B-$190B in April. The latest range is therefore the company’s third upward reset this year, following its initial 2026 spending step-up and April guidance increase.
The Q2 outlay shows that the higher annual plan is already being deployed rather than merely reserved. That makes execution of the buildout—and the return on a much larger capital base—more consequential for Google than when its 2025 capex plan was about $75B.
First-order effects
- Google has raised both ends of its 2026 capex range by $15B from April, expanding the funding envelope for its current infrastructure program.
- A $44.92B Q2 spend, up 100% year over year, gives the revised forecast operational weight and raises the near-term importance of deploying that capital efficiently.
Second-order effects
- The larger Google commitment increases pressure on other major platforms to preserve access to compute capacity; Meta had already outlined a far higher 2026 capex plan of its own.
- Infrastructure suppliers and financing partners gain a clearer signal that demand from hyperscalers is persisting through 2026, though the rapid guidance changes also make order and capacity planning less predictable.
Third-order effects
- If repeated upward revisions become common, AI infrastructure budgets may be managed as rolling capacity commitments rather than fixed annual allocations, concentrating more strategic leverage with companies able to fund and operate large buildouts.
- The same pattern raises compute execution risk: sustained spending will increasingly be judged on whether additional capacity translates into durable product or cloud demand, rather than on the scale of capex alone.
The trend: This is another data point in the shift from episodic data-center investment to an AI-driven, continually repriced compute-capacity race.