Google projects its capex at $195B to $205B in 2026, after saying in April that it will spend up to $190B; Q2 capex grew 100% YoY to $44.92B; GOOGL falls 7%+
Julia Love /Bloomberg:
Context & Ripple Effects
Google began 2026 with a $175B-$185B capital-spending outlook, then lifted it to $180B-$190B in April. The latest range is therefore a third escalation of an already aggressive infrastructure plan, following its initial 2026 capex forecast and the April guidance increase.
The Q2 spending figure gives the revised annual outlook operational weight rather than leaving it solely a forecast. It also extends a sharp shift from Alphabet's roughly $75B 2025 capex expectation, as reported in its prior-year investment plan.
First-order effects
- Google's finance and infrastructure teams must fund a higher 2026 spending envelope, with Q2 capex already at $44.92B and the full-year range now $195B-$205B.
- The revised guidance makes Google a materially larger near-term buyer of data-center capacity and the equipment required to build it, while increasing the importance of converting that spend into usable AI and cloud capacity.
Second-order effects
- Suppliers and data-center partners face stronger demand visibility from Google, while rival cloud platforms are pressured to show whether their own infrastructure plans can keep pace.
- For investors, the higher capex burden puts greater focus on the returns from Google Cloud and AI-related services, rather than on spending growth alone.
Third-order effects
- Repeated upward revisions suggest AI infrastructure budgets are becoming rolling commitments: capacity plans can be reset during the year as deployment needs change.
- If this pattern persists across large platforms, competitive advantage may increasingly depend on financing and executing physical compute buildouts—not only on model and product development—while raising execution risk if demand or deployment timelines disappoint.
The trend: Google's revisions are one data point in an AI infrastructure arms race in which hyperscalers repeatedly expand compute commitments as capacity needs become clearer.