Meta, Google, and Microsoft Triple Down on AI Spending (www.wired.com)

Microsoft, Meta and Google used their quarterly results to signal a massive ramp-up in AI infrastructure spending: Meta raised its full-year capex to $70–72 billion (from a prior $66–72B) and expects spending to be “notably larger” next year, after reporting $51.24B revenue (up 26%). Alphabet now forecasts $91–93B in 2025 capex (vs. an earlier $75B) after a record quarter of $102.3B revenue and 650M monthly Gemini users. Microsoft reported $77B revenue, a 74% year-over-year jump in quarterly capex to $34.9B, and said FY2026 capex growth will outpace FY2025. All three are plowing cash into data centers, GPUs and AI teams (Meta is hiring aggressively while reorganizing teams), and cite AI-driven gains in ads, cloud and VR. Parallel mega-deals underline the scale: Nvidia offered up to $100B to OpenAI contingent on 10 GW of centers, OpenAI plans 30 GW of capacity (~$1.4T economic footprint), and Microsoft has committed $13B to OpenAI (taking a $3.1B near-term hit). This is significant because it turns AI into a capital-intensive infrastructure race with material implications for chip supply, cloud economics and product roadmaps. Technical trends include “fungible” data-center designs that can be retooled year-to-year, continuous hardware refreshes to ride Moore’s Law, and software-led efficiency to amortize costs. Analysts warn of bubble risk and stranded capacity from multi-year builds, but vendors say phased investments and flexible data-center architecture reduce downside. Key things to watch: GPU availability/pricing, capex pacing, monetization of AI services, and whether demand justifies these multibillion-dollar bets.