Big Tech AI capex number
Coverage cites estimates that Big Tech will spend $720 billion on AI in 2026 — a headline number driving continued hardware and services demand narratives. That scale underpins aggressive vendor roadmaps and hyperscaler procurement cycles. (fool.com)
Amazon signaled about $200 billion in 2026 capital spending, saying most of that outlay will fund AWS data centers and capacity projects. (cnbc.com) Alphabet guided capital expenditures of $175 billion–$185 billion for 2026, telling investors the increase is tied to compute capacity for Google Cloud and DeepMind. (cnbc.com) Meta’s Q4 2025 filing set full‑year 2026 capex at $115 billion–$135 billion, with CFO Susan Li attributing the jump to investments for “Meta Superintelligence Labs” and core infrastructure. (investor.atmeta.com) Microsoft reported $34.9 billion in capex in Q1 FY2026 and $37.5 billion in Q2, putting first‑half 2026 capex at about $72.4 billion and establishing a near‑$100 billion annualized run‑rate for AI short‑lived assets like GPUs and CPUs. (microsoft.com) Oracle raised fiscal‑2026 capex guidance to roughly $50 billion and said it expects to raise $45–50 billion of gross proceeds to fund a global OCI expansion tied to customer commitments and GPU installs. (gurufocus.com) Markets reacted: Amazon shares plunged in after‑hours trading after the $200B projection, while Alphabet and Meta also saw stock pullbacks when they disclosed outsized 2026 capex ranges. (finance.yahoo.com) Hyperscalers’ guidance shows a disproportionate tilt toward “short‑lived” compute assets — Microsoft said roughly half of recent capex went to GPUs/CPUs — and Oracle reported multi‑billion quarterly GPU and data‑center investments, signalling accelerated demand for accelerators, networking, and systems. (microsoft.com) Analysts and banks flag power, foundry and component constraints as real supply‑side limits: Goldman Sachs projects sharp data‑center power demand growth through 2026, and several analyses highlight TSMC’s central role in meeting chip fabrication needs. (goldmansachs.com)