The numbers are getting absurd, and not in the fun way. Amazon, Microsoft, and Alphabet have collectively guided their 2026 capital expenditure to roughly $725 billion, a figure that represents approximately a 77% jump from the $410 billion previously projected for 2025.
The spending spree is driven by one thing: insatiable demand for AI infrastructure. Data centers, servers, networking gear, and most importantly, chips.
The spending breakdown
Amazon is leading the pack with around $200 billion in projected 2026 capex. Microsoft trails closely at approximately $190 billion. Alphabet, not content to be left behind, raised its 2026 forecast to a range of $175 billion to $205 billion after bumping the number up again in July 2026.
Meta rounds out the hyperscaler quartet with a projected $115 billion to $135 billion in capital spending.
Amazon’s company leadership has indicated that a considerable portion of AWS capital expenditure for 2026 already has pre-set customer commitments lined up for monetization in 2027 and 2028.
Alphabet’s aggressive posture comes on the heels of an 82% annual growth rate in Google Cloud revenues. Alphabet posted a historical quarterly cash burn of $5.9 billion in Q2 2026 while simultaneously raising its capex forecast by $15 billion.
The chip supply chain under pressure
Amazon has been developing its custom Trainium chips, with cumulative commitments exceeding $225 billion. Alphabet continues investing in its proprietary TPU (Tensor Processing Unit) architecture.
Analysts predict hyperscaler capital expenditures could reach into the trillions by 2030 or 2031. The significant capex revisions that started in 2024 and 2025 appear to have been the opening act, not the main event.
What this means for crypto-adjacent investors
When hyperscalers are vacuuming up every available chip, that creates supply pressure that affects GPU availability and pricing for everyone else, including crypto miners. Bitcoin miners who rely on GPU-adjacent supply chains for next-generation hardware should be watching these capex numbers closely.
Building $725 billion worth of data centers requires staggering amounts of electricity. AI data centers and Bitcoin mining operations increasingly compete for the same power resources, particularly in regions with cheap, abundant energy. As hyperscalers lock up power purchase agreements and build out grid connections, the competitive landscape for crypto mining energy access gets tighter.
As AI infrastructure scales, tokenized data assets and decentralized computing networks like Render and Akash position themselves as alternative providers of GPU compute.