AI capex estimates just got raised again. $Amazon.com(AMZN)$ , $Meta Platforms, Inc.(META)$ , $Alphabet(GOOGL)$ , $Microsoft(MSFT)$ combined AI infrastructure spending for 2026 is now guided around $720B–$745B. Last quarter it was ~$695B–$725B.
Rough breakdown: $Amazon.com(AMZN)$ $220B, $Alphabet(GOOGL)$ ~$195B–$205B, $Microsoft(MSFT)$ $175B, $Meta Platforms, Inc.(META)$ ~$130B–$145B.
There has been deleveraging and margin pressure across parts of the market. Short-term volatility could stick around. But the medium-to-long-term thesis still looks intact to me. Every major hyperscaler keeps pointing to compute shortages, strong cloud demand, pricing power, and higher semiconductor and networking investment.
The AI buildout is shifting from hype to actual infrastructure. That is why I keep watching upstream beneficiaries — semiconductors, networking, power, NeoClouds. Companies like $IREN Ltd(IREN)$ , where prepayments are running at 45%, show just how strong demand for available AI capacity remains.
The market can debate the cycle. The infrastructure build is still accelerating.
Comments