Particle.news

Hyperscalers Scale Up AI Spending as Meta Launches Muse Spark 1.1 and Alphabet Sells Stock

The wave of record capital expenditures is lifting chip and memory makers and forcing new financing and product plans to prove that AI buildouts will turn into durable revenue.

Overview

  • Meta released Muse Spark 1.1 on July 9, a multimodal, agentic model positioned for paid access and developer use as the company doubles down on a $125 billion–$145 billion 2026 capex plan.
  • Alphabet shifted its funding strategy in early June by announcing roughly $84.75 billion in equity financing and raising 2026 capex guidance toward $180 billion–$190 billion to expand data centers and custom chips.
  • Analysts now estimate U.S. hyperscalers will spend roughly $750 billion to $805 billion on AI infrastructure this year, a jump that is reshaping data‑center designs and stressing supply of high‑bandwidth memory.
  • Pick‑and‑shovel suppliers such as Micron, NVIDIA and TSMC are reporting outsized revenue gains from the rush for chips and memory, with Micron citing triple‑digit sales growth tied to AI demand.
  • Major investors and research desks are projecting multi‑trillion AI spending over coming years but warn that power, financing and monetization hurdles could make returns uncertain and keep stock volatility high.