Particle.news
Download on the App Store

AI Market Shifts From Chatbots to a Compute Arms Race

Massive spending on data centers, GPUs and integrated systems is testing whether new capacity can be turned into steady cloud revenue.

Overview

  • Alphabet has told investors it will spend $180 billion to $190 billion on capital expenditures in 2026 and said 2027 capex will rise significantly, a plan meant to add large-scale AI compute capacity.
  • Google Cloud revenue grew 63% year over year in Q1 and the segment’s contracted backlog roughly doubled to $462 billion, a sign demand is outpacing current supply.
  • Morgan Stanley reported on Sunday that Alphabet may be renting about 110,000 Nvidia GB300 GPUs from SpaceX — an estimate that would equal roughly 220 megawatts of power use and about $11.04 billion in annualized expense — and the figure is presented as a reported estimate, not company-confirmed.
  • Nvidia is expanding beyond chips with its Blackwell platform, plans to ship the Vera Rubin six‑chip system by year‑end, and projects about $1 trillion in orders for Blackwell and Vera Rubin across 2026–2027, while its CUDA software creates high switching costs for developers.
  • Wall Street has raised price targets on Alphabet as some analysts argue heavy AI capex could be monetized into higher cloud and enterprise AI sales, but investors remain split because rising compute intensity could also squeeze margins if monetization lags.