Particle.news

Tech Giants Push AI Into Team Tools, Chips and Token Factories

The industry is moving from model research to commercial systems that must confront rising token demand, energy needs and deployment limits.

Overview

  • Anthropic has started rolling out Claude collaboration products with Claude Data live and Whiteboards and Sheets in testing to place its AI directly inside teams’ daily workflows.
  • Lenovo unveiled a 'Wanquan' hybrid token factory plan for a Zhangjiakou mother site that targets large-scale compute, 100MW of smart compute capacity and annual token production aimed at serving continuous agent-style workloads.
  • IDC projects rapid physical expansion through 2030, forecasting global data-center cabinets and related facility spending to more than double and warning that power, cooling and supply chains will constrain where capacity can actually grow.
  • ByteDance is reported by Bloomberg to be preparing a Muse-like assistant and plans large multiyear AI investment while Baidu announced a Kunlun chip roadmap with the M300 next year and a multi-chip N-architecture launch in 2028.
  • AWS CEO Matt Garman said orbiting data centers are unlikely to be commercially viable for 10–20 years, underscoring that far-term ideas do not solve the near-term problem of soaring per-use compute demand and strained infrastructure.