Particle.news

Tencent Boosts AI Compute with Massive Q2 Spending to Build Cloud Capacity

The company says rising domestic GPU supply by late 2026 or early 2027 will let it shift more capacity to cloud customers and end the current high‑investment phase.

Overview

  • Tencent reported a sharp rise in capital expenditure for the second quarter, spending RMB 527.8 billion mainly on AI infrastructure to buy GPUs and related systems for model training and inference.
  • The company posted revenue of RMB 2,047.9 billion and a Non‑IFRS net profit of RMB 684 billion for Q2 while warning that heavy one‑off AI spending will reduce cash available for stock buybacks.
  • Executives said Tencent will keep accelerating compute purchases through the end of this year into early 2027, managing allocation between internal AI projects and external cloud leasing to balance product needs and customer demand.
  • Tencent Cloud plans to add near‑package optics (NPO) supercomputing nodes in Q4 2026 and industry estimates suggest single‑quarter domestic procurement could reach about 100,000 GPU cards.
  • If domestic GPU supply does increase as Tencent expects, cloud customers should see more available capacity and lower internal prioritization of enterprise models, and investors should watch for a return to lower capex after the stated one‑off phase ends.