AI Price War Reshapes Data Centre Demand, Capacity Deals
OpenAI and Anthropic cut model prices, including OpenAI lowering GPT-5.6 Luna API pricing by 80% to $0.20 per million input tokens and $1.20 per million output tokens and Anthropic pricing Claude Opus 5 at about half the cost of its prior flagship, in response to Chinese rivals such as DeepSeek, Moonshot’s Kimi, and Zhipu’s GLM undercutting Western models by up to nine times. The pricing shift changes how AI workloads are served by pushing more inference volume through GPU-dense data centre infrastructure, with capacity demand, usage-based billing, and flexibility in power and cooling becoming central to the operating model. It matters because US hyperscale and colocation operators, plus data centre markets in the Gulf, Southeast Asia, and parts of Europe, may see higher demand for GPU capacity or lose workloads to Chinese-hosted and open-weight alternatives, affecting long-term leases, PPAs, and build-to-suit deals.
Read Source