GDS expands AI data centers as Chinese chip and model activity picks up
China’s AI compute news centered on data-center expansion, cooling and memory supply, model platform updates, and financing tied to computing-power usage.
• GDS raised its full-year guidance and plans to spend 10 billion yuan on data-center expansion this year, with chairman and CEO Huang Wei attributing order growth mainly to AI demand in China.
• A report on China’s planned new-type power grid said investment in the grid will reach 5 trillion yuan and be more than 80% higher than during the 14th Five-Year Plan period, with the grid framed as support for “computing power + green power.”
• TrendForce said liquid cooling is becoming standard in high-end AI infrastructure as rack-scale AI servers reach hundreds of kW, with liquid-cooling penetration in AI chips expected to rise from about 33% in 2025 to 53% in 2026.
• POWEV’s SINKER brand launched a 48GB DDR5-5600 RDIMM server memory module based on domestically made DRAM dies and compliant with JEDEC standards.
• Wuyang Auto Control said its optical-module liquid-cooling business has produced samples and entered proofing and small-batch supply, while warning that mass production remains uncertain.
• Alibaba Cloud launched the Qwen-Audio-3.0 speech model series on the Qwen AI platform and Alibaba Cloud Bailian platform, offering API access and Token Plan subscriptions.
• DeepSeek’s API peak-valley pricing took effect, setting off-peak prices at half of peak-hour prices and defining peak hours as 9:00-12:00 and 14:00-18:00 Beijing time.
• Biren Technology forecast first-half 2026 revenue of about 1.15 billion to 1.30 billion yuan, up roughly 1,852% to 2,107% year on year, while expecting its adjusted loss to narrow 35% to 47%.