September 30, 2026 — today in AI.
1. AI Data Centers as Strategic Assets: Power, Cooling, and Semiconductors Take Center Stage
Singapore Semiconductor Industry Association exec Ang Wee Seng stated at EE Power Asia 2026 that AI has shifted data centers from processing-focused to thermal and power-focused infrastructure. Thermal management and electricity supply now carry equal weight with chip selection in competitive calculus.
2. General Compute Buys Large Fleet of Cerebras Chips to Enhance AI Inference Output
General Compute announced Sept. 29 it is purchasing a large fleet of Cerebras Systems wafer-scale chips to run alongside Nvidia GPUs in a hybrid inference architecture. This signals that single-vendor GPU strategies are giving way to heterogeneous platforms optimized for latency and cost arbitrage.
3. Why Big Tech Is Building Its Own AI Chips
Hyperscalers are building custom silicon to cut inference costs, reduce GPU supply dependency, and improve data center efficiency at scale. Vertical integration of chip design is now cost-justified; margin pressure and supply constraints are driving in-house silicon programs.
4. AI Industry Needs to Earn $6 Trillion by 2031 to Justify Data Centres
Data center sizes and costs are doubling every 12–16 months due to surging Nvidia and SK Hynix chip prices plus networking equipment inflation. The economics of AI infrastructure are tightening; breakeven timelines are extending and capital requirements per inference unit are rising.