AI Data Center Power Collaboration

In late July, Delta Electronics announced an expanded partnership with Infineon Technologies to develop silicon‑carbide power components for next‑generation AI data‑center architectures, including 800 V HVDC and solid‑state transformers. The deal built on earlier joint work on high‑power vertical modules, while SongChuan Precision reported an 80 % revenue surge in its AI‑DC product line and disclosed a $4 billion capital‑spending plan to add capacity in Mexico, Vietnam and other sites to meet rising demand for AI data‑center power modules.

By early August, industry leaders were emphasizing that AI compute growth is now limited more by power consumption and data movement than by raw processing. At SEMICON Taiwan 2026, organizers highlighted this shift with dedicated platforms on ASIC and AI chip design, high‑bandwidth memory, and silicon photonics, stressing the need for co‑design across compute, memory and interconnect. Experts also noted that intelligent cluster middleware is essential to balance power use with workload demands, underscoring the centrality of power‑per‑token and memory‑speed considerations in AI training and inference.

Together, the snapshots trace a progression from specific corporate partnerships and capacity expansion toward a broader, industry‑wide focus on collaborative solutions to the power challenges of scaling AI data‑center workloads.