Vaire Computing’s Ice River chip uses reversible logic and adiabatic computing to recycle energy during calculations. This proof-of-concept shows how AI chips might cut power waste, but scaling it to production remains the big challenge.
Hyperscale Data has announced a major expansion at its Michigan campus to host NVIDIA’s Blackwell architecture. The move marks one of the largest AI infrastructure buildouts in the Midwest, aimed at supporting high-demand GPU workloads for research, enterprise, and cloud operations.
Gigabyte’s new AI Top CXL R5X4 card lets select workstations add up to 1TB of DDR5 ECC memory, offering massive performance headroom for AI training and demanding content creation, but only if you have the right motherboard and budget.
Google partners with MediaTek to develop components for its next-gen TPUs, aiming for energy-efficient AI chips and reducing reliance on others.
The demand for AI computing power continues to grow as HPE secures a $1 billion deal with Elon Musk's X for high-performance servers optimized for AI workloads.