Future Thinking Lab
The AI Bottleneck Isn't the Chip, It's the Package
5:47
Why are AI chips still hard to get? Often the holdup isn't the transistors. It's the step after the chip is printed: advanced packaging.
In this episode, Helen explains how AI accelerators work with HBM (high-bandwidth memory) stacked right beside the GPU. She covers what TSMC's CoWoS packaging (Chip on Wafer on Substrate) actually does, and why the reticle limit, the largest area a lithography…