The most important signals in chips this year are less about transistor counts and more about the surrounding constraints: AI compute demand, government chip policy, dependence on mature manufacturing nodes, advanced-packaging bottlenecks and power semiconductors. Packaging, the step that stitches multiple chip components into one high-performance part, has become a limiting factor for the most advanced AI hardware.
Power is the other quiet constraint, as data centres full of accelerators strain electricity supply and cooling.
What it means: the limits on AI are increasingly physical, such as packaging capacity and power, not only design. That is why availability and cost of AI compute stay tight even as new chips launch, and why efficient use of what exists matters more than waiting for the next part.