The relentless pursuit of nanometer scaling is colliding with a physical wall, turning conventional power delivery architectures into a critical bottleneck. While Intel has long championed the shift to backside power delivery as a pivotal technological breakthrough, the sheer demands of transistor density now necessitate even more versatile solutions. With the upcoming 14A2 process node, the company intends to deploy a hybrid power delivery scheme leveraging both sides of the wafer—a strategic move designed to maintain voltage stability amidst the extreme miniaturization of transistor architectures.
For decades, the semiconductor industry has adhered to a conventional paradigm: all interconnects, including power delivery and data lines, were routed above the active transistor layer. This architecture inevitably created a routing bottleneck in the upper metal layers, where power rails competed for limited real estate with signal traces. Intel bet on a radical topological shift by moving the power delivery system to the backside of the silicon wafer. However, recent data suggests that even this ambitious transition may serve as a mere stepping stone toward even more complex configurations.
The primary challenge today lies in the scaling of the M0 metal layer. In the base version of the 14A process node, Intel aims to achieve a 28nm pitch between adjacent transistor structures. However, further progress and the push toward a 21nm threshold will require a modified iteration of the technology—14A2. This is where the concept of dual-sided power delivery enters the fray. While the backside of the wafer will remain dominant, the front side is poised to regain strategic importance.
This shift is dictated by the fundamental laws of physics: as interconnect geometry shrinks below 21nm, electrical resistance begins to grow exponentially. The infrastructure of nano-through-silicon vias (nTSV), which in theory should provide efficient power delivery from the rear, struggles to keep pace with ultra-high transistor densities. This results in voltage drops (IR-drop), which jeopardize chip stability and constrain maximum clock frequencies.
To mitigate these risks, Intel plans to redistribute the power load. A hybrid layout will allow the front side of the die to power auxiliary elements, thereby offloading the primary arteries on the backside. This approach creates the necessary power headroom and enables a more aggressive increase in component density without sacrificing energy efficiency.
The roadmap for these innovations is as follows: engineering samples of the base 14A process are expected by late 2028, with mass production slated for 2029. The 14A2 version is expected to arrive later, marking the next evolutionary step. Notably, Intel expects the initial 14A version to deliver a 1.3x increase in transistor density compared to the current 18A process, underscoring the company's drive to reclaim technological leadership in the advanced semiconductor segment.
The Evolution of Data Density in Kioxia Memory
