Bspdn Confront the Frontside Interconnect Congestion Problem
# Backside Power Delivery: Step 1 — Confront the Frontside Interconnect Congestion Problem
## 1. Power and Signal Have Always Shared One Stack
From the earliest planar CMOS processes through today's finest-pitch logic, power delivery
(the VDD and VSS rails that feed every transistor) and signal routing (the wires that carry
logic state between gates) have been built in the same back-end-of-line (BEOL) metal stack,
on the same side of the wafer, in the same limited number of layers above the transistor.
For decades this was simply how interconnect worked, and it was not a serious constraint:
early processes had so few transistors per unit area that there was plenty of metal to go
around for both jobs.
## 2. Density Growth Turned a Convenience Into a Bottleneck
As transistor density climbed generation over generation, the amount of signal wiring that
needed to cross a given area of chip grew faster than the number of metal layers available to
carry it. Power delivery still needed its own share of that same frontside stack -- wide,
low-resistance straps running across the die to keep supply voltage stable -- and those wide
power straps occupy routing tracks that signal wiring would otherwise use. The two needs
compete directly for the same physical resource.
## 3. Why This Series Exists
This series gives backside power delivery network (BSPDN) technology the dedicated treatment
it has not yet received in this knowledge base. Four prior transistor-architecture series --
most explicitly Gate-All-Around / Nanosheet Transistors (2020s) Step 12 and CFET Step 15 --
referenced BSPDN in passing as "a companion innovation" without explaining the problem it
solves, the process steps it requires, or how its commercial implementations differ. This
step opens that explanation from first principles: the congestion problem itself, before any
discussion of the fix.
## 4. Framing for the Series
BSPDN is treated here as a current, actively-ramping interconnect innovation -- being
introduced and reaching early production at the most advanced nodes -- not as a mature,
universally adopted standard. Later steps will walk through the wafer-flip process, nano-TSV
formation, buried power rail integration, and the three named commercial implementations
(Intel PowerVia, TSMC Super Power Rail, Samsung's approach) in turn.