Answers are grounded in this paper and cite their sources.
Continuous complementary metal–oxide–semiconductor scaling has driven successive gains in computing performance, energy efficiency, and integration density. Beyond the 1‐nm node, however, lateral shrinkage yields diminishing returns as channel electrostatics and transport, parasitic resistance and capacitance, and lithographic cost cease to scale together. This review examines scaling across materials, devices, cells, and systems. At the channel level, SiGe‐ and Ge‐rich channels, two‐dimensional semiconductors, aligned carbon nanotubes, and III–V nanowires provide mobility, injection velocity, or electrostatic advantages, but require low‐resistance contacts, controlled dielectric interfaces, polarity control, uniformity, and wafer‐scale integration. At the device and cell levels, highly stacked and vertical‐channel gate‐all‐around field‐effect transistors (FETs) increase footprint‐normalized current or decouple gate length from lateral footprint, while complementary FETs and Flip FETs increase complementary‐device density and enable dual‐sided routing; source/drain resistance, capacitance, buried‐device access, high‐aspect‐ratio fabrication, and thermal budgets remain limiting. At the chip and system levels, system–technology co‐optimization, represented by TSMC's system‐integration roadmap and Huawei's τ‐Scaling and LogicFolding, redirects optimization from transistor density toward time and energy for useful computation. Connecting these levels within a materials–process–device–circuit–system framework establishes whether materials and vertical architectures deliver real current, density, and time benefits or merely transfer bottlenecks across the stack.
No takes yet. Share an insight, caveat, or question.
Zhao et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: