Transistors are now only tens of atoms wide, and that's the actual reason quantum matters
Intel's 18A process began ramping into high-volume production at its Arizona fab in late 2025, with the first chips broadly available from January 2026, retiring the FinFET transistor design that had governed chip manufacturing since 2011. TSMC's 2nm entered volume production in the fourth quarter of 2025.
Those names are labels, not measurements. The real critical dimensions are larger, but still tiny: Intel's 18A has a gate pitch of about 50 nanometres, and IBM's 2nm-class demonstration transistors had a gate length of 12 nanometres. Silicon's atomic lattice spacing is about 0.54 nanometres, so that gate is only about 22 lattice spacings long. The features that matter are now tens of atoms across.
The physics, not the marketing
At that scale, electrons stop behaving the way circuit design assumes. Quantum tunnelling, an electron passing through a barrier it classically shouldn't cross, becomes a real source of current leakage. This isn't a defect better equipment fixes. It's the physics of how small a silicon switch can get before it stops switching reliably. That's why the industry's response has shifted from shrinking further to architecture changes, gate-all-around transistors, backside power delivery, that extend silicon's life without asking it to get much smaller.
Why quantum computing is a different answer, not a faster one
A classical bit is on or off, switched by transistors. A qubit uses superposition and entanglement to process information in a way that doesn't hit the same atomic ceiling, because it was never trying to shrink a classical switch past its limit. It's a different computing paradigm.
That distinction is part of why serious capital is moving in. IBM committed more than $10 billion to quantum computing over five years in June 2026, funding its roadmap toward Starling, a fault-tolerant system targeted for 2029. This isn't speculative capital chasing a trend. It's one of the oldest names in computing hardware reading the same physics everyone else is reading.
Worth watching
Silicon keeps improving through architecture and packaging for a while yet. It stops improving through the method that built the entire industry, making the transistor smaller, because there isn't much smaller left. Quantum is gaining serious capital in part for that reason. It's worth tracking closely from here.
Sources
- Intel, "Intel Unveils Panther Lake Architecture: First AI PC Platform Built on 18A"
- Focus Taiwan, "TSMC officially begins 2nm chip volume production in Q4 2025"
- Wikipedia, "2 nm process" (gate pitch and gate length figures for Intel 18A and IBM's 2nm-class transistors)
- arXiv, "Quantum Computing's Classical Problem, Classical Computing's Quantum Problem"
- IBM Newsroom, "IBM Commits More Than $10 Billion to Quantum Computing"