Moore's law and process-node scaling

Prediction that components per chip double regularly as cost per function falls; it became the industry's roadmap for node shrinks.

Chips & processorsHardwareLandmark
  1. 1965

    First version

    Gordon Moore (Fairchild); scaling rules by Robert Dennard (IBM)

    Image: Fairchild Camera and Instrument Corporation · Unclear; no licence stated (corporate wordmark, trademark; 1965 advertisement) · Electronics magazine, 19 April 1965 (vol. 38 no. 8), p. 14 Fairchild advertisement; worldradiohistory.com scan (PDF page index 15)
  2. 1974
    Dennard scaling

    Dennard scaling

    Rules for shrinking MOS dimensions make each node faster, cheaper and lower-power.

    Image: Mister rf · CC BY-SA 4.0 · Wikimedia Commons
  3. 1975
    Revised pace

    Revised pace

    At IEDM Moore slows the projection to a doubling every two years.

    Image: Pauli Rautakorpi · CC BY 3.0 · Wikimedia Commons
  4. 2005
    Multi-core era

    Multi-core era

    Dennard scaling breaks down as clock speeds stall; extra transistors go into multiple cores per chip.

    Image: Mckaysalisbury · CC BY-SA 3.0 · Wikimedia Commons
  5. 2011
    FinFET era

    FinFET era

    22nm Tri-Gate transistors keep scaling going as planar transistors run out.

    Image: Fritzchens Fritz from Berlin · CC0 · Wikimedia Commons
  6. 2019
    EUV node

    EUV node

    TSMC N7+ is the first EUV process in high-volume customer products.

    Image: Bechology · CC BY-SA 4.0 · Wikimedia Commons
  7. 2025
    2nm class: nanosheets and backside power

    2nm class: nanosheets and backside power

    Intel 18A pairs gate-all-around transistors with backside power wiring; TSMC N2 nanosheets also enter volume production.

    Image: 极客湾Geekerwan · CC BY 3.0 · Wikimedia Commons

Contested: Moore's 1965 paper projected yearly doubling; in 1975 he slowed it to 'a doubling every two years'.

See it on the timeline