HN Debrief

IBM debuts sub-1 nanometer chip technology

  • Hardware
  • Semiconductors
  • Infrastructure
  • Enterprise

IBM posted a press release for a new research-stage chip technology it brands as “sub-1 nanometer” or “7 angstrom.” The company says it uses a vertically stacked “nanostack” architecture to keep scaling logic chips moving forward, and says production could be possible in about five years. The big clarification is that this is not a chip with sub-1 nm horizontal features etched across the die. The visible structures in IBM’s own images are much larger. The “0.7 nm” label is being used as a node name tied to claimed density or performance equivalence, not as a literal physical dimension.

Treat the “0.7 nm” label as a process-generation brand, not a physical measurement. If you track chip roadmaps or vendors, compare on transistor density, power, performance, and commercialization path instead of node names alone.

Discussion mood

Skeptical but not dismissive. Most comments saw a genuine research milestone buried under marketing that overstates what “0.7 nm” means, and they were far more interested in density, PPA, and licensing prospects than in IBM’s headline.

Key insights

  1. 01

    Node names now mean generation, not size

    Modern process labels work as shorthand for expected PPA and density, not as ruler-accurate dimensions on silicon. That makes IBM’s wording less shocking inside the industry than it looks from the outside, but it also means node numbers have lost most of their value for cross-vendor comparison. Intel 10 nm and TSMC 7 nm were cited as the classic reminder that the same number can hide very different physical implementations.

    If you compare foundries or chip vendors, build your scorecard around density, performance per watt, and delivered products. Treat node labels as branding unless the company also publishes the underlying metrics.

      Attribution:
    • formerly_proven #1
    • api #1
    • jadar #1
  2. 02

    IBM sells process know-how more than chips

    IBM’s likely business here is not volume manufacturing. It is licensing, technology transfer, and joint development with companies that do own production fabs. Rapidus was mentioned as an example for IBM’s 2 nm work. Albany matters because it is where IBM, ASML, and state-backed research infrastructure meet. The lab helps de-risk future manufacturing even if IBM never ships wafers at scale from that site.

    Read IBM semiconductor announcements like upstream platform bets. The value is in who adopts the process and which ecosystem partners line up around it, not in whether IBM becomes a merchant foundry.

      Attribution:
    • wmf #1
    • topspin #1 #2
    • petcat #1
  3. 03

    The technical details are in papers, not the press release

    The release is marketing-heavy, but commenters pointed out that IBM has published scientific material behind it and that outside analysts have already done long technical breakdowns. That changes the posture from "empty stunt" to "real research that is being oversold." The gap is not whether there is engineering substance. It is whether the public-facing label describes that substance honestly.

    For semiconductor announcements, do not stop at the corporate headline. Check whether there is a paper, a process deck, or a credible analyst teardown before deciding if the claim is fluff or signal.

      Attribution:
    • IanCutress #1
    • wmf #1
  4. 04

    IBM chips still matter in hidden enterprise systems

    IBM’s silicon is easy to miss because it mostly lives behind mainframes, POWER servers, and old system-of-record workloads rather than consumer devices. Banks, retailers, telecom gear, and government systems still run on IBM platforms. That gives IBM a strategic reason to stay close to advanced chip research even after offloading most production.

    Do not mistake low consumer visibility for irrelevance. If your company touches finance, retail operations, telecom, or government infrastructure, IBM hardware decisions can still ripple into your stack.

      Attribution:
    • ternaryoperator #1
    • lizknope #1
    • xxpor #1
  5. 05

    Physics is not the only limit anymore

    Comments on the physical floor were more nuanced than "atoms are the limit." Silicon transistors already run into quantum leakage and material constraints, but manufacturing, cooling, and 3D integration are just as binding. The bottleneck is no longer simple planar shrink. It is getting useful switching behavior, thermals, and yield from more complex structures and materials.

    Expect future progress to come from packaging, architecture, materials, and specialized designs as much as from raw transistor shrink. Product roadmaps that assume old Moore’s Law-style scaling will keep arriving on schedule are setting themselves up for disappointment.

      Attribution:
    • vitally3643 #1
    • marcosdumay #1
    • mrguyorama #1

Against the grain

  1. 01

    The sub-1 nm label is materially misleading

    The harshest critics did not accept the usual "node names are marketing" defense. They argued IBM’s own images show features far larger than 1 nm and that the company is borrowing the aura of atomic-scale fabrication for something closer to a density-equivalence claim. From that angle, this is not harmless shorthand. It is a category error that invites the wrong mental model.

    If you communicate deep-tech milestones to customers or investors, avoid labels that imply a physical fact you are not actually demonstrating. The short-term attention is not worth the credibility hit.

      Attribution:
    • buran77 #1
    • adrian_b #1
    • scythe #1
    • TallGuyShort #1
  2. 02

    IBM has a long history of research without market payoff

    Some commenters saw the announcement through IBM’s institutional track record rather than the device physics. IBM has produced plenty of impressive lab work, shed its fabs, and often failed to turn breakthroughs into products it controls. That makes every new node headline feel detached from actual market power.

    Watch for named manufacturing partners, transfer timelines, and product commitments before assigning strategic weight to IBM research news. Technical novelty alone does not tell you who will capture the value.

      Attribution:
    • lizknope #1
    • d3m0t3p #1
    • bozhark #1

In plain english

ASML
The Dutch company that dominates the market for the most advanced chip lithography machines.
High NA EUV
High numerical aperture extreme ultraviolet lithography, a newer chipmaking system that can print smaller features with higher optical precision than earlier EUV tools.
PDK
Process design kit, the package of rules, models, and files chip designers need to design chips for a specific manufacturing process.
POWER
A family of IBM server processors used in enterprise systems.
PPA
Power, performance, and area, the three core tradeoffs chip designers use to judge a process or design.

Reference links

Technical analysis and background

IBM business and commercialization

Lithography and infrastructure

History of limits and scaling

IBM research history