Cited by 1,120Advanced Thermodynamics and Statistical MechanicsComputability, Logic, AI AlgorithmsBlack Holes and Theoretical Physics

The DOI matches its Crossref record; 95 of the 96 references it cites were found in a registry or an index; none of the 82 sources its claims cite is checked for support yet.

Ultimate physical limits to computation

Open access
Lloyd, Seth·2000·Nature

Abstract

Computers are physical systems: the laws of physics dictate what they can and cannot do. In particular, the speed with which a physical device can process information is limited by its energy and the amount of information that it can process is limited by the number of degrees of freedom it possesses. Here I explore the physical limits of computation as determined by the speed of light c, the quantum scale ℏ and the gravitational constant G. As an example, I put quantitative bounds to the computational power of an ‘ultimate laptop’ with a mass of one kilogram confined to a volume of one litre.

Overview

AI snapshot

Written from the paper’s full text.

TL;DR

Lloyd derives physical limits on computation from the speed of light, quantum scale, and gravitational constant, using a 1-kg, 1-L “ultimate laptop” as an example. Energy limits speed to \(2E/\pi\hbar\) operations per second, while entropy limits memory, yielding \(5.4258\times10^{50}\) operations per second and about \(10^{31}\) bits in the thermal/black-body estimate.

Key findings

  • The ultimate laptop has mass 1 kg, volume 1 L, and energy \(E=mc^2=8.9874\times10^{16}\) J; it can perform a maximum of \(5.4258\times10^{50}\) logical operations per second and, using all memory, about \(10^{19}\) operations per bit per second.
  • Speed is energy-limited: an elementary logical operation in time \(\Delta t\) requires \(E \ge \pi\hbar/2\Delta t\), so a system with average energy \(E\) can perform at most \(2E/\pi\hbar\) operations per second; this total is independent of parallel versus serial architecture.
  • Memory is entropy-limited: information is \(I=S(E)/k_B\ln 2\). A black-body estimate for 1 kg in 1 L gives \(T=5.87\times10^8\) K, \(S=2.04\times10^8\) J/K, and \(I=2.13\times10^{31}\) bits; a 1-kg black-hole limit gives \(3.827\times10^{16}\) bits and the same \(5.4258\times10^{50}\) operations per second, but fully serial computation.
  • Error correction imposes a thermal load: black-body radiation allows \(B=7.195\times10^{42}\) bits per m² per second; with surface area \(10^{-2}\) m² and about \(10^{50}\) operations per second, the error rate must be less than \(10^{-10}\) per operation, with energy throughput \(4.04\times10^{26}\) W, suggesting operation below the maximum speed.

Methods

  • Theoretical derivations from \(c\), \(\hbar\), \(G\), and \(k_B\), using the quantum speed limit from time-energy uncertainty and the Margolus-Levitin result; universal logic is discussed via Toffoli/Controlled-Controlled-NOT operations.
  • For memory, the 1-kg, 1-L system is modeled as a collection of elementary-particle modes in the canonical ensemble at fixed average energy; maximum entropy is estimated, with a lower bound from black-body photons and a black-hole entropy calculation for the compressed limit.
  • Comparisons are made to existing quantum computers (NMR, optical-atomic, ion-trap) and to parallel/serial trade-offs via communication time versus bit-flip time.

Limitations

  • The paper states that there is no guarantee these limits will ever be attained; black-hole computation is likely extremely difficult or impossible to realize, complete control over all degrees of freedom is difficult, packaging issues make the memory limit unlikely, and exact Planck-scale behavior requires a theory of quantum gravity.
  • The paper states that error-correction thermal loads suggest operating slower than the maximum, and that Avogadro-scale memory would face severe input/output problems (e.g., \(10^{23}\) bits at \(10^{12}\) bits/s would take about 10,000 years for a serial read/write).
  • Evident from the text: the numerical estimates depend on simplifying assumptions, including zero angular momentum and electric charge, a black-body lower bound, species-count estimates, and applying the canonical ensemble to a closed system with acknowledged subtlety.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/35023282

Cites

95 of 95 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

0 of 82 cited sources checked · 0 supported · 0 flagged

Show all 98 claims

  • It may state two separate claims as one. 33
  • Probably not a substantial claim. 15
  • The quote only partly states this claim. 1
47 of 47
  • The authors state this as an assumption.

    We have complete control over the actual state of the ultimate laptop, and are able to guide it through its logical steps while insulating it from all uncontrolled degrees of freedom.

    “As above, we assume that we have complete control over the actual state of the ultimate laptop, and are able to guide it through its logical steps while insulating it from all uncontrolled degrees of freedom.”
  • The authors state this as an assumption.

    The computer encodes erroneous bits in the form of black-body radiation at the characteristic temperature 5.87 × 10^8 K of the computer's memory.

    “To estimate the rate at which it can reject errors to the environment, assume that the computer encodes erroneous bits in the form of black-body radiation at the characteristic temperature 5 . 87 × 10 8 K of the computer's memory.”
  • The authors state this as an assumption.

    The volume occupied by the computer can be modeled as a collection of modes of elementary particles with total average energy E.

    “The idea is simple: model the volume occupied by the computer as a collection of modes of elementary particles with total average energy E .”
  • The authors state this as an assumption.

    The only conserved quantities other than the computer's energy are angular momentum and electric charge, which are taken to be zero.

    “For this calculation, assume that the only conserved quantities other than the computer's energy are angular momentum and electric charge, which we take to be zero.”
  • The authors state this as an assumption.

    Energy and entropy are dominated by black body radiation consisting of photons.

    “A lower bound on the entropy can be obtained by assuming that energy and entropy are dominated by black body radiation consisting of photons.”
  • The authors state this as an assumption.

    All available bits are used.

    “In keeping with the prescription that the ultimate laptop operates at the absolute limits given by physics, in what follows, we assume that all available bits are used.”
  • The authors state this as an assumption.

    Control of highly energetic systems in which all energy is put to use is possible in principle.

    “In discussing the computational power of physical systems in which all energy is put to use, we assumed that such control is possible in principle, although it is certainly not possible in current practice.”
  • The authors state this as an assumption.

    It is possible to prepare initial states of black-hole or quantum-gravity systems that cause their natural time evolution to carry out a computation.

    “At the greater extremes of a black hole computer, we assumed that whatever theory (string theory, M theory?) turns out to be the correct theory of quantum matter and gravity, it is possible to prepare initial states of such systems that causes their natural time evolution to carry out a computation.”
8 of 8

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Error-rate formula disagrees with the defined time ratio

    Minor

    Using the paper's definitions of tfl ip and t com gives 2 tfl ip/t com = π ̄hcS/(2 k B ln 2 ER). That does not equal the expression printed for the maximum error rate.

  • Black-hole energy per bit disagrees with the stated temperature

    Minor

    Substituting R S = 2 Gm/c2 into the stated temperature gives k B T = ̄ hc3/(8πGm). The stated energy per bit is therefore 2 ln 2 k B T, not ln 2 k B T/2-a factor-of-four discrepancy.

  • Figure 2 gives an incompatible communication time

    Minor

    Figure 2 says communication across the laptop takes about 10^9 seconds, while a bit flips about 10^19 times per second. Those values give roughly 10^28 flips per communication time, contradicting the paper's stated ratio of about 10^10.

3 of 3

Citations

Cites

Flagged first, then by citations

Show all 96 references

1 of the 96 references couldn’t be matched to a known paper.

Research works printed but not linked: 10

  1. Levitin, L. B. in Proceedings of the 3rd International Symposium on Radio Electronics part 3, 1–15 (Varna, Bulgaria, 1970).
  2. Margolus, N. & Levitin, L. B. in Proceedings of the Fourth Workshop on Physics and Computation—PhysComp96 (eds Toffoli, T., Biafore, M. & Leão, J.) (New England Complex Systems Institute, Boston, MA, 1996).
  3. Lecerf, Y. Machines de Turing réversibles. C.R. Acad. Sci. 257, 2597–2600 (1963).
  4. Seitz, C. L. et al. in Proceedings of the 1985 Chapel Hill Conference on VLSI (ed. Fuchs, H.) (Computer Science Press, Rockville, MD, 1985).
  5. Younis, S. G. & Knight, T. F. in Proceedings of the 1993 Symposium on Integrated Systems, Seattle, Washington (eds Berrielo, G. & Ebeling, C.) (MIT Press, Cambridge, MA, 1993).
  6. Margolus, N. in Complexity, Entropy, and the Physics of Information, Santa Fe Institute Studies in the Sciences of Complexity Vol. VIII (ed. Zurek, W. H.) 273–288 (Addison Wesley, Redwood City, 1991).
  7. Cory, D. G., Fahmy, A. F. & Havel, T. F. in Proceedings of the Fourth Workshop on Physics and Computation—PhysComp96 (eds Toffoli, T., Biafore, M. & Leão, J.) 87–91 (New England Complex Systems Institute, Boston, MA, 1996).
  8. Zel'dovich, Ya. B. & Novikov, I. D. Relativistic Astrophysics (Univ. of Chicago Press, Chicago, 1971 ).
  9. Novikov, I. D. & Frolov, V. P. Black Holes (Springer, Berlin, 1986).
  10. Thorne, K. S., Zurek, W. H. & Price R. H. in Black Holes: The Membrane Paradigm Ch. VIII (eds Thorne, K. S., Price, R. H. & Macdonald, D. A.) 280– 340 (Yale Univ. Press, New Haven, CT, 1986).

Cited by

Sorted by citations

Items 201–300 of 999

Metadata

Authors
Lloyd, S.
Affiliations
Massachusetts Institute of Technology (US)
Corresponding author
Lloyd, Seth
Venue
Nature
Published
2000-08
Volume
406
Issue
6799
Pages
1047-1054
ISSN
  • 0028-0836 (ISSN-L)
  • 1476-4687
OpenAlex
W1615222761
Cited by
1,120
References
96
Citations by year
2026: 49 · 2025: 78 · 2024: 69 · 2023: 63 · 2022: 78 · 2021: 64 · 2020: 71 · 2019: 81 · 2018: 83 · 2017: 63 · 2016: 45 · 2015: 39 · 2014: 34 · 2013: 31 · 2012: 21
FWCI
10.36
Type
journal article
Language
en
Open access
Locations
  • Nature (publishedVersion)
  • PubMed (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • CERN Document Server (European Organization for Nuclear Research) (submittedVersion)
  • arXiv (Cornell University)
Topics
  • Physical Sciences › Physics and Astronomy › Statistical and Nonlinear Physics › Advanced Thermodynamics and Statistical Mechanics
  • Physical Sciences › Computer Science › Computational Theory and Mathematics › Computability, Logic, AI Algorithms
  • Physical Sciences › Physics and Astronomy › Nuclear and High Energy Physics › Black Holes and Theoretical Physics
Keywords
physical limits of computation, computational power, speed of light, quantum scale, gravitational constant, degrees of freedom, information processing

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-10-06, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-06 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review