Sandberg, Armstrong & Cirković (2017) argue that a civilisation can perform vastly more irreversible computation by waiting for the cosmic microwave background to cool. Bennett, Hanson & Riedel (2019) reply that the gain is illusory because cold reservoirs already exist. Both sides, with their math, side by side.
arXiv:1705.03394
Foundations of Physics 49:820 · DOI 10.1007/s10701-019-00289-5
Erasing one bit at temperature T dissipates at least E_min = k_B · T · ln 2 joules. For N irreversible bit-erasures, the total minimum energy cost is E = N · k_B · T_sink · ln 2 where T_sink is the temperature of the cold reservoir into which entropy is dumped. Reversible computation has no such bound — you can in principle do unlimited reversible operations for zero energy — but most useful computation produces irreversible steps (memory creation, error correction, decisions). The factor is linear in T_sink, which is the lever both sides argue over.
If a civilisation is forced to use the CMB as its cold reservoir (because it has saturated all local entropy sinks and is now dumping entropy into the only available wide-open low-temperature surface), then waiting Δt years cools the CMB by a factor of a(t0+Δt)/a(t0). In ΛCDM with current parameters, the late-era expansion gives a(t) ≈ exp(HΛ(t−t0)) with HΛ ≈ 1.8×10⁻¹⁸ /s ≈ 5.7×10⁻¹¹ /yr. Waiting 10¹¹ years cools the CMB by exp(5.7) ≈ 300×; waiting 10¹² years cools it by exp(57) ≈ 10²⁵×. The Landauer cost drops in proportion. Sandberg et al. estimate that by waiting until the CMB is cold enough, a civilisation could perform on the order of 10³⁰ times more irreversible operations than if it computed now.
The 2019 reply argues the premise fails. Today's universe is not in maximum-entropy equilibrium — there are stars (hot reservoirs at ~6000 K), an interstellar medium, the CMB itself (2.725 K), and engineered cold sinks (dilution fridges at millikelvin, vacuum-pumped traps colder still). A civilisation that wants to do irreversible computation can dump entropy into any reservoir whose temperature is below its compute temperature. The CMB is just one option, and not the coldest one available. Furthermore, the universe contains other low-entropy resources (gravitational potential, baryon-asymmetry, etc.) that today can be used as entropy sinks. "This can be done at any time, and is not improved by waiting for a low cosmic background temperature." The aestivation gain over a current-tech civilisation is therefore not 10³⁰ but more like 10⁰–10³ — significant but not enormous, and entirely contingent on engineering you could equally well do today.
Sandberg side: E_now = N · k_B · 2.725 · ln 2 · (1 − r); E_wait = N · k_B · T_CMB(Δt) · ln 2 · (1 − r) where r is the reversible fraction and T_CMB(Δt) uses the ΛCDM expansion above. Bennett side: same Landauer with T_sink set to the user-chosen "cold reservoir today" temperature. The third row of each panel compares the two costs; the headline ratio is the punchline.
The Macro Transcension Hypothesis and aestivation are different answers to the Fermi paradox: MTH says civilisations compress into black-hole substrates and stop expanding; aestivation says they're alive but dormant, waiting. Both predict observable silence; both predict no Dyson swarms. The Bennett critique, if correct, removes most of the motivation for aestivation specifically — but doesn't rule it out for civilisations that genuinely want to wait. The aestivation gain at any wait time can be compared to the ergosphere ops/sec from Bekenstein-Landauer-Lloyd Tool 4, Panel 2: that's the MTH counter-proposal in concrete numbers.
Sandberg, Armstrong & Ćirković 2017 (J Brit Interpl Soc 70:406; arXiv:1705.03394) formalised the aestivation hypothesis: any sufficiently advanced civilisation that is compute-limited (rather than energy-limited) has an incentive to "sleep" through the early universe and resume operation only when the cosmic microwave background has cooled enough to dramatically lower the Landauer floor. The CMB is currently 2.725 K; in a flat ΛCDM universe it cools as 1/a, so by t ≈ 10²⁵ yr it reaches ~10⁻¹⁵ K. The compute speedup factor over the same energy budget is roughly (T_now/T_future)^n where n depends on whether the bound is Landauer (n=1) or thermal noise (n=2) — for n=1 and a factor 10¹⁵ in temperature, that's a 10¹⁵× speedup per joule.
Heat death (matter-dominated era ends): ~10¹⁴ years. Stellar epoch end: 10¹⁴ years (no new star formation). Degenerate-stars era: 10²⁰ years. Proton decay theoretical: 10³⁴ years. CMB temperature at 10¹⁰⁰ years ≈ 10⁻⁹⁰ K. For an aestivating civilisation, the optimal wake-up time is set by the trade-off between (a) cooler universe = cheaper compute and (b) shrinking accessible energy budget as stars die and the universe expands. Sandberg et al. argue the sweet spot is around 10²⁰ to 10²⁵ years from now.
Bennett, Hanson & Riedel 2019 (Foundations of Physics 49:820; DOI 10.1007/s10701-019-00289-5) argued that mere waiting doesn't solve the Fermi paradox because the dark-energy-dominated late universe expands accessible matter away faster than waiting can recover compute efficiency. Aestivators would lose more in energy budget than they gain in Landauer floor reduction. The debate is alive — a related literature on the Black Hole Era and ultra-late-universe physics continues.