Thermodynamic Budgeting for Hybrid Quantum-Classical Computing: A Three-Layer Energy Model with Explicit Arithmetic and a 15-Month Validation Roadmap

In the paper's words

Hybrid quantum-classical (HQC) computing is the dominant near-term paradigm, yet its energy accounting is rarely made explicit: the classical control, readout, cryogenic, and orchestration hardware surrounding the quantum processing unit (QPU) can dominate the total power budget, eroding any thermodynamic advantage the quantum subsystem might offer. We develop a quantitative framework that partitions the energy cost of hybrid workflows into three layers — quantum execution (including cryogenics), the classical-quantum interface (control/readout), and classical compute — parameterized by cycle time, shot count, error-correction overhead, and data volume crossing the boundary. We derive closed-form per-iteration energy expressions, a delegation (partitioning) criterion for assigning subtasks to the quantum layer, and a break-even condition under which quantum acceleration yields net thermodynamic benefit.

The abstract, verbatim, and its opening sentence is the claim. Every line below is quoted from the text.

Numbers it reports

1. For a representative variational-workload reference case (100 physical qubits, 10,000 shots/iteration, 1,000 iterations), we compute a total workflow energy of 27.7 MJ, of which the cryogenic/quantum layer accounts for 98.304%, the interface 0.49%, and classical compute 1.21%; the quantum subroutine must deliver a classical-equivalent speedup of ≈ 69× for thermodynamic break-even.

   from Abstract

2. We further show that (i) workload sharing across 10 concurrent workflows reduces total energy ≈ 6.8× and raises the non-quantum share to 11.5%, (ii) measurement batching reduces interface energy by a computed 8.3–10×, and (iii) at the device level, reversible control logic bounded by the Landauer limit ($k_{B} T \ln 2 \approx 2.87 \times 10^{-21} J$ at 300 K) offers up to a computed 5× reduction of the control-layer energy share, though this share is itself small at system scale.

   from Abstract

3. Workload W: a VQE-style loop in the pattern of [8,10]: $N_{\mathrm{iter}} = 1,000$ optimizer iterations; $N_{\mathrm{shots}} = 10,000$ shots per iteration (standard error ∝ $1/\sqrt{N} = 1/\sqrt{10},000 = 0.01$, i.e., ~1% statistical resolution); $t_{\mathrm{shot}} = 100 \mu s = 1 \times 10^{-4} s$ per shot (conservative for a ~100-qubit, depth-~20 superconducting circuit including reset and readout); $N_{q} = 100$ physical qubits, no error correction (NISQ regime of [8]).

   from 4.1 Reference workload and parameters

4. The quantum subroutine must deliver a classical-equivalent wall-clock speedup of ≈ 69× for thermodynamic break-even at baseline parameters.

   from 4.3 Break-even speedup

5. With amplitude-interception-style batching per [5]: full per-circuit batching reduces $n_{ex}$ from 100 to 10 exchanges per iteration, giving a 10.0× interface-energy reduction.

   from 4.5 Interface batching gain

6. With partial batching (2 of 10 bases unbatchable due to non-commuting observables, common in the chemistry settings of [10]), 12 exchanges remain: reduction = 1.05/0.126 = 8.33× .

   from 4.5 Interface batching gain

How it is done

1. We derive closed-form per-iteration energy expressions, a delegation (partitioning) criterion for assigning subtasks to the quantum layer, and a break-even condition under which quantum acceleration yields net thermodynamic benefit.

   from Abstract

2. As classical data centers approach power-delivery and cooling limits, any claim that quantum processors will "offload" computation must be tested thermodynamically, not merely asymptotically.

   from 1. Introduction

3. QELF's locality categories are, in energy terms, proxies for our interface data-volume term: quantum-local subtasks incur minimal interface cost, while communication-bound subtasks pay the interface tax per byte; our model makes that tax explicit and could give QELF classes quantitative energy semantics.

   from 2. Background and Related Work

4. We model one hybrid cycle (one optimizer iteration of a variational loop) as three additive energy terms: $E_{\mathrm{cycle}} = E_{Q} + E_{\mathrm{IF}} + E_{C}$ Quantum execution term. $E_{Q} = P_{\mathrm{cryo}} \times t_{\mathrm{cycle}} + P_{q,op} \times t_{Q}$, where $P_{\mathrm{cryo}}$ is the always-on cryogenic base power (dilution refrigerator plus compressors, approximately duty-cycle-independent), $P_{q,op}$ the incremental power of active qubit control, $t_{Q}$ the active quantum time per cycle, and $t_{\mathrm{cycle}}$ the full cycle time.

   from 3.1 System model and energy partition

What would falsify it

1. A device-level complement: a reversible-control-layer analysis bounded by Landauer's principle, with explicit falsifiability criteria, and a 15-month staged roadmap for simulation and validation.

   from 1. Introduction

2. Months 13–15: sensitivity analysis, falsification testing, and write-up.

   from 3.5 Simulation and validation plan (15 months)

3. Limitations and falsifiability.

   from 6. Discussion

Where the claim stops

1. The reversible-control analysis (R6) shows real device-level savings (5× on control energy per logical operation), but the system-level budget shows why this alone cannot deliver sustainability: the control layer is a fraction of a percent of baseline workflow energy.

   from 6. Discussion

2. Our result reads as a warning: at 100 noisy qubits, thermodynamic break-even requires an advantage class NISQ algorithms have not yet demonstrated.

   from 6. Discussion

3. (1) Parameter values are motivated estimates, not measurements; the structural claim that survives uncertainty is that cryogenic base load scales weakly with qubit count while dominating the budget, so utilization , not component efficiency, is the control knob.

   from 6. Discussion

Where to read

1. Abstract279 words · 1 min
2. 1. Introduction338 words · 1 min
3. 2. Background and Related Work902 words · 4 min
4. 3. Methods398 words · 2 min
5. 4. Analysis1,115 words · 5 min
6. 5. Results278 words · 1 min
7. 6. Discussion542 words · 2 min
8. 7. Conclusion125 words · 1 min
9. Appendix A. Divergence report263 words · 1 min