Production-Cost Ratio — research note (Tier-1 item 4, aggregate leg)

BSAHI — producing-side instrument Produced: 2026-09-16

The instrument

tools/research/production_cost_ratio.py + data/production_cost_ratio.json. Deterministic from frozen daily series (hash-rate TH/s, miners-revenue USD/day, transaction-fees BTC/day, market-price USD — all under captured-data/historical/). Per era:

power_GW = hashrate_TH/s x network-average ASIC efficiency (J/TH) energy_cost_$day = power_x_hours x electricity $/kWh (assumption; scenario range shown below) production_value_$day = miners revenue (subsidy + fees) production_cost_ratio = energy_cost / production_value (network viability) fee_coverage_of_production = fee_usd_day / energy_cost (block-space's share)

Electricity scenarios — the assumption, made explicit

The single global price of $0.05/kWh is an assumption (grade C), not a measurement. Rather than hide it behind one number, the model publishes the production-cost ratio across plausible electricity prices. The ratio is linear in $/kWh, so this is the assumption shown as a range:

era	$0.03/kWh	$0.05/kWh	$0.08/kWh	$0.10/kWh	$0.15/kWh
2012	0.00	0.00	0.00	0.00	0.00
2013	0.00	0.00	0.00	0.00	0.00
2014	0.01	0.03	0.04	0.05	0.07
2015	0.09	0.15	0.23	0.29	0.44
2016	0.27	0.45	0.73	0.91	1.36
2017	0.19	0.32	0.51	0.64	0.96
2018	0.27	0.44	0.71	0.88	1.33
2019	0.33	0.56	0.89	1.11	1.67
2020	0.44	0.73	1.18	1.47	2.20
2021	0.12	0.20	0.33	0.41	0.61
2022	0.27	0.45	0.72	0.91	1.36
2023	0.36	0.59	0.95	1.19	1.78
2024	0.36	0.59	0.95	1.19	1.78
2025	0.40	0.66	1.06	1.32	1.99
2026	0.50	0.84	1.34	1.68	2.52

Bold = above 1.0 (the modeled network-wide energy bill exceeds what miners earn — subsidy + fees — on the flow-cost basis). Years that cross the wall:

$0.03/kWh: none
$0.05/kWh: none
$0.08/kWh: 2020, 2025, 2026
$0.10/kWh: 2019, 2020, 2023, 2024, 2025, 2026
$0.15/kWh: 2016, 2018, 2019, 2020, 2022, 2023, 2024, 2025, 2026

The point is not which price is "right" — it is that the producing-side stress is visible as a function of a decision variable instead of a hidden constant.

First reading (2026-09-16)

era	EH/s	GW	energyⱼₖ (kWh/day)	cost $/day	prod value $/day	cost/value	fee % of prod. cost
2017	6.3	2.5	60.7M	3.03M	9.44M	0.32	35.7
2019	66.8	6.7	160.2M	8.01M	14.4M	0.56	5.0
2020	120	8.4	202M	10.1M	13.8M	0.73	8.1
2023	380	14.4	347M	17.3M	29.2M	0.59	10.9
2024	636	20.3	488M	24.4M	41.2M	0.59	14.0
2025	937	26.2	630M	31.5M	47.6M	0.66	1.6
2026	953	23.8	572M	28.6M	34.1M	0.84	0.7

Total is in the CBECI-consistent ballpark (2017 ≈ 22 TWh/yr, 2026 ≈ 209 TWh/yr; efficiency table is a documented estimate, grade C).

Reading

2026: energy cost ≈ 84% of miner revenue at $0.05/kWh. Near the flow-cost viability wall; at $0.06/kWh the network's aggregate energy bill would exceed its revenue. This is the producing-side stress (boundary class E) in its strongest form yet — a cost squeeze, not a fee shortage symptom.
Block-space fees pay 0.7% of the energy cost that produces the blocks they live in (2026). The fee market internalizes ≈ nothing of the physical production cost. (2012-2017 fee coverage looks high only because the estimated energy bills were tiny — grade-C baseline effect.)
The trend is rising: cost/value went 0.32 (2017) → 0.56 (2019) → 0.73 (2020) → 0.59-0.66 (2023-25) → 0.84 (2026). Pressure is concentrated in subsidy-halving years as subsidy steps down against a growing energy bill.

Honesty block

Efficiency ($/kWh, J/TH) are documented assumptions, not measurements (grade C). The ratio is linear in both. The hashrate + revenue + fee + price legs are frozen primary series (grade B).
Aggregate only — regional split NOT included. The Cambridge-mining-map / regional-electricity capture that row 13 names is still grade D. This closes the "producing side has no instrument" gap, not the regional-granularity gap.
No number above is the regional production cost; do not read the two as the same thing.

Next step for this row

Regional share capture (Cambridge map or pool-country attribution) + regional electricity series (EIA/IEA) → row 13 D → B. Aggregate stays C (assumption-bound) until a measured network-average efficiency series is captured.

Data source: data/production_cost_ratio.json — deterministic, frozen-input instrument; re-runnable offline.

← All research · ← Back to Learn