Methodology

Core idea

Given a quantity of electrical energy (in kWh), the calculator estimates how much Bitcoin the global mining network would statistically produce with that energy, under current network conditions. It also estimates an equivalent count of AI inferences for comparison.

Formulas

hashesPerKwh = (JOULES_PER_KWH / AVG_EFFICIENCY) × HASHES_PER_TH
btcPerHash = blockSubsidy / (difficulty × 2³²)
btcPerKwh = hashesPerKwh × btcPerHash
btcEquiv = userKwh × btcPerKwh
sats = btcEquiv × 100,000,000
fiatUsd = btcEquiv × btcUsdSma200
aiInferences = (userKwh × 1,000) / AI_WH_PER_INFERENCE

Constants and assumptions

AVG_MINING_EFFICIENCY = 27 J/TH — approximate fleet-wide average of active mining hardware.
JOULES_PER_KWH = 3,600,000
HASHES_PER_TH = 10¹²
DIFF_CONSTANT = 2³² (4,294,967,296)
AI_WH_PER_INFERENCE = 1 Wh — approximate energy for one large-model inference.
Block reward uses subsidy only (fees excluded).
Fiat equivalent uses the 200-day simple moving average of BTC/USD to smooth volatility.

Scale details

Block reward fraction: btcEquiv / blockSubsidy, displayed as "1 / N of a block".
Network time: the duration for which the user's energy would power the entire network. userJoules / networkPowerWatts, where networkPowerWatts = hashrateTHs × AVG_EFFICIENCY.
Household equivalent: the user's kWh divided by a country's average monthly household consumption.

Daily snapshot

The calculator reads a local snapshot.json file containing network difficulty, hashrate, the 200-day SMA price, and the current block subsidy. This file is regenerated daily at 00:10 UTC by a server-side process. No external API calls are made from the browser.

Data sources

Difficulty & hashrate: derived from Bitcoin Core RPC or public blockchain APIs.
BTC/USD 200-day SMA: computed from daily close prices via exchange data aggregators.
Block subsidy: deterministic from block height (currently 3.125 BTC post-April 2024 halving).
Household electricity data: sourced from IEA and national energy agencies (approximate).

Limitations

The 27 J/TH efficiency is a fleet-wide estimate. Individual hardware varies from ~15 to 60+ J/TH.
Transaction fees are excluded; actual miner revenue per hash is higher.
Mining economics (pool luck, payout methods, operational costs) are not modeled.
The 200-day SMA lags actual price; it is intentionally conservative.
The AI inference estimate is a rough order-of-magnitude figure. Real-world inference costs depend heavily on model size, hardware utilization, and batch size.
Snapshot data may be up to ~24 hours stale.

AMP → Bitcoin v1.0