Bitcoin Hash Rate

Bitcoin hash rate is rendered on a logarithmic right axis in exahashes per second. The span is enormous — roughly 10⁻⁶ EH/s in 2011 to over a thousand today, nine orders of magnitude. Only a log scale preserves the structure across that range. BTC price sits muted on the left log axis behind the network line for cycle context.

The first thing to internalise is that no instrument anywhere counts this number. There is no oracle polling every miner and summing throughput. Every published figure — this page's included — is a model fitted to two observable quantities: the protocol's current difficulty target, and the realised time between blocks. Ben Celermajer put it cleanly: “In a distributed process like mining, it is near impossible to obtain reliable hash rate figures from the universe of miners.” Today's reading is 908.1 EH/s, a 30-day change of −5.8%, against an all-time high of 1305.5 EH/s on 2025-10-24.

The standard derivation, documented on the Bitcoin wiki, is straightforward. Difficulty D is calibrated every 2,016 blocks so that the previous epoch would have been found at the rate of one block every ten minutes:

hashrate ≈ D · 2³² / 600

where 600 is the target block interval in seconds and 2³² is the size of the search space per difficulty unit. The same two quantities back out a daily hash-rate series, smoothed into a continuous EH/s line. One gigahash per second is 10⁻⁹ EH/s; days under one TH/s are dropped to keep the log axis legible. Full derivation lives on methodology; provenance is documented on data sources.

The asymmetry to remember: short-window readings carry meaningful variance from the Poisson statistics of block discovery alone. Most reference dashboards smooth seven days for that reason. This page surfaces the daily series and the 30-day rate of change, with the smoothing happening at read time on the chart's own zoom levels — so the noise is on the page, not hidden behind it.

Two readings carry the chart. The first is the absolute level — today's EH/s versus the all-time high — for context on where the network sits in its long-run growth curve. The second is the 30-day rate of change, which surfaces the local regime: rebuilding hard, growing steadily, or capitulating. The level moves over years; the rate of change is what carries information about what miners are doing this month. The bands below carry the historical context for each rate-of-change window.

Six order-of-magnitude milestones bracket the network's history, plus the single-day all-time high. Crossings are computed inline from the daily-close hash-rate series; spot-price-on-day comes from the daily-close history. The pre-2020 cadence stitches in coarser data, so the early crossing dates resolve to the day data is available rather than the day each threshold was first breached intraday. The compression is the story: it took years to climb from one TH/s to one EH/s, and the network has since added hundreds of EH/s in a single cycle.

The headline story of the post-2016 hash-rate climb is not energy. It is efficiency. Each generation of flagship Bitcoin ASIC has hashed more per joule than the last, and the cumulative improvement from the Antminer S9 in 2016 to the S21 Pro in 2024 is roughly six- to seven-fold. That decoupling is what makes the security-budget framing of hash rate misleading on its own: today's ATH is bought partly with new energy and partly with the same energy doing more work. The table below traces the generational ladder.

Hash rate is a power × efficiency product. If energy spend stayed flat from 2016 to 2024, the same farms running S9s would have produced one-sixth the EH/s the same farms running S21 Pros produce today. The 6-7× efficiency gain is therefore an upper bound on the share of today's ATH that is “free” hash — bought by replacing rigs rather than adding wattage. In practice the rate of efficiency adoption is uneven (older farms run mixed fleets) but the directional point holds: a 10× rise in hash rate over a half-cycle is not a 10× rise in real-resource commitment. Hash rate is therefore a biased security-budget proxy, and the bias has gotten stronger every cycle.

It is also a proxy that double-counts in the wrong direction during a chip transition. When a new flagship ships, farms swap fleets and the EH/s line steps up sharply even if their power draw — and their dollar commitment to securing the chain — barely moves. Anyone reading the all-time-high headline as “the network has never been more expensive to attack” is conflating throughput with cost. The honest version of that claim runs through issuance and miner revenue, not the raw hash figure.

The 2021 China-ban window is the most important data event in the entire series, and the cleanest case study in what hash rate actually measures. Bitcoin's network hash rate fell from a pre-shock peak of 198.5 EH/s on 2021-04-15 to a trough of 58.5 EH/s on 2021-06-27 — a 70.6% drop in roughly ten weeks — as Chinese provinces, chiefly Inner Mongolia and Sichuan, pushed local mining operators offline. Recovery to the pre-shock level reached 2022-01-01 and the network printed a fresh ATH shortly after. The canonical Cambridge mining-map record documents the geographic redistribution of those rigs from China to the United States, Kazakhstan, and Russia.

Read the 2021 collapse as an inventory-relocation event, not an economic capitulation. The rigs were not unprofitable — they were illegal where they stood. The same physical ASICs that printed the trough were containerised, shipped, and re-energised to print the recovery. That is why the rate-of-change pane shows the steepest decline and one of the steepest rebuilds in the series back to back: the same fleet, briefly dark in transit. It is the cleanest example of why hash rate describes ASIC logistics as much as it describes miner conviction.

A single day near the record is a coin-flip, not a confirmation. The daily estimate can swing several percent on the Poisson statistics of block discovery alone — a fortnight of fast blocks reads as rising hashpower with no rig added. The August 2024 post-halving window is instructive: the network slow-bled into the Capitulation band for weeks as the reward halving squeezed margins, but a handful of those daily prints were pure block-luck noise, not hardware leaving. Trust the 30-day rate of change; treat any one-day “new ATH” within a few percent of the old one as statistically indistinguishable from it.

Relocations and weather masquerade as conviction. The 2021 China ban (the 70.6% drop above) was rigs in shipping containers, not miners surrendering. The same pattern recurs at smaller scale every summer: the Texas grid curtailments of July–August 2023, when ERCOT paid miners to power down during heat-wave peaks, dented the series for reasons that had nothing to do with the security budget or miner sentiment. A drop is not automatically a capitulation; check whether the cause was economic before you read it as one.

Hash rate lags price — it does not lead it. Fantazzini and Kolodin's 2020 study Does the Hashrate Affect the Bitcoin Price? finds unidirectional Granger-causality from price to hashrate in their second sub-sample (December 2017 to February 2020); their first sub-sample (August 2016 to December 2017) shows neither direction as significant. The popular hash-rate-is-bullish narrative, which treats hash-rate ATHs as a leading indicator for price, has weak empirical footing. Hardware is ordered after price rises, installed months later, and capitulates after price falls — the causality runs the wrong way for forecasting.

The headline number sits inside two big mining pools. As of late 2025, Foundry USA and AntPool together account for roughly half of the network's hash rate — Foundry runs in the 30% range, AntPool in the high-teens, with combined share fluctuating around 50–60%. The security-budget reading of an all-time-high hash rate ought to be discounted by that intermediary concentration: the network is growing, but the surface against a co-ordinated rewrite is narrower than the raw EH/s figure suggests.

If you accumulate on a schedule, hash rate asks nothing of you. It is the slowest, most descriptive lens on the network's growth and the least useful for short-window timing. Read it for orientation — the way you would read a power-law corridor — not for entries. The actionable cousin is the smoothed Hash Ribbon, which is built on this exact series and lives on its own page.

If you are trying to time the cycle, the 30-day rate-of-change pane carries more information than the level. Steep negative readings have bracketed the 2018, 2021, and 2024 capitulation windows; sustained acceleration above 5% per 30 days has tracked every recovery into a fresh ATH. But pair the rate of change with the Hash Ribbon's regime label and with Puell's miner-revenue lens before acting on it — and always ask whether a drop was economic (margins) or mechanical (relocation, curtailment) before you read it as a signal.