What Is Bitcoin Hashrate?

Bitcoin hashrate is the total amount of computing power that miners point at the Bitcoin network, measured as the number of hash calculations the network performs every second. As of September 7, 2026, it sits at roughly 900 exahashes per second, which works out to about 78 septillion guesses per day. Written out, that is 78 followed by 24 zeros.

That number cannot be measured. There is no meter attached to the network. Every bitcoin hashrate figure you have ever read, including the one above, is an inference drawn from how quickly blocks are being found. Understanding that single fact explains most of the confusion around the metric, including why two respected trackers can disagree by 50 exahashes per second on the same afternoon.

This guide covers what hashrate measures, how it is estimated, how it interacts with mining difficulty, what it does and does not tell you about Bitcoin's security, and why 2026 has been the strangest year in the metric's history.

Manage your Bitcoin and other cryptocurrencies securely with the self-custody Bitcoin.com Wallet app.

Key Takeaways

• Bitcoin hashrate measures guesses per second, not transaction throughput or network speed. As of September 7, 2026, the network runs at roughly 900 exahashes per second (EH/s), where one exahash equals one quintillion hashes.
• Hashrate is estimated, never observed. It is derived from mining difficulty and observed block times, which is why published figures differ between sources and why the seven-day average is the industry default.
• A higher hashrate does not make transactions confirm faster. Difficulty retargets every 2,016 blocks to hold block times near ten minutes regardless of how much power joins or leaves.
• 2026 broke the historic pattern. Hashrate peaked near 1,157 EH/s in October 2025 and has spent more than 300 consecutive days below that record, the longest drawdown in a decade, even as BTC recovered from its lows.
• Miners are selling megawatts to AI instead. More than $70 billion in AI and high-performance computing contracts have been announced across the public mining sector, and AI infrastructure earns roughly ten times more per megawatt than mining does.
• Security scales with cost, not just with the number. Assembling 51% of current hashrate would take well over a million machines and several gigawatts of power, before an attacker spent a cent on electricity.

What Is Bitcoin Hashrate?

Bitcoin hashrate is the combined rate at which all miners on the network perform SHA-256 hash calculations, expressed in hashes per second (H/s). It is a measure of computational effort, and by extension of the money, hardware and electricity currently committed to producing Bitcoin blocks.

A hash is the output of a one-way mathematical function. Feed SHA-256 any data and it returns a fixed-length string of 64 hexadecimal characters. Change one character of the input and the output changes completely, in a way nobody can predict in advance. That unpredictability is the whole point. Because there is no shortcut from a desired output back to an input, the only way to find a hash below a specific numeric threshold is to keep trying different inputs and check each result. That is mining, and it is described in the original Bitcoin whitepaper as proof of work.

So bitcoin hashrate is a guess rate. It counts how many attempts per second the entire network makes at a problem that has no method other than brute force.

The numbers involved outgrew normal units years ago, which is why mining discussions are full of prefixes. Each step up the scale is a factor of one thousand.

Hashrate units and what each represents

The scale of growth is easier to feel as a ratio. In January 2009, the network was doing roughly 7.2 million hashes per second. Today it does roughly 900 quintillion. That is an increase of about 126 trillion times in seventeen years.

How Bitcoin Mining Turns Electricity Into Hashes

Bitcoin mining hashrate is produced by a loop that runs constantly on every machine connected to the network. The loop has four steps.

1. Assemble a candidate block: A miner gathers pending transactions from the mempool, adds a coinbase transaction paying itself the block subsidy plus fees, and builds an 80-byte block header. The header contains the previous block's hash, a summary of the transactions, a timestamp, the current difficulty target and a field called the nonce.
2. Hash the header: The miner runs SHA-256 over those 80 bytes and gets a 256-bit number back.
3. Compare the result against the target: If the resulting number is lower than the network's current target, the miner has found a valid block. It broadcasts the block, other nodes verify it in milliseconds, and the miner collects the reward.
4. Change the nonce and repeat: Almost always, the hash is too high. The miner increments the nonce, producing entirely different input data, and hashes again.

Step four is where hashrate lives. A machine advertising 200 TH/s is completing that comparison 200 trillion times per second. Across the whole network, the expected number of attempts needed to find one block right now is about 540 sextillion, or 5.4 followed by 23 zeros. On average that takes ten minutes, though individual blocks arrive anywhere from seconds to over an hour apart because each attempt is statistically independent of the last.

The economics are worth being blunt about. Every one of those 78 septillion daily attempts except the winning few consumes electricity and produces nothing of direct value. That apparent waste is the mechanism. The cost of the failed guesses is what makes rewriting Bitcoin's history expensive, and expense is the security model.

How Bitcoin Hashrate Is Measured, and Why Trackers Disagree

There is no sensor counting hashes. Miners do not report their attempts to anyone, and there would be no way to verify the reports if they did. What the network does publish, in every block, is two hard facts: the current difficulty and the timestamp. Hashrate is reverse-engineered from those.

The estimate works like this. At a given difficulty, the expected number of hashes required to find one block is the difficulty multiplied by 2^32, which is 4,294,967,296. If blocks are arriving every ten minutes on average, then dividing that expected workload by 600 seconds gives an implied hashrate:

Hashrate ≈ (difficulty × 2^32) ÷ average block time in seconds

Run the current numbers. Difficulty stood at 125.81 trillion following the adjustment at block 963,648 on August 22, 2026. Multiply by 2^32 and divide by 600, and you get roughly 900.6 EH/s. That is where the headline figure comes from, and you can reproduce it in a calculator in about fifteen seconds.

Now the complication. Block times are random. On any given day the network might find blocks faster or slower than ten minutes purely by luck, with no change in the underlying hardware. Feed a single lucky day into the formula and it will report a hashrate "record" that never existed. Feed in an unlucky day and it will report a collapse that never happened.

Different providers handle this differently, which is the whole reason published figures diverge.

Hashrate estimation methods and their effects

The practical consequences are real. In mid-August 2026, one widely used chart showed 872 EH/s while another showed 915 EH/s, and neither was wrong. They were answering slightly different questions. The same thing happens with records. Bitcoin's all-time high has been variously reported as about 1,157 EH/s in October 2025 on a seven-day basis, and as 1.44 ZH/s in September 2025 under a more sensitive methodology. This is not new either. In December 2021, three popular trackers simultaneously put the network's peak at 180, 186 and 197 EH/s.

The takeaway for readers is simple. When you see a bitcoin hashrate figure, check the averaging window before comparing it to anything. A seven-day average is the safest default, and comparing a seven-day figure to someone else's instantaneous reading is meaningless. You can see the difficulty, block times and pool data the estimates are built from on the mempool.space mining dashboard.

Hashrate and Mining Difficulty

Hashrate and mining difficulty are two halves of a feedback loop, and confusing them is the most common error in hashrate explanations.

Hashrate is the input: how much power miners choose to supply. Difficulty is the protocol's response: how hard it makes the target so that blocks keep arriving roughly every ten minutes no matter how much power shows up.

Every 2,016 blocks, which is about two weeks, every full node independently recalculates difficulty. Each node compares how long the previous 2,016 blocks actually took against the 20,160 minutes that a ten-minute average implies, then scales difficulty by that ratio. Faster than schedule means difficulty rises. Slower means it falls. There is no vote, no committee and no announcement. The rule sits in the Bitcoin Core source code and every node arrives at the same answer from the same block timestamps. Consensus rules cap any single retarget at a fourfold increase or a 75% decrease, which prevents a single freak epoch from breaking the schedule.

This is why the honest answer to "what happens when bitcoin hashrate drops" is: blocks get slower for up to two weeks, and then the network fixes itself.

Early 2026 provided an unusually clean demonstration. In late January, Winter Storm Fern brought extreme cold to Texas and other major United States mining regions. Miners with demand-response agreements powered down rigs, some voluntarily to sell power back to the grid at premium rates and some to avoid straining it. Foundry USA, the largest mining pool, lost roughly 60% of its hashrate, close to 200 EH/s. Network hashrate fell from around 1.13 ZH/s to as low as 663 EH/s over January 25 and 26, a seven-month low. Block times stretched toward twelve minutes.

Nothing broke. Transactions kept confirming, fees stayed normal, and BTC's price traded through the episode without reacting. Two weeks later the arithmetic caught up and difficulty fell 11.16% on February 7, 2026. Because easier difficulty means better margins, and because Texas miners came back online faster than the cold receded, the following epoch ran ahead of schedule and difficulty jumped 14.73% on February 19. Two of the most volatile consecutive epochs on record, and the ten-minute average held.

Notable Bitcoin difficulty adjustments in 2026. All figures as of September 7, 2026. Difficulty retargets roughly every two weeks.

The pattern across 2026 tells the story better than any single number. Ten of the seventeen adjustments moved lower, and difficulty finished August roughly 15% below where it started the year. Every attempted recovery since June has been erased within an epoch or two.

Why Hashrate Matters for Bitcoin's Security

Hashrate is Bitcoin's security budget expressed in computation. Rewriting confirmed history requires out-mining everyone else, which means an attacker needs to control more than half of the network's hashrate. That is the 51% attack, and hashrate is the direct measure of what it would cost.

The arithmetic is worth walking through, because the abstract version undersells it. Charles Schwab published a detailed breakdown of Bitcoin attack economics in mid-2026 using much the same approach. At roughly 900 EH/s, a majority share means assembling about 459 EH/s of your own. Using 200 TH/s machines, that is around 2.29 million units. Using top-tier 300 TH/s machines, about 1.53 million. At current retail prices for high-end hardware, the purchase order alone runs into the billions of dollars, and that is before you find somewhere to plug it in: an attack fleet that size would draw close to 6 gigawatts, comparable to the output of several large power stations, dedicated to nothing but the attack.

There is also a manufacturing constraint that pure dollar figures miss. Roughly 1.5 to 2.3 million machines is a meaningful fraction of global annual ASIC production, so an attacker could not simply place an order and wait. Buying that much capacity would be visible to the market long before it came online.

It is equally important to be precise about what a majority attacker could actually do, because this is routinely overstated. With more than half the hashrate, an attacker could reorganize recent blocks, censor transactions from being confirmed, and double-spend their own coins by rewriting the chain after a payment was accepted. They could not spend coins from wallets they do not hold keys for, create bitcoin out of thin air, or change the 21 million supply cap. Those limits come from cryptographic signatures and consensus rules that hashrate cannot override.

The genuine concern is not the raw total. It is who directs it. Individual miners overwhelmingly point their machines at pools rather than mining solo, and pool operators have historically decided which transactions go into blocks.

Bitcoin mining pool distribution. Seven-day averages from mempool.space, week ending August 31, 2026. Pool shares are estimated from blocks mined and shift week to week.

Four pools account for roughly 69% of the network. That figure looks alarming until you separate two different things. Pools do not own the hashrate; they coordinate it. Miners can and do repoint their machines within hours, which is a form of accountability that raw share numbers hide. The sharper risk was that pool operators controlled block construction, and that is being addressed: during 2026, pools representing close to 75% of global hashrate committed to the Stratum V2 protocol, which hands transaction selection back to individual miners regardless of which pool they mine with.

Concentration also creates a mundane operational risk that has nothing to do with attacks. When one pool's miners cluster in one region, a cold snap in Texas becomes a global network event, which is precisely what Winter Storm Fern demonstrated.

What Bitcoin Hashrate Does Not Tell You

Four claims about hashrate circulate widely, including on pages that currently rank well in search results. All four are wrong.

• It does not make transactions confirm faster - This is the most persistent error. Block times target ten minutes whatever the hashrate, because difficulty adjusts to hold that average. Doubling the network's hashrate overnight would produce two weeks of slightly faster blocks, then a difficulty increase that puts things back at ten minutes. Confirmation speed depends on fee competition for space in blocks, not on hashrate.
• It does not set bitcoin's price - Causation runs mostly the other way, and with a lag. Higher prices raise mining revenue, which funds hardware and power contracts, which shows up as hashrate months later. 2026 has made the point emphatically: hashrate has sat around 20% below its October 2025 peak while BTC recovered from roughly $63,000 in early August to about $79,400 on September 7, 2026. Anyone treating hashrate as a price signal in 2026 would have read it backwards.
• It does not measure decentralization - Hashrate is a single scalar. In principle, 900 EH/s could come from one operator in one building. How that power is distributed across firms, pools, countries and grids is a separate question with separate data, covered in the pool and geography tables above.
• It does not track energy use one-for-one - Machines keep getting more efficient, so hashrate and electricity consumption have gradually decoupled. The network can add hashrate while holding power flat, and it frequently has.

Hashprice, Efficiency and the Shutdown Price

If you want to understand why hashrate moves, watch hashprice. It is the single most useful metric in mining and the one least explained to general readers.

Hashprice is the daily revenue a miner earns per unit of hashing power, usually quoted in dollars per petahash per second per day. It combines four inputs: the BTC price, network difficulty, the block subsidy and transaction fees. When hashprice falls below a machine's operating cost, that machine loses money every hour it runs, and rational operators switch it off. That threshold is the shutdown price, and it is why hashrate is elastic rather than fixed.

As of August 31, 2026, hashprice stood at $39.36 per petahash per second per day. For context on how brutal the last year has been, it peaked around $63 in July 2025, fell to roughly $35 to $37 by November 2025, and bottomed near $28 to $29 in the first half of 2026, an all-time post-halving low. CoinShares documented that collapse in detail in its Q1 2026 Bitcoin mining report, estimating that 15 to 20% of the global fleet was losing money at $30 hashprice.

Two structural pressures sit underneath those numbers.

The first is the halving. April 2024 cut the block subsidy from 6.25 to 3.125 BTC. The next halving is expected in 2028 at block 1,050,000, roughly 85,000 blocks away as of early September 2026. Each halving cuts miner revenue per block in half overnight while costs continue.

The second is the fee market, which has not filled the gap. Transaction fees made up just 0.64% of total miner revenue in the week ending August 31, 2026, around 20 BTC against roughly 3,214 BTC in total block rewards. For a network that will eventually rely on fees alone, that share is the number worth watching.

Against that, miners have one reliable lever: efficiency, measured in joules per terahash. Lower is better.

ASIC mining efficiency by era. Efficiency figures reflect best commercially available hardware, not fleet averages. As of April 2026, leading fleets operated at roughly 15 to 16 J/TH, while older fleets remained above 24 J/TH.

That roughly tenfold efficiency gain since 2016 is what has allowed hashrate to grow far faster than energy use, and it is also what forces the refresh cycle. A fleet running at 25 J/TH competing against one at 13 J/TH pays nearly twice the electricity bill for the same revenue, which is why hashprice troughs push older machines off the network permanently rather than temporarily.

Bitcoin's Hashrate History

Because difficulty is recorded on-chain and hashrate is not, the cleanest way to see the long arc is to convert historical difficulty into implied hashrate using the formula above.

Bitcoin difficulty and implied hashrate by era. Implied hashrate calculated by Bitcoin.com as difficulty multiplied by 2^32, divided by 600 seconds. Actual hashrate at each point differed depending on whether blocks were running fast or slow.

The step changes in that table map onto hardware generations. CPUs gave way to GPUs around 2010, which were roughly a hundred times faster per machine. Field-programmable gate arrays had a brief run in 2011. Purpose-built ASICs arrived in 2013 and made everything before them obsolete within months; difficulty rose in 30 of that year's 31 adjustments.

Two disruptions stand out. In mid-2021, China banned mining outright. Roughly half the network went offline within weeks, hashrate fell from around 180 EH/s in May to under 90 EH/s by July, and the protocol responded with the largest downward difficulty adjustment in its history at nearly 28%. Machines were crated up and shipped to the United States, Kazakhstan, Russia and Canada. The network recovered its previous high within about nine months, and the United States emerged as the dominant mining jurisdiction.

The second is happening now. The network crossed 1 zettahash per second for the first time in late August 2025 and peaked around 1,157 EH/s that October. It has not been back since. As of September 7, 2026, hashrate had spent 320 consecutive days below its record, the longest such stretch in a decade, with the seven-day average sitting near 914 EH/s at the end of August. What makes this drawdown different from 2021 is that no government banned anything and no storm caused it. Miners chose to redirect capacity elsewhere.

Where Bitcoin's Hashrate Comes From

Hashrate has a physical address, and knowing where it sits explains a lot about how the metric behaves.

Estimated bitcoin hashrate by country. Estimates for Q2 2026 from mining data provider Hashrate Index. The top three jurisdictions account for roughly 65% of the network. Country-level figures are approximations, since miners are not required to disclose location and some operate in jurisdictions where mining is officially prohibited.

Two things in that table are worth pausing on. China still hosts around 12% of the network five years after banning mining, which says something about how enforceable such bans are. And Paraguay, Ethiopia and Oman are new arrivals in the top ten, drawn by stranded hydro and gas power that has no other buyer.

On energy, the most careful public estimate comes from the Cambridge Bitcoin Electricity Consumption Index, which put the network's annualized draw at roughly 138 terawatt-hours as of mid-2026, around 0.5% of global electricity consumption. Cambridge pairs that central figure with a lower bound near 74 TWh and an upper bound near 242 TWh, precisely because the exact mix of hardware on the network cannot be observed. Cross-checking is instructive: 138 TWh per year is an average draw of about 15.8 gigawatts, which across 900 EH/s implies a fleet-wide efficiency near 17.5 J/TH including data-center overhead. That is consistent with a network running a mix of current and previous-generation machines, which is a useful sanity check on both numbers. Cambridge's April 2025 industry survey also put the sustainable share of mining's energy mix at 52.4%, combining 42.6% renewables with 9.8% nuclear.

Which brings us to the defining development of 2026. Bitcoin miners own something that has suddenly become far more valuable than mining: energized data-center capacity. Over $70 billion in AI and high-performance computing contracts have now been announced across the public mining sector. The economics are lopsided. Building mining infrastructure costs roughly $700,000 to $1 million per megawatt. AI infrastructure costs roughly $8 million to $15 million per megawatt, and the revenue scales accordingly. Several listed companies that were pure-play miners in 2024 now describe themselves as data-center operators that also mine, and CoinShares projected that listed miners could derive as much as 70% of revenue from AI by the end of 2026, up from around 30% earlier in the year. Individual deals are visible in public filings on SEC EDGAR, including multi-billion-dollar, multi-year leases running into the 2030s.

This has broken the feedback loop that governed hashrate for a decade. Historically, a recovering BTC price pulled hashrate back within months. A fifteen-year hyperscaler lease cannot be paused because bitcoin rallied. Whether the October 2025 peak gets retaken now depends on whether mining margins can outbid AI tenants for the same megawatt, which is a genuinely open question rather than a matter of waiting for price.

None of this threatens the network in absolute terms. 900 EH/s remains an extraordinary amount of committed computation, and difficulty adjusts downward to protect the margins of whoever stays.

How to Check Bitcoin's Hashrate Yourself

You do not need to trust any single figure, including ours. The underlying data is public.

1. Start with the difficulty, not the hashrate: Difficulty is recorded in every block and is not an estimate. Any block explorer will show the current value and the countdown to the next retarget.
2. Check the seven-day average, not the daily reading: Almost every provider offers both. The daily figure is noise; the seven-day figure is signal.
3. Do the arithmetic if you want to verify: Multiply difficulty by 4,294,967,296 and divide by 600. If a published hashrate is wildly different from your answer, blocks are currently running fast or slow, which is itself informative.
4. Compare at least two sources: If they agree closely, the estimate is stable. If they diverge sharply, block times are volatile and any single reading should be treated with caution.
5. Look at average block time as a cross-check: Blocks consistently over ten minutes means hashrate has left and difficulty has not caught up yet. Consistently under ten minutes means the opposite.
6. Check pool distribution while you are there: It tells you more about network health than the headline number does.

Conclusion

Bitcoin hashrate measures the computing power miners commit to the network, in hashes per second, and it functions as a running tally of the cost of attacking Bitcoin's history. As of September 7, 2026, it stands near 900 EH/s, roughly 20% below the record set in October 2025 and holding there longer than after any previous peak, not because of a ban or a crackdown but because miners are finding better uses for their megawatts.

Treat the metric with the precision it deserves. It is an estimate, not a measurement. It says nothing about transaction speed and very little about price. What it does say, clearly, is how much real-world capital and energy currently stand behind the ledger.