Bitcoin Hashrate Growth in 2026: The Climb to 2 Zettahashes

For fifteen years, one line on every Bitcoin mining chart behaved like a law of nature. Network hashrate, the total amount of computing power aimed at guessing valid blocks, went up. It rose through bear markets, through exchange failures, through a 2021 Chinese mining ban that knocked more than half the network offline in a single summer. Every time it fell, it came back higher. By the autumn of 2025 the line had pushed briefly above one zettahash per second, a figure with twenty-one zeros in front of it, and the cost of breaking that much computing power had never looked steeper.

Then 2026 broke the pattern. Hashrate posted its first quarterly decline in six years, and it did so for a reason Bitcoin had never faced at scale: not a ban, not a price collapse, but a better-paying tenant for the same electricity. Miners started leasing their power and their buildings to artificial intelligence. The question hanging over the industry for the rest of this year is no longer whether hashrate can grow. It is whether it will, and at what price of Bitcoin the climb toward two zettahashes restarts.

The curve that only pointed up, until 2026

Hashrate growth is the closest thing Bitcoin has to a heartbeat. It is the outward sign that people are pointing more machines, more chips and more electricity at the network, and for most of Bitcoin’s life the direction was so reliably upward that analysts treated a rising hashrate as a given. The falls that did happen were violent and short. The China ban of mid-2021 was the sharpest: Beijing outlawed domestic mining, roughly half the network went dark within weeks, and difficulty recorded its largest single downward adjustment on record. Within a year the machines had relocated, mostly to the United States, and the hashrate was making new highs.

2026 is different in kind, not just in degree. The network’s mining power has slipped roughly 17% below its record, and difficulty, the metric that tracks hashrate with a lag, fell below its year-earlier level for only the second time in Bitcoin’s history. The first time was the China ban. This time nobody banned anything. Miners with working, plugged-in machines are choosing to run fewer of them for Bitcoin and redirect the electricity elsewhere. That is a first, and it changes how you have to read the whole growth story.

What hashrate growth actually measures

A hash is one attempt at solving the puzzle that seals a Bitcoin block: run the block’s data through the SHA-256 function, check whether the result falls below a target number, and if it does not, change one field and try again. Hashrate is how many of those attempts the whole network makes each second. Mining is guessing, at planetary scale, and hashrate is the guess rate.

One subtlety trips up almost every newcomer: nobody can observe hashrate directly. There is no meter on the network. Instead it is inferred from how quickly blocks are arriving relative to the current difficulty. The working estimate is hashrate is approximately difficulty multiplied by two to the power of 32, divided by the 600-second block target. Plug in a difficulty of about 125.8 trillion and you get a figure near 900 exahashes per second, which is why every tracker publishes a slightly different number and why any honest snapshot is a range, not a single point. The gap between a noisy single-day reading and a smoothed seven-day or thirty-day average is wide enough that some data providers have at times flashed record highs the network never actually reached, so seasoned analysts track the moving averages rather than the spot figure. Growth, then, means the estimated guessing capacity of the network expanding over time, measured across the ladder of units below.

The jump from the kilohashes of Satoshi’s laptop to the near-zettahash of 2025 is roughly eighteen orders of magnitude in sixteen years. No consumer technology has scaled like it. That history is what makes the 2026 flatline notable: it is the first time the ladder has stopped climbing without a gun to its head.

The 2026 stall, and why no government ordered it

The table below is the network as it stands in late August 2026, after a long squeeze and a sharp, late-summer relief rally in the Bitcoin price. Read it as a freeze-frame of a growth engine idling rather than accelerating.

Difficulty sits at 125.81 trillion after the retarget around 22 August, according to the CoinWarz difficulty chart, down from a November record of 155.97 trillion. That is a decline of roughly 19% from the top. The current level is also below where difficulty stood a year earlier, which, as CoinDesk noted, has happened only once before in Bitcoin’s history. Hashrate Index has attributed the 2026 contraction to weak mining economics, the migration of power and capital toward AI and high-performance computing, regional outages, curtailment in Texas and disruption tied to Iran. Every item on that list is a business decision or a local shock, not a coordinated crackdown. The network is shrinking its computing footprint because, at recent prices, a meaningful slice of it did not pay.

Engine one: the price, and a beta that keeps rising

The single biggest driver of hashrate growth has always been the Bitcoin price, and the relationship runs one way. Price leads, hashrate follows. When Bitcoin rallies, mining revenue per machine rises, marginal rigs turn profitable, operators plug in idle inventory and order new fleets, and months later the extra machines show up as higher hashrate. When price falls, the sequence reverses. Hashrate is a lagging, price-sensitive indicator dressed up as an engineering metric.

What is new in 2026 is how tightly the two are now coupled. Analysts at JPMorgan led by Nikolaos Panigirtzoglou put a number on it: the beta of mining difficulty to the Bitcoin price has risen to 0.62 over the past six months, meaning difficulty now moves far more readily with price than it used to. Their explanation is blunt: when Bitcoin trades below its production cost, higher-cost miners power down, the hashrate declines, and difficulty adjusts lower. The bank pegs all-in production cost near $78,000 and, for much of the year, Bitcoin traded under it. That is the mechanism behind the stall. With so many operators sitting right on their cost floor, small price moves flip large blocks of capacity on and off.

It also explains why the late-August rally matters so much. Bitcoin has climbed back to around $77,200, above the JPMorgan cost line for the first time in months, which lifts revenue for the marginal machine. That same rally, tracked in our coverage of the hashprice relief rally and Bitcoin’s security bill, pushed the revenue-per-unit figure up sharply. At roughly $77,200 and a network near 900 to 920 EH/s, a first-principles estimate (about 144 blocks a day times 3.125 BTC times the price, divided by the network’s petahashes) puts hashprice close to $38 per petahash per day, up from the $31.89 Hashrate Index last posted on 17 August when Bitcoin sat near $64,000. Whether that translates into fresh hashrate growth depends on the price holding, not spiking for a week.

Engine two: ASIC efficiency, more hashes per watt

The second engine is quieter and more relentless: the machines keep getting more efficient. Efficiency in mining is measured in joules per terahash, the energy a rig burns to make a trillion guesses. Lower is better. And because a mining site is usually capped by the megawatts it can draw, efficiency is what lets hashrate grow even when the power budget is fixed. Swap old boxes for new ones with the same wall draw and the site’s hashrate climbs by itself.

The current flagship, Bitmain’s S21 XP, is rated at about 13.5 joules per terahash, roughly seven times more efficient than the S9 that defined the last cycle. That gain is why the network can, in principle, keep growing its hashrate while its electricity use grows far more slowly. It is also why falling difficulty does not automatically mean fewer machines: a fleet of new S21 XPs can replace two or three times its number of S19s at the same power, quietly pushing raw hashrate up while the megawatt total stays flat. In 2026, though, this engine is running into the other two. Efficient hardware only grows hashrate if someone buys and racks it, and capital for new rigs is scarce. The big fleet-upgrade wave, when operators swapped S19-class boxes for S21-class machines, is largely spent, and the next efficiency leap will need both cheaper chips and the balance-sheet room to deploy them at scale. A more efficient miner sitting in a warehouse adds nothing to the network; only one drawing power and hashing does, and in a year when balance sheets point at AI, a good deal of that efficiency is waiting on the shelf.

Engine three: the megawatt, now contested by AI

Here is the part of the story that has no precedent. For Bitcoin’s entire history, the binding constraint on hashrate growth was access to cheap electricity, and mining was the highest-value use of a stranded or off-peak megawatt that anyone had found. That is no longer true. Training and serving large AI models needs enormous, steady power and the physical shells (substations, cooling, land, grid interconnects) that Bitcoin miners spent a decade building. The AI buildout will pay far more for a megawatt than mining can, and it signs multi-year contracts that a volatile hashprice cannot match.

The scale of the pivot is hard to overstate. Publicly listed miners have announced more than $70 billion in AI and HPC contracts, and CoinShares expects some firms to earn up to 70% of their revenue from AI by the end of 2026, against roughly 30% today. Hut 8’s AI portfolio alone has been valued around $26.6 billion. Every megawatt that flows to a frontier-model tenant is a megawatt not making SHA-256 guesses, which is why the same demand that powers the boom in verifiable AI compute and on-chain model markets is now the ceiling on Bitcoin’s hashrate. For the first time, hashrate growth has to compete for its raw input against a buyer with deeper pockets.

This is not a total loss for the network. A site that hosts AI can keep a mining hall as a flexible, interruptible load, spinning it up when power is cheap and idling it when the grid or a tenant needs the electricity. The arithmetic is stark at the level of a single megawatt: at the current hashprice an efficient fleet earns roughly $107 from a megawatt-hour of power, and an older machine closer to $41, while a signed AI lease can fix a higher rate for years without the price risk of a block reward. But it does mean the days of miners expanding hashrate simply because power was available are over. Power now has a rival bidder, and that bidder usually wins.

Engine four: capital, redirected

Hashrate grows when someone spends money on machines and the buildings to house them. In 2026 the people who used to spend that money on rigs are spending it on graphics processors and data-center fit-outs instead. The reason is survival. In late March, publicly listed miners were losing roughly $19,000 on every bitcoin they produced, and a business that loses money on its core product does not order more capacity for that product. It looks for a new one.

So the capital cycle that historically fed hashrate growth has been rerouted. Miners have sold Bitcoin from their treasuries to fund the transition, taken on non-recourse financing against signed AI leases, and pointed their build pipelines at compute tenants rather than fresh mining halls. The effect on hashrate is straightforward: fewer new machines, older fleets left to age out, and a network that stops compounding upward. It is the mirror image of a bull market, when mining profits are plowed straight back into more mining. The engine is not broken; it has simply been aimed at a different target, and it will not swing back toward hashrate until mining once again offers a better return on a dollar of capital than an AI lease does.

The difficulty thermostat: growth’s built-in brake

Bitcoin has a self-correcting mechanism that keeps blocks arriving about every ten minutes no matter how much or how little hashrate is online. Every 2,016 blocks, roughly every two weeks, the network measures how long that batch actually took and adjusts difficulty to bring the pace back to target. More hashrate makes blocks come faster, so difficulty rises. Less hashrate slows them, so difficulty falls. This thermostat is why a network that has lost 17% of its computing power still produces blocks on schedule, and it is the clearest read on which way hashrate is actually moving.

The 2026 record, drawn from the CoinWarz difficulty chart, is a sawtooth that trends gently down. June brought the year’s largest single drop, more than 10%, and the most recent retarget shaved off another 1.3%, with the next one, due around 6 September, projected slightly negative again. A falling difficulty is the thermostat doing exactly what it is meant to: rebalancing the puzzle for a smaller field of miners so the survivors are not punished for the departures. It is worth stressing that difficulty is a consequence of hashrate, not a cause. It cannot make hashrate grow. It can only keep the network functioning while hashrate decides, on economic grounds, whether to expand or retreat.

The forecast: 1.8 to 2 zettahashes, on one condition

Despite the stall, the most-cited industry forecast still points up. CoinShares’ 2026 mining report projects the network reaching 1.8 zettahashes by the end of this year and 2 zettahashes by the end of March 2027, the latter nudged one month later than a prior estimate. But the forecast comes with a condition stapled to it, and the condition is the whole story: it assumes Bitcoin recovers toward $100,000. The report is explicit that the path splits by price.

If prices stay below $80,000 for the rest of the year, CoinShares expects hashprice to keep sliding as more rigs go dark, which would pull hashrate down rather than up. A sustained drop under $70,000 could trigger a larger capitulation. Only a move toward $100,000 restores the economics that fund growth. James Butterfill, the firm’s head of research, has called this one of the most challenging periods for miners since the last halving, with the weighted-average cash cost to produce a bitcoin around $79,995 in the fourth quarter of 2025. In other words, the climb to two zettahashes is no longer an engineering trajectory the network follows automatically. It is a bet on the Bitcoin price, and the market gets a vote.

There is an arithmetic wrinkle worth naming. The roughly $38 hashprice implied by today’s $77,200 price is already near the $37 CoinShares models for a $100,000 Bitcoin. That is not a contradiction. Hashprice is revenue divided by hashrate, and the $100,000 case assumes the network has grown toward 1.2 zettahashes by then, spreading the same daily reward across more machines. Today’s smaller network earns a similar figure at a lower price. Growth, put plainly, dilutes each machine’s paycheck unless the price rises to match it. That is the treadmill every miner runs on.

The halving clock underneath the growth curve

Every price scenario above sits on top of a countdown that never stops. Roughly every four years the block subsidy, the brand-new bitcoin paid to miners for each block, is cut in half. It dropped to 3.125 BTC in April 2024 and is scheduled to fall to about 1.5625 BTC in 2028. Because the subsidy still accounts for more than 99% of mining revenue, with transaction fees running well under 1% of the block reward, a halving at a flat Bitcoin price cuts miner income almost in half overnight. The network has absorbed three of these before, but each one lifts the price a miner needs just to keep the same machines switched on.

That is why the climb to two zettahashes has a deadline stapled to its price tag. To reach 1.8 zettahashes at the end of 2026 and hold that level through the 2028 halving, Bitcoin would have to rise enough to offset both the extra machines diluting each unit of revenue and the subsidy cut waiting at the end of the decade. Fees could in theory take up the slack, and brief spikes during past demand surges have hinted at what a fee-funded security budget might look like, but those spikes have so far faded within days. Until fees become a durable share of revenue, hashrate growth stays hostage to two clocks at once: how high the price climbs, and how soon the next halving lands. The AI pivot simply adds a third bidder for the same electricity while both clocks keep ticking.

Why hashrate growth still matters: the security budget

Hashrate is not vanity. It is the price of attacking Bitcoin. To rewrite recent history or double-spend, an attacker needs to out-compute the honest network, which means acquiring a majority of its hashrate, and every exahash added to the total raises that bar. A 17% drop in hashrate is, mechanically, a 17% discount on a majority attack. That is why the growth curve is a security curve in disguise.

How high is the bar today? Campbell Harvey, a finance professor at Duke, has modeled a one-week majority attack at around $8 billion, or, as he frames it, about 50 basis points of the value of bitcoin, once you assume the attacker also shorts Bitcoin in offshore derivatives markets to profit from the crash the attack would cause. Not everyone buys the threat. Matt Prusak, president of American Bitcoin Corp, has argued to reporters that economic feasibility kills the 51% thesis, since accumulating that much hardware would take years and exchanges would freeze suspicious activity first. Both can be true: the attack is wildly expensive and getting cheaper at the margin as hashrate slips. This is the security-budget lens that ties mining to the other half of the mining-staking world, where restaking and its great unwinding are wrestling with the same question from the proof-of-stake side: how much economic weight actually stands behind the guarantee, and what happens when that weight shrinks.

Where the next zettahash would come from

If hashrate does resume its climb, geography will shape where the machines land. The United States remains the center of gravity, hosting roughly 37% of global hashrate according to Hashrate Index’s Q2 2026 heatmap, followed by Russia near 17% and China around 12%, with the top three jurisdictions accounting for roughly two-thirds of the network. The growth at the edges is the interesting part: Paraguay has climbed toward 4% on cheap hydro, and Gulf states, Ethiopia and other frontier hosts have added capacity where power is abundant and industrial policy is friendly.

The AI pivot is quietly redrawing this map too. Sites that can host both mining and high-performance computing (stable grids, reliable cooling, room to expand) are winning capital, which favors the United States, Canada and the Gulf over frontier plays chasing the very cheapest kilowatt. A megawatt in Texas that can flip between mining and an AI tenant is worth more than a megawatt in a jurisdiction that can only mine. So the next zettahash, if it arrives, is likelier to be built in places that offer optionality than in places that only offer cheap power. That is a subtle shift in the growth engine’s center of gravity, from lowest-cost electricity toward most-flexible infrastructure.

What it means for miners and holders

For miners, 2026 is a filter. Operators running efficient fleets at genuinely cheap power, or sitting on sites they can convert to AI, are surviving and in some cases thriving. Operators running old machines at merchant power rates are the ones switching off, and their departure is most of the hashrate decline. The industry is not dying; it is consolidating toward efficiency and optionality, and the CoinShares figure that 15% to 20% of the global fleet runs at a loss at recent prices is a rough measure of how much capacity sits on the bubble.

For holders, hashrate remains a useful confidence signal, but it must be read correctly. A rising hashrate has historically meant miners are betting real capital on Bitcoin’s future, and a falling one has meant stress. In 2026 the fall means something more ambiguous: not that miners have lost faith, but that they have found a second business that pays better right now. The transaction-fee side of the security budget, which spikes when demand for block space surges as it did during the Ordinals boom and its flight to Bitcoin’s blue chips, remains too small and too erratic to fill the gap; fees have been running well under 1% of the block reward. The honest read is that hashrate growth has become a leveraged bet on the Bitcoin price, and a rising price is the only thing that reliably turns the engine back on.

For the network as a whole, none of this is existential. Bitcoin kept producing blocks every ten minutes even when more than half its hashrate vanished in a single summer, because the difficulty thermostat guarantees the chain keeps moving no matter how many machines unplug. What 2026 changes is the shape of the growth story, not its survival. The question has shifted from how fast hashrate can compound to whether it compounds at all in a year when the most profitable thing a miner can do with a megawatt is rent it to a company training a model. The answer, on nearly every analyst’s spreadsheet, reduces to the same variable: the price of Bitcoin between now and year-end.

The regulatory line: mining is not a security

One risk that has receded is the legal one, at least in the United States. In March 2025 the SEC’s Division of Corporation Finance issued a staff statement that proof-of-work mining, whether solo or through a pool, does not amount to the offer or sale of a security, on the reasoning that miners earn rewards through their own computational effort rather than the efforts of others. That removed a cloud that had hung over the industry and helped cement the United States as the network’s leading host. Mining itself is not a licensed financial activity, which is why the regulators most relevant to a miner are usually energy and environmental agencies, not securities watchdogs.

The broader US posture toward crypto has warmed in parallel, from a strategic Bitcoin reserve to the stablecoin framework we cover in the GENIUS rulebook reaching offshore. For hashrate growth, the policy signal is favorable but secondary. Clear rules make it easier to finance mining and to list miners as public companies, yet they do nothing to change the core arithmetic. As long as an AI tenant pays more for a megawatt than a block reward does, friendly regulation will not, by itself, restart the climb. Only price will.

Frequently Asked Questions

What is Bitcoin hashrate growth?

Hashrate is the total number of guesses per second the Bitcoin network makes trying to solve blocks, and hashrate growth is that figure rising over time as more machines, chips and electricity come online. It is measured in exahashes per second (EH/s), with the whole network running near 900 EH/s in August 2026, and it is the main real-time gauge of how much industrial effort is securing the chain.

Why did Bitcoin hashrate fall in 2026?

Because miners chose to redirect power and capital to artificial intelligence, not because of a ban or a crash. With Bitcoin trading below JPMorgan’s estimated $78,000 production cost for much of the year, higher-cost rigs powered down, and listed miners signed more than $70 billion in AI and HPC contracts that pay more per megawatt than mining does. It is the first time hashrate has fallen year over year for a reason other than a government crackdown.

Will Bitcoin hashrate reach 2 zettahashes?

CoinShares forecasts 1.8 zettahashes per second by the end of 2026 and 2 zettahashes by the end of March 2027, but only if Bitcoin recovers toward $100,000. If the price stays below $80,000, the firm expects hashrate to decline further as more miners exit, and a sustained move under $70,000 could trigger a larger capitulation. The milestone is now conditional on price rather than a near-automatic trajectory.

Does falling hashrate make Bitcoin less secure?

At the margin, yes, but not dangerously so at current levels. A lower hashrate means a majority attack costs less, and a 17% drop is roughly a 17% discount on that attack. Even so, researchers such as Duke’s Campbell Harvey still estimate a one-week majority attack at around $8 billion, and the network has never suffered a successful 51% attack. Security is lower than at the 2025 peak but remains extremely expensive to breach.

How is Bitcoin hashrate measured if it cannot be observed directly?

It is estimated, not metered. The network infers hashrate from how fast blocks are arriving relative to the current difficulty, using the relationship that hashrate is approximately difficulty times two to the power of 32, divided by the 600-second block target. Because this depends on recent block timing, which is random in the short run, different trackers report slightly different numbers, so a network hashrate is best read as a range rather than a single precise value.

By Marcus Okafor, mining and network security correspondent at HOGE Wire.