Hashrate Growth and the Price of Security: Bitcoin vs. Ethereum

Bitcoin’s network hashrate sits close to 860 exahashes per second in early August 2026, off its late 2025 peak but still representing an enormous, continuously reinvested pool of computing power. Watching that number climb or fall has become a habit for anyone following Bitcoin mining, but the number by itself does not explain much. Hashrate growth is really a story about capital: who is spending money to secure a network, why they keep spending more of it even as profit margins shrink, and what that spending actually buys. Ethereum’s validators are running a parallel experiment on the same question, growing a very different kind of security budget through staked ETH rather than electricity. This piece walks through both: what drives Bitcoin’s hashrate higher, and occasionally lower, what a would-be attacker would actually have to spend to overwhelm it, and how that compares to the economics of growing security through proof of stake.

What Hashrate Growth Actually Measures

Hashrate is the number of guesses Bitcoin’s mining machines make every second while trying to find a valid block. Each guess is a SHA-256 hash of the candidate block header combined with a changing number called a nonce; the network is not solving anything meaningful in a mathematical sense, it is brute-forcing a number below a target threshold, over and over, as fast as possible. The unit scales the same way storage does, from small to large:

• Kilohash: a thousand hashes per second
• Megahash: a million hashes per second
• Gigahash: a billion hashes per second
• Terahash: a trillion hashes per second, roughly what a single ASIC produces
• Petahash: a thousand terahashes per second
• Exahash: a quintillion hashes per second, the unit Bitcoin’s total network hashrate has used for most of the past decade
• Zettahash: a thousand exahashes per second, a threshold Bitcoin first crossed in 2025

When people say hashrate is growing, they mean the combined guessing speed of every machine connected to the network keeps climbing.

That growth is not one smooth line, and headlines that treat it as a single bullish or bearish signal usually miss what is actually happening underneath. Growth compounds two separate things: more machines being plugged in, and each new generation of machine guessing faster per watt than the last. Growth in the first driver is a bet on Bitcoin’s future price and fee revenue; growth in the second is closer to an industrial manufacturing trend, similar to how each generation of GPU or CPU packs more transistors onto the same size chip. Both feed the same headline number, but they mean very different things for miner economics.

The framing that matters most for this piece is that hashrate is Bitcoin’s security budget made visible. Every exahash of guessing power is machinery, electricity and financing that has to be outspent by anyone who wants to rewrite Bitcoin’s transaction history. Ethereum secures itself differently, through capital that stakers lock up and risk losing rather than electricity that miners burn, but the underlying question is identical: how expensive is it to make the network lie? Comparing the two growth curves side by side is the point of this piece.

Where Bitcoin’s Hashrate Stands Right Now

As of August 5, 2026, Bitcoin’s network hashrate is running at roughly 860 exahashes per second at block height 961,157, according to CoinWarz, down about 5.6% over the previous 24 hours and 3.8% over the past week, though still up close to 9% over the trailing month. That kind of short-term whipsaw is normal; hashrate is not measured directly, it is inferred from how quickly blocks are actually being found relative to the current difficulty target, so any single reading can look noisy even while the underlying trend is clear.

The bigger picture is that hashrate is still coming down from a genuine record. The network first pushed past one zettahash per second during 2025 and peaked at somewhere around 1.1 to 1.16 zettahash that October, according to CoinShares’ Q1 2026 Bitcoin Mining Report, around the same time Bitcoin itself set an all-time price high near $126,000. Current levels sit a bit more than 20% below that peak, a real pullback but far from a collapse, and one that has already partly reversed itself more than once this year as the difficulty adjustment keeps recalibrating around whatever hashrate actually shows up.

The two metrics that matter for context, price and hashprice (miner revenue per unit of hashrate per day), have been telling different stories for most of 2026. Bitcoin’s price has traded in the mid $60,000s for weeks, well below the October 2025 high, while hashprice has sat near multi-year lows through much of the same stretch even as raw hashrate stayed historically elevated. That gap, computing power staying high while the revenue it generates stays depressed, is the single most important thing to understand before reading too much into any hashrate chart.

The Engine Behind the Curve: ASIC Efficiency and Capital Cycles

Two forces push hashrate higher, and only one of them is optional. The first is pure capital deployment: mining companies and private operators buying more machines and plugging them into more megawatts. The second is efficiency. Each new generation of ASIC, application-specific integrated circuit, squeezes more hashrate out of the same watt of electricity, which means total network hashrate can rise even when nobody buys a single new machine, simply because operators are swapping older units for newer ones as they reach the end of their useful economic life.

The efficiency curve has been dramatic. Bitmain’s Antminer S9, the workhorse of the 2016 to 2018 mining boom, ran at about 98 joules per terahash. The S19 generation that dominated the 2020 to 2022 cycle cut that closer to 30 joules per terahash, and later revisions pushed it lower still, according to spec data compiled by D-Central. The current flagship, the Antminer S21 XP, is rated at 13.5 joules per terahash while delivering 270 terahashes per second, per Bitmain’s own specification sheet. That is roughly a sevenfold efficiency gain in under a decade, in an industry that also has to keep the lights on.

Efficiency gains explain why hashrate can keep setting records even in a year when miner profit margins are being squeezed hard. A miner does not need Bitcoin’s price to rise to add hashrate; replacing a rack of older units with current-generation machines adds capacity while cutting the electricity bill for the same output, which is one reason hashrate growth and miner profitability can move in opposite directions for months at a time.

The Difficulty Adjustment: Bitcoin’s Built-In Growth Governor

Bitcoin targets a new block roughly every ten minutes no matter how much hashrate is pointed at the network. It holds that pace through the difficulty adjustment: every 2,016 blocks, about two weeks, the protocol compares how long that batch of blocks actually took against the ten-minute target and moves the difficulty threshold up or down to compensate. More hashrate chasing the same block reward pushes difficulty up at the next adjustment; hashrate leaving the network pulls it back down. It is a pure feedback loop with no human input, written directly into Bitcoin’s consensus code.

2026 has been an unusually volatile year for that mechanism. Current difficulty sits at 126.23 trillion as of block 961,157 on August 5, 2026, with the next adjustment expected around August 8 and forecast to tick up only modestly, according to CoinWarz. That follows a stretch of unusually large swings in both directions over the preceding ten weeks.

Swings of five to ten percent in a single adjustment window used to be rare; they became almost routine in the first half of 2026 as miners reacted in near real time to a hashprice that kept testing multi-year lows, switching machines off during unprofitable stretches and back on when conditions improved. The difficulty adjustment is, in effect, Bitcoin’s thermostat: it does not stop hashrate from falling, it just makes sure blocks keep arriving on schedule regardless of how much or how little computing power actually shows up.

What Growing Hashrate Is Supposed to Buy: Modeling the Cost of Attack

All of that spending exists to answer one question: what would it cost someone to overpower the network and rewrite its history? A 51% attack requires controlling a majority of hashrate for long enough to outrun the honest chain, letting an attacker reverse their own transactions or censor others. Nobody has pulled this off against Bitcoin itself; the arithmetic below is a large part of why.

Duke University finance professor Campbell Harvey put a number on it in 2025: sustaining a majority-hashrate attack for one week would cost roughly $6 billion, split between about $4.6 billion in mining hardware, $1.34 billion in data center construction and around $130 million in electricity, according to reporting on the paper by crypto.news. On Harvey’s numbers, that works out to about a quarter of one percent of Bitcoin’s total network value, which sounds cheap until you remember it is $6 billion that has to be spent on hardware nobody else particularly wants to sell you quickly, followed by a chain reorganization the entire industry would notice within minutes.

Harvey has since refined the model to account for something the original version left out: an attacker does not need to profit from mining at all if they can profit from the attack itself. Pairing a hashrate acquisition with a large short position in Bitcoin derivatives on liquid, largely offshore venues turns a falling price during the chaos of an attack into the payoff, rather than a cost that has to be recouped through block rewards, an update to the thesis that makes the attack marginally more attractive on paper even as it stays wildly expensive in practice.

Not everyone buys the thesis at all. Matt Prusak, president of American Bitcoin Corp, pushed back when Bloomberg asked him about Harvey’s work, arguing that “economic feasibility kills the 51% thesis” since accumulating that much hashpower would take years in practice, not a single transaction, while exchanges and custodians would likely freeze suspicious activity long before an attacker could cash out, according to the same crypto.news report. Both views agree on the underlying point: the theoretical cost of attacking Bitcoin is a specific, calculable number, and it climbs every time the network’s hashrate and price both grow.

The Other Growth Curve: How Ethereum Prices Security Through Staking

Ethereum abandoned mining entirely at the September 2022 Merge, and it grows its security budget through an entirely different mechanism: validators lock up ETH as a bond, and the network can destroy, or slash, part of that bond if they misbehave. There is no electricity bill and no hardware to overpower; the thing an attacker has to acquire is a large enough share of the staked ETH itself.

That pool of staked ETH keeps growing. As of early August 2026, roughly 41.5 million ETH, about 34% of the entire supply, sits staked across roughly 895,060 active validators, according to validatorqueue.com, which tracks the Ethereum beacon chain directly. Demand to join is still strong enough to create a real backlog: the same tracker shows about 2.48 million ETH waiting in the entry queue, with an estimated wait of roughly 43 days before new validators can start earning rewards, Ethereum’s equivalent of Bitcoin miners racing to plug in new machines. Those rewards come from a mix of protocol issuance and transaction priority fees, paid to whichever validator is chosen to propose each block, rather than from a fixed block subsidy like Bitcoin’s.

The validator count itself is behaving strangely relative to that growth, and the reason is a 2025 upgrade. Ethereum’s Pectra upgrade, which activated on mainnet on May 7, 2025, included a change (EIP-7251) that raised the maximum effective balance a single validator can hold from 32 ETH to 2,048 ETH, according to the Ethereum Foundation’s own mainnet announcement. Before that change, a staker with 640 ETH had to run 20 separate validators; afterward, they can run one. Large staking operators have been consolidating accordingly, which means total staked ETH and validator count are no longer telling the same story: stake keeps climbing while the validator count grows more slowly than it otherwise would, as fewer, larger validators do the same job.

Pricing an Attack Under Proof of Stake

Ethereum’s version of the majority-attack math looks completely different from Bitcoin’s, because the resource being acquired is the asset itself rather than external machinery. Broadly, an attacker needs to control just over one third of staked ETH to disrupt finality, stalling the chain from finalizing new blocks, or just over two thirds to actively finalize a dishonest version of history. Both thresholds are published protocol constants, not estimates.

The economics work against the attacker in a way Bitcoin’s model does not. Acquiring that much ETH on the open market, worth many tens of billions of dollars at current prices for the two-thirds threshold, would push the price against the attacker the entire way up. Once in control, the honest majority of the network can coordinate a response Bitcoin has no real equivalent of: a social slashing event, where the rest of the network votes to burn the attacker’s stake entirely, destroying the capital that funded the attack in the first place. Bitcoin has no way to claw back a majority miner’s hardware; Ethereum can, in principle, claw back a majority attacker’s capital.

This is the core structural difference between the two growth curves. Bitcoin’s security is denominated in a sunk, external cost, hardware and electricity that keeps existing whether or not an attack happens. Ethereum’s is denominated in the asset under attack itself, which can be partly destroyed as a direct penalty. Both models get more expensive to attack as they grow, but they get expensive in different currencies.

Two Growth Curves, Side by Side

Put the two mechanisms next to each other and the differences are sharper than the similarities.

The regulatory line at the bottom of that table is worth pausing on, and it gets fuller treatment later in this piece, but the short version is that both activities now sit on the same side of a line US securities regulators drew in 2025 and 2026, even though the underlying economics could hardly be more different.

Restaking: A Third Growth Curve Stacked on Top of Staking

Ethereum’s staking growth is not the end of the story; a third layer has grown on top of it. Restaking protocols let validators reuse the same staked ETH to help secure additional services, oracles, bridges, data availability layers and other infrastructure collectively called AVSs, or actively validated services, in exchange for extra yield, effectively renting out the same security budget twice. EigenLayer pioneered the model and still holds the largest share of restaked capital, with tens of billions of dollars in assets restaked across its ecosystem at last count.

The pitch is that restaking compounds security growth for free: the same staked ETH now backs more things, without asking anyone to lock up new capital. The counterargument, examined in Restaking’s Next Act: EigenCloud Bets on Verifiable AI, is that it compounds risk instead, since a slashing event triggered by one AVS’s bug or exploit can now reach into capital that was originally staked to secure something else entirely. Not every project wants that trade-off; some designs deliberately skip the token and the shared-slashing model altogether, an approach covered in Restaking Without a Token: Symbiotic, Karak and OpenGDP. Growth in restaked value is not the same kind of growth as growth in base-layer staked ETH; it is leverage layered on top of it, and leverage cuts both ways.

Concentration Risk on Both Sides of the Ledger

Growth in either metric says nothing about how many hands actually control it, and both networks have a similar answer: fewer than the scale of the numbers might suggest. On the mining side, three pools, Foundry USA at 25.3%, F2Pool at 17.7% and AntPool at 16.8%, currently account for roughly 60% of Bitcoin’s blocks between them, according to Hashrate Index’s live pool tracker, even though tens of thousands of individual machines ultimately point at those pools. A pool operator, not any individual miner, decides which transactions go into the next block, which has already produced real censorship disputes over sanctioned transactions in past years.

Ethereum’s staking side shows a milder but comparable pattern. Lido, the largest liquid staking provider, currently controls around 24% of all staked ETH, down from a peak near 32% in 2023 as competitors have chipped away at its share, according to CCN. That decline is a genuine decentralization win relative to 2023, when Lido’s share sat close enough to the one-third finality-disruption threshold to worry researchers, but it still means roughly a quarter of Ethereum’s entire security budget runs through a single protocol’s smart contracts and node operator set.

Neither network has solved concentration by growing. Bitcoin’s hashrate can set a new record while its pool-level Nakamoto coefficient, the smallest number of entities needed to control the majority, stays stuck at three or four. Ethereum’s staked ETH can climb past 41 million while its largest single operator still commands a share large enough to draw regulatory attention on its own. Growth and decentralization are different axes, and both ecosystems are still working out how to move on both at once.

Where the Capital Is Coming From: AI Money Meets Restaking Money

Hashrate growth used to be funded almost entirely by mining revenue reinvested into more machines. That is no longer true. The largest US-listed miners have leaned on AI and HPC, high-performance computing, infrastructure deals, debt raises and equity issuance to fund expansion that increasingly has as much to do with data center real estate as with hashrate. Some of that capital still buys ASICs; a growing share buys power contracts, land and cooling infrastructure that could just as easily host GPUs as SHA-256 chips, blurring what a rising hashrate number even represents.

The cost of that capital is not fixed, and interest rate policy reaches into hashrate growth more directly than it might seem. A hawkish rate path raises the cost of the debt miners have used to fund fleet expansion and data center buildouts, tightening exactly the kind of financing that turned into new exahashes over the past two years, a dynamic explored in Bitcoin Halving Cycle Math: A Hawkish Fed Tests the Cycle. Restaking has its own, separate capital story running in parallel: institutional allocators have started treating restaked ETH yield as a fixed-income-adjacent product, pulling in capital that has little to do with Bitcoin’s rate environment at all, evidence that the two growth curves this piece compares are increasingly funded out of different pools of money, not just secured through different mechanisms.

How Regulators Treat Mining Rewards vs Staking Rewards

US securities regulators have, over the past two years, converged on treating both activities the same way: as not securities transactions, at least under current staff guidance. The SEC’s Division of Corporation Finance said in a March 2025 statement that proof-of-work mining, whether solo or through a pool, does not implicate securities law, since rewards come from a miner’s own computational contribution rather than a third party’s managerial effort, the standard question under the Howey test, according to The Block.

A broader release in March 2026 extended that same reasoning to staking, along with wrapping and no-consideration airdrops, under the SEC’s Project Crypto initiative. “This is what regulatory agencies are supposed to do: draw clear lines in clear terms,” SEC Chair Paul Atkins said of the release, according to the SEC’s own newsroom announcement. The practical effect is that a miner plugging in an ASIC and a validator bonding ETH now sit on the same side of a regulatory line that, as recently as 2023, looked far less settled, particularly for staking-as-a-service products that had drawn separate SEC enforcement attention. None of this is binding law; staff statements can be withdrawn by a future commission. But it is the clearest signal either industry has gotten, and a fuller look at how that broader enforcement posture has actually played out runs in The SEC’s Report Card on Its Own Crypto Enforcement.

What Could Bend Either Growth Curve

Both growth curves face real tests ahead. Bitcoin’s next halving, expected around April 2028, will cut the block subsidy from 3.125 to 1.5625 BTC, mechanically cutting miner revenue in half unless price or fee income compensates; the resulting shakeout among less efficient hashrate is a running debate across this site’s halving-cycle coverage. If hashprice stays depressed into that event the way it has through most of 2026, expect another round of the same consolidation that has already reshaped the public mining sector once this cycle.

There is also a simpler, more mechanical risk sitting underneath all the efficiency gains celebrated earlier in this piece: an ASIC glut. Manufacturers keep shipping faster chips regardless of whether hashprice can support them, and a sudden wave of cheap, only-slightly-outdated hardware hitting the secondhand market, whenever a large miner restructures or fails, can push hashrate higher for reasons that have nothing to do with confidence in Bitcoin and everything to do with clearing inventory.

On the staking side, the entry queue backlog itself is worth watching: a 43-day wait to start earning yield is a real opportunity cost, and if it lengthens further it could eventually push new capital toward liquid staking tokens, restaking, or competing chains rather than the base layer, concentrating influence exactly where the section above already flagged it as thin. Restaking’s slashing risk is the other open question: the model has not yet been tested by a genuine cascading failure across multiple AVSs at once, and how validators and token holders respond to a first real cascade will say more about the durability of stacked security growth than any single TVL chart can.

The honest conclusion is that neither hashrate nor staked ETH is a number that only goes up because the underlying asset is good. Both are the output of thousands of independent capital-allocation decisions, responding to price, yield, financing costs and regulatory clarity in something close to real time. Watching either curve bend, not just climb, is usually where the more interesting story is.

Frequently Asked Questions

What is Bitcoin’s hashrate and why does it matter?

Hashrate is the total number of guesses per second Bitcoin’s mining machines make while searching for a valid block, measured in units from megahashes up to zettahashes. It matters because it is the clearest proxy for how much money and electricity would have to be outspent by anyone trying to rewrite Bitcoin’s transaction history; a higher hashrate generally means a more expensive network to attack.

Why does Bitcoin’s hashrate keep growing even when mining is unprofitable?

Growth comes from two separate sources: new capital buying additional machines, and each ASIC generation squeezing more hashrate out of the same electricity than the last one. Even during stretches when hashprice is near multi-year lows, replacing older, less efficient machines with newer ones can push total network hashrate higher while cutting an individual miner’s power bill, so hashrate and miner profitability do not always move together.

How much would it actually cost to attack Bitcoin’s network today?

Duke University’s Campbell Harvey estimated in 2025 that sustaining majority control of Bitcoin’s hashrate for one week would cost around $6 billion in hardware, data center construction and electricity, a figure he has since revisited using models that involve derivatives markets. Critics counter that the practical logistics of acquiring that much hashpower quickly, without anyone noticing, make the theoretical math less alarming than it looks.

Is Ethereum’s staking growth comparable to Bitcoin’s hashrate growth?

They answer the same underlying question, how expensive is it to attack the network, but through different mechanisms. Bitcoin’s security comes from sunk hardware and electricity costs that keep existing whether or not an attack happens. Ethereum’s comes from staked ETH that can be partly destroyed through slashing if validators act dishonestly, so growth in staked ETH represents capital that is directly at risk rather than an external, sunk cost.

What could cause Bitcoin’s hashrate to fall significantly?

The clearest historical trigger is a sustained drop in hashprice, miner revenue per unit of hashrate, which pushes higher-cost operators to power down until Bitcoin’s difficulty adjustment recalibrates. Bitcoin’s 2028 halving, which will cut the block subsidy in half, is the next scheduled event likely to test this again, alongside any renewed regulatory or energy-policy pressure in major mining jurisdictions.

Yuki Tanaka covers Bitcoin mining, staking infrastructure and network security for HOGE Wire.