Hashrate & mining pools

Mining is the part of Bitcoin most people get wrong, and the part where the most money is made and lost. The dashboard on this site shows a number — somewhere between 700 and 800 exahashes per second — that I want to unpack, because behind that figure sits a global industrial system burning real electricity to defend a public ledger. As a Thai engineer running Bitcoin Core since 2017, who has resisted the temptation to plug an S19 into his own breaker box, I want to walk through what hashrate actually measures, how pools work, what hashprice tells you, and what the energy debate looks like when you set down the slogans.

What hashrate actually measures

Hashrate is the total number of attempts per second, across every miner on Earth, to find a valid block hash. Today the network sits near 700–800 EH/s — 7 × 10²⁰ to 8 × 10²⁰ hashes per second. A modern Bitmain S21 does roughly 200 TH/s on its own, so the global network is equivalent to three to four million such machines running flat out, every second, forever.

You don’t measure hashrate directly. There is no metering on the network. What you observe is the rate at which valid blocks are found and the difficulty target enforced by the protocol; from those two quantities you back out an estimated hashrate. mempool.space/mining shows live numbers smoothed over rolling windows. Treat any instantaneous figure with skepticism — block-finding is a Poisson process and short windows are noisy. The 7-day or 30-day average is the honest number.

What a “hash” is

A hash is one round of SHA-256 applied to the 80-byte block header. The header contains the previous block hash, the Merkle root of transactions, a timestamp, the difficulty bits, and a 32-bit nonce. Miners vary the nonce (and the extranonce inside the coinbase, which changes the Merkle root) and rehash, looking for an output that, interpreted as a 256-bit number, sits below the current target. Lower target = harder problem = higher difficulty. A successful hash is just a number with enough leading zeros — pure brute force, no shortcut. SHA-256 has no known structural weakness that lets you skip the search.

Mining hardware, four eras

Mining hardware has gone through four generations:

• 2009–2010: CPUs. Satoshi mined on a CPU. Megahashes per second.
• 2010–2013: GPUs. Graphics cards parallelize SHA-256 by orders of magnitude. Gigahashes per second.
• 2013: FPGAs. A brief window where field-programmable arrays squeezed more efficiency out of the same paradigm.
• 2013–today: ASICs. Dedicated chips that do nothing but SHA-256, etched at the cutting edge of process node availability. A modern ASIC outputs 200–400 terahashes per second at roughly 15–20 joules per terahash.

The trajectory is one-way. You cannot meaningfully mine Bitcoin on a CPU or GPU today; difficulty has long priced them out by factors of millions. Bitcoin mining is now the most efficient SHA-256 industry on the planet.

Hashprice — the mining business in one number

Hashprice is dollars earned per petahash of compute per day. It is the single number that captures the mining economy.

hashprice ≈ (subsidy + fees) × BTC_price × 86400 / network_hashrate

If price rises while hashrate stays flat, hashprice rises and miners get richer. If hashrate climbs while price stays flat, hashprice falls and inefficient rigs get unplugged. After every halving, hashprice halves overnight on the subsidy side, and miners with marginal margins go offline until either price rises or the next difficulty adjustment makes the survivors more profitable per joule. Hashrate Index tracks hashprice in real time and is the closest thing to a stock ticker the mining industry has.

A mining business that does not understand hashprice is not a business. It is a leveraged bet on price, dressed up.

What a pool actually is

Solo mining today is a lottery. With 800 EH/s on the network and a single S21 at 200 TH/s, your share is roughly one in four million. You might find a block once a year, or never. The variance ruins your cash flow.

A pool fixes the variance. Members point their hashrate at the pool’s coordinator, the pool combines the work, and when the pool finds a block it distributes the reward proportionally. Members get smooth daily payouts instead of a lottery ticket. Common payout schemes include PPLNS (pay per last N shares), FPPS (full pay per share, where the pool eats the variance), and PPS+. The pool charges 1–4% for absorbing the variance.

Critically, a pool is usually not the company that owns the miners. It is a coordination layer over independent operators. ViaBTC, F2Pool, AntPool — these are mostly aggregators, not industrial mining firms. MARA Pool is an exception where a single company runs both its own hashrate and the pool.

Top pools today

The latest 24-hour pool distribution comes from the mempool.space pools API. At time of writing the top names are Foundry USA, AntPool, F2Pool, ViaBTC, Binance Pool, and MARA Pool. The top five together typically account for 75–85% of blocks. Foundry alone has at times sat above 30%.

Those numbers look alarming if you read them as ownership. They are not ownership numbers. They are coordination numbers. A miner who is unhappy with their pool’s policy — say, censorship of certain transactions — can repoint their hashrate to a different pool in minutes. The pool brand on a block is a thin layer on top of a much more distributed reality.

Pool centralization vs hashpower centralization

This is the distinction that most takes online get wrong. Hashpower is owned by thousands of mining operations, scattered across continents, holding wildly different machines. Pools are a small number of coordination services those operations choose to use. The risk of pool centralization is real — a pool can attempt transaction censorship, can reorganize at the margins, can build templates that exclude certain outputs — but it is bounded by the fact that the underlying miners can leave at any time.

The community gets nervous when a single pool crosses 35–40% share, and it is worth being nervous about. It is also worth remembering that the 2014 GHash.io incident, where a pool briefly exceeded 50%, resolved precisely because miners voluntarily left. Coordination is not capture.

Stratum V2

Stratum V2 is the protocol upgrade that addresses this concern at the technical layer. Under the legacy Stratum V1, the pool builds the block template and tells the miner what to hash. The miner has no say in which transactions go in the block. Under Stratum V2, miners can construct their own templates and submit them to the pool, while still receiving pooled payouts. The censorship surface collapses. As of 2026 adoption is still partial, but the direction is clear, and serious operators are migrating.

The energy story without the slogans

Bitcoin mining uses around 0.4% of global electricity, with a credible range of 0.2% to 0.6% depending on methodology. The Cambridge Bitcoin Electricity Consumption Index is the best public source. That is a real number and not a small one.

It is also a number with unusual properties. Mining is uniquely interruptible — a rig switches off and on in seconds with no ramp cost — which makes it the ideal demand-side counterparty for stranded or curtailed generation: hydro spillage in wet seasons, flared methane at oil wells, off-peak nuclear, surplus wind and solar. Texas grid operators have explicitly used miners as a flexible load to stabilize ERCOT.

The honest framing is not “Bitcoin is green” or “Bitcoin is dirty.” Both are lazy. Mining is energy-intensive by design, increasingly clean by economic incentive, and one of the few large electrical loads on Earth that strengthens rather than strains a grid. The Cambridge mining map is the geographic picture.

Geography after the China ban

Before mid-2021, China hosted around two-thirds of global hashrate. The May 2021 ban scattered that capacity within months. The United States took the largest share — Texas, Georgia, the Dakotas — followed by Russia, Kazakhstan, Canada, Malaysia, and Paraguay. ASIC mining migrates to wherever electricity is cheap. The map redraws every couple of years.

Block reward today

Each block today pays 3.125 BTC of subsidy plus typically 0.05–0.5 BTC in fees. With around 144 blocks per day, that is roughly 450 BTC issued daily plus fees. At a $75,000 price, the network pays around $33 million per day — $12 billion per year — to the entire mining industry. That is Bitcoin’s security budget: the dollar-denominated cost an attacker would have to outspend to attempt 51% control. Every joule miners burn is value defending the chain.

Halving impact on miners

Each halving cuts the subsidy in half. Margins squeeze, inefficient miners unplug, difficulty adjusts down, the survivors regain profitability. The long arc, by design, is a transition from issuance-funded security to fee-funded security. Today fees are a small fraction of revenue. By 2032 they will need to be a meaningful fraction, and by 2140, when the last fragment of subsidy is issued, fees will be the entire game.

Reading the dashboard

When you look at the mining tile on this site, four numbers matter:

• Current hashrate in EH/s — the headline figure, noisy on short windows.
• 24h or 7d average — the smoother truth.
• Pool distribution over the last 24 hours — who found blocks recently.
• Top pool share — anything above 35% deserves a second look.

I do not mine. Thailand’s residential power rates keep climbing, and home-miner economics never penciled out past hobby tier. But I watch hashrate every week. It tells me how hard it would be for a hostile actor to rewrite the chain my coins live on. As long as it keeps climbing, the network gets more expensive to attack — the whole point of proof of work.