What is proof of work?

Proof of work (PoW) is a category of consensus mechanism that protects a blockchain through computational cost. Participants have to compute repeatedly until they find a result that meets the protocol's requirement before they get the chance to publish a new block, while other nodes can check that result quickly.

The point is not to have computers solve the transactions themselves, but to attach a public, measurable cost to proposing a block. Finding a qualifying result usually takes a large number of attempts; verifying one is easy. That asymmetry lets nodes compute an accumulated work score for a stretch of history from the proofs inside its blocks. An attacker who wants to replace that history has to redo the corresponding work and catch up with the network.

The problem proof of work addresses

An open network cannot decide who keeps the ledger on the basis of "one vote per participant", because the same person can create any number of identities. Proof of work does not count identities. It asks whoever wants to compete to submit a verifiable computational result. Creating more identities does not by itself improve the chance of success; more hash power — that is, the ability to complete more hash attempts per unit of time — does raise the probability of finding a result.

That computation helps the network handle two questions at once: who gets the opportunity to publish a candidate block, and which stretch of history nodes keep using when several valid branches appear at the same time. In proof-of-work networks such as Bitcoin, nodes compare the work accumulated by each branch rather than simply comparing which branch contains more blocks.

Proof of work does not replace transaction validation. Whoever finds a qualifying result can only propose a candidate block; it does not let them change the protocol rules. Every fully validating node still independently checks transaction signatures, whether assets have already been spent, block format and the proof of work itself. A block that breaks the rules is rejected even though hash power was spent on it.

From candidate block to network confirmation

Proof-of-work networks differ in algorithm and parameters, but the basic process can be divided into the following stages.

Assembling a candidate block

Participants who compete to produce blocks are usually called miners. A miner selects content from transactions that are not yet confirmed, assembles a candidate block as the protocol requires, and prepares the block header and other data that will go into the hash computation. A miner can decide which valid transactions to include and in what order, but cannot make an invalid transaction valid.

Hashing repeatedly

A cryptographic hash function converts an input into a result of fixed length. The miner changes the nonce — a value used for repeated attempts — or another variable field in the block, computing the hash again and again until the result satisfies the target condition the protocol has set.

As long as the hash function the protocol uses has no known effective break, there is no reliable shortcut to a qualifying result. Whether any single attempt succeeds is hard to predict, so more hash power means more attempts per unit of time and a higher probability of finding a qualifying result first. A participant with less hash power may still succeed first: ownership of an individual block is not decided in advance.

Broadcasting and validating the block

Once a miner finds a qualifying result, the block is broadcast to the network. A node that receives it does not need to redo all the attempts the miner made; it only computes and checks whether the result satisfies the target condition, and verifies that the block and its transactions follow the other rules.

Only nodes that complete this validation keep relaying and processing the block as a valid candidate. The distinction matters: mining competes to propose a block, and node validation decides whether a block qualifies to be accepted.

Continuing along the branch with more accumulated work

Two miners may publish different but equally valid blocks at close to the same time, which temporarily leaves the network with branches. Nodes follow the protocol's chain selection rule to decide which branch to continue on. For example, Bitcoin nodes select the valid chain with more accumulated work; as further work is added, one branch usually pulls ahead and the other is abandoned.

Transactions that were in the abandoned branch do not disappear automatically. If a transaction is still valid, it can return to the pending state and be included in another block later. If it conflicts with a transaction in the winning branch, it will not take effect again.

Target, difficulty, hash power and accumulated work

Four terms cause most of the confusion around proof of work.

Target is the numeric threshold a qualifying hash has to satisfy. Taking the common form "the hash value must be less than or equal to the target", a smaller target means fewer results qualify and more attempts are needed on average.

Difficulty is another way of expressing how hard it is to find a qualifying result. Higher difficulty normally means a stricter target. Many proof-of-work networks adjust difficulty periodically so that average block production stays roughly in line with the pace the protocol intends even as total hash power changes. The adjustment method and period are set by the individual network.

Hash power, or hash rate, is how many hash attempts can be completed per unit of time. It reflects computational input. It does not indicate how many nodes a miner controls, and it is not the same as work already accumulated. When choosing between branches, nodes rely on the work proven by blocks, not on estimates of live hash power.

Accumulated work is the sum of the work scores the protocol derives from the target or difficulty of each valid block. That score represents how many hash attempts generating this history is expected to take on average. It is not a direct measurement of how many times miners actually failed, how much electricity they used, or their current hash rate. Nodes can calculate the score independently from block data, so they do not have to trust costs that miners report themselves.

Confirmation is probabilistic

Proof of work generally provides probabilistic confirmation rather than making a new block permanently unchangeable the moment it appears. As further blocks are added after the block containing a transaction, the work required to replace the history holding that transaction also grows, so the chance of being affected by a chain reorganization — where a recently adopted branch is replaced by another branch — normally falls.

"Wait for a few confirmations and it is safe" has no fixed answer that applies everywhere. It depends on the distribution of hash power in the network, block production speed, the value of the transaction, how much risk the recipient can absorb, and whether there is a network outage or an unusual fork. The confirmation count a wallet or platform displays only states how many blocks have followed that transaction; it is not an unconditional guarantee.

A chain reorganization does not necessarily mean an attack took place either. Network latency can leave honest miners working on top of different blocks for a while, after which the branch with more accumulated work resolves the conflict. The more recent blocks a reorganization covers, the rarer it usually is — and the more its effects deserve attention.

What majority hash power can do

If one participant, or a group acting together, controls enough effective hash power over a sustained period, it may extend its own branch faster than the rest of the network. This kind of majority hash power attack is often called a 51% attack, but "51%" is only a convenient label; actual success also depends on duration, network conditions and the attacker's objective.

A miner with that advantage might try to exclude or delay certain transactions, reorganize recent blocks, or pay someone and then withdraw its own payment through another branch, which is a double spend. It may also keep other miners' blocks out of the branch the network eventually adopts.

But an advantage in hash power is not permission to rewrite the ledger at will. An attacker cannot produce someone else's valid signature out of nothing, cannot spend assets it does not control the keys for, and cannot make honest validating nodes accept blocks that break supply rules or transaction validation rules. Proof of work settles competition between valid histories; it does not switch off the validity rules.

Mining rewards are not proof of work itself

Many public networks reward whoever produces a block with newly issued assets and transaction fees, offsetting equipment, electricity and operating costs. Those rewards give miners a reason to participate, but the form of the reward is not what defines proof of work.

Proof of work describes how verifiable computation is submitted, how blocks are competed for, and how histories are compared. Block subsidies, transaction fees, supply caps and reward reductions belong to an individual network's economic rules. The total income miners receive from subsidy and fees is often called the network's security budget, because it sustains their continued investment in hash power. Those income rules affect the security conditions of proof of work, but they cannot be equated with the consensus mechanism itself.

A mining pool is not a new consensus mechanism either. A pool combines the hash power of several participants to mine cooperatively and distributes proceeds under an agreement. It can reduce how much an individual participant's income fluctuates. But when a large share of hash power depends on a small number of pool operators, those operators' outages, transaction filtering or coordinated behaviour affect more participants.

Costs and trade-offs

Proof of work brings real resource costs into a digital network. An attacker that wants to keep catching up with the honest network generally has to commit equipment, electricity and operational resources, and honest miners have to bear the same categories of cost. These resource costs set an economic threshold for rewriting history and for sustaining an advantage over honest miners, and they are part of the proof-of-work security model. Overall security still depends on nodes validating correctly, on the distribution of hash power, on network connectivity, on the hash function and on software implementations.

That also creates clear trade-offs:

• Energy and equipment costs: large numbers of hash attempts consume electricity and call for specialized or high-performance hardware. The actual energy mix and emissions impact depend on region, equipment, where the load comes from and how the figures are measured.
• Concentration of hash power: scale, cheap electricity, chip supply and pool operations can concentrate hash power among a small number of participants.
• Waiting for confirmations: probabilistic confirmation needs time to accumulate, and one block cannot be treated as absolute settlement.
• A changing security budget: when block subsidies decline, or asset prices and fee income change, miner revenue and total network hash power may change with them.
• Coordinating rule upgrades: hash power can influence block production and branch competition, but protocol upgrades still depend on how software, nodes, miners, users and other participants adopt the rules.

Energy use also cannot be reduced to "every transaction necessarily consumes this much electricity" by averaging over a transaction count. Miners compete for blocks, and how many transactions a block can hold, whether it aggregates batch settlement data submitted by other systems, and whether empty blocks occur are all determined by the specific system and how it is used. When assessing energy impact, be explicit about whether the discussion concerns total network consumption, a unit of time, a single block, or the marginal effect of a category of transaction.

Proof of work compared with other consensus mechanisms

Proof of work uses external computational resources to limit the competition to produce blocks. Proof of stake generally has validators who stake the protocol's asset take part in proposing and voting, with rewards or penalties constraining behaviour. A permissioned network may simply specify which institutions or nodes can take part in consensus.

All of these mechanisms have to define who can propose data, how other nodes validate it, which result is adopted when there is a conflict, and when a transaction can be treated as sufficiently settled. They differ in which resource limits influence, which participants they assume may act dishonestly, and who bears the cost of security.

So a mechanism cannot be judged safe or suitable on the single question of whether it consumes electricity. The comparison also has to cover who can take part, how much hardware or staked asset is needed, whether influence is concentrated, how transactions are confirmed, whether recovery after an attack is possible, how nodes check data, and who upgrades the rules.

FAQ

Is proof of work the same thing as mining?

Not exactly. Mining is the process in which participants compute and compete to propose blocks; proof of work also covers how other nodes validate the computational result and how the network compares several valid histories. The two words are often used together in everyday speech, but they do not cover the same ground.

Do miners compute the transactions?

No. Miners mainly hash block-related data over and over, looking for a result that meets the target. Transaction execution and validity checks belong to a separate set of protocol rules.

Will the miner with the most hash power always produce the next block?

Not necessarily. More hash power means more attempts in the same period and therefore a higher probability, but any single round of competition still involves randomness. Long-run results tend to track each participant's share of hash power more closely.

Can proof of work prevent every double spend?

No. It raises the cost of reorganizing history through accumulated work, and risk normally falls as confirmations accumulate, but double spends remain possible when hash power is concentrated, confirmations are insufficient or the network behaves abnormally.

Do all blockchains use proof of work?

No. Some networks use proof of stake, permissioned voting or other mechanisms. The same network may also change consensus mechanism through a protocol upgrade.

Related concepts

Continue with consensus mechanism, block, blockchain node and cryptographic hash, or return to the Crypto Glossary.

Sources

• Bitcoin white paper — the original design account of proof of work, branch competition, accumulated work and confirmation probability.
• Bitcoin Developer Guide: block chain — block header hashing, target, difficulty adjustment, accumulated work and chain reorganization.
• Bitcoin.org: FAQ — the limits of miner authority, confirmations, hash power attacks and mining incentives.
• NISTIR 8202: Blockchain Technology Overview — the general proof-of-work process, its role in consensus and its resource trade-offs.