Proof of work is a consensus mechanism that selects block producers based on who can demonstrably burn the most computational effort. A miner constructs a candidate block, hashes its header, and checks whether the result is below a target value. Because hash functions produce effectively random outputs, the only way to find a valid hash is to try different nonce values until one works — there is no shortcut, no shortcut key, no clever algorithm that avoids the brute force. The miner who finds a valid hash first gets to publish the block and claim the reward, and all other miners verify the work instantly by hashing the header once. Producing the proof is expensive; verifying it is free.
This asymmetry is what makes proof of work useful as a consensus mechanism. An attacker who wanted to rewrite a chain’s history would have to redo all the mining work for every block since the point of the attack, at the same pace as (or faster than) the honest network. For a chain like Bitcoin, where the cumulative mining work represents tens of billions of dollars of hardware and electricity, this is economically prohibitive in any realistic scenario. The chain’s immutability is not a legal promise or a social convention; it is a statement about how much energy would have to be spent to change it.
Why Satoshi Chose It
The original Bitcoin whitepaper framed proof of work as the solution to a specific problem: in an open peer-to-peer network where anyone can join, how do you decide whose version of history is the real one? You cannot vote based on identity because identity is not meaningful in a permissionless network; creating new identities is free, so any vote-based system can be trivially attacked by flooding it with fake identities (a Sybil attack). You need a voting system where votes are expensive in some real, physical sense that cannot be faked.
Satoshi’s answer was to make votes cost CPU cycles. “One CPU, one vote” — though the reality quickly became “one ASIC, one vote” as specialised hardware took over — gives each participant voting power proportional to their share of computational effort. An attacker cannot fake computational effort; they have to actually do it. The energy the network consumes is, in effect, the cost of maintaining an unforgeable record of who said what and in what order.
This was a genuinely novel combination of existing ideas (Adam Back’s Hashcash had used proof of work for anti-spam, and the cryptography was standard), and it is why Bitcoin is frequently described as a significant intellectual achievement regardless of what you think about the current state of the industry it spawned.
The Energy Debate
Proof of work’s most visible cost is electricity. Bitcoin’s network currently consumes somewhere between 100 and 200 terawatt-hours per year, comparable to a mid-sized country. Ethereum used to consume a similar amount before it moved to proof of stake in 2022, at which point its energy consumption dropped by roughly three orders of magnitude overnight. Other major PoW chains (Dogecoin, Litecoin, Monero, Kaspa) consume less than Bitcoin but are still non-trivial energy users.
The defense is that the energy is doing useful work: securing a global settlement network against tampering. The critique is that the same security could be achieved with proof of stake at a tiny fraction of the energy cost, and that the defense is essentially begging the question — you are only using proof of work because you believe proof of work is necessary, and the energy consumption is the direct consequence of that belief. Whether PoW is actually necessary, or whether PoS provides comparable security with vastly lower externalities, is one of the more genuine disagreements in the space. Both sides have real arguments; both sides have blind spots.
What Proof of Work Gets You That Proof of Stake Doesn’t (Maybe)
The strongest case for proof of work is its relative simplicity and its track record. Bitcoin’s consensus rules have been stable and secure since 2009, and the security model is easy to reason about: whoever has the most hash rate wins, and acquiring hash rate requires buying real hardware and paying real electricity bills. There are no “weak subjectivity” checkpoints, no staking derivatives creating reflexive feedback loops, no edge cases where attackers can use old keys to forge alternate histories.
The PoS counter is that Bitcoin’s security is itself dependent on a continuing subsidy from block rewards, and as those rewards shrink with each halving, transaction fees will have to pick up the slack. Whether fee-only mining will be sufficient to secure the network long-term is a real open question that nobody knows the answer to, and the argument that PoW is more battle-tested partially evaporates once the battle-testing is happening in a regime the chain has never actually experienced.
Both systems work. Both have problems. Bitcoin is the biggest proof-of-work chain and will probably remain one for a long time. Ethereum is the biggest proof-of-stake chain and has demonstrated that the alternative can work at scale. The choice between them is increasingly more about tradeoffs than about which one is correct, and the tradeoffs are not symmetric enough for one side to claim final victory.