A hard fork is a change to a blockchain’s consensus rules that is not backward compatible: blocks or transactions valid under the new rules are rejected by nodes still running the old software. Among the blockchain fundamentals, it is the most decisive way a network can change how it stores and agrees on data, because every participant must choose which rule set to follow.
In crypto & blockchain, a hard fork can be a planned, coordinated network upgrade in which nearly all nodes adopt the new rules, or a contentious split in which part of the community keeps the old rules and two separate chains with separate coins continue from a shared history.
How a hard fork changes consensus rules
A hard fork works by expanding or altering the set of valid blocks so that the new set is no longer a subset of the old one. Every full node independently validates each block against its own copy of the rules. When an upgraded node accepts a block that an old node considers invalid, the two nodes stop agreeing on the chain from that block height onward.
Most hard forks are scheduled for an activation point, usually a specific block height, an epoch or a timestamp written into the new client software. Before that point the upgraded and old nodes follow the same chain. At activation, the first block that uses a new rule is the moment of divergence, and any node that has not upgraded is effectively left on a separate chain, often with very little hash power or stake behind it.
The table below shows typical changes that require a hard fork and why old nodes cannot accept them.
| Change | Example | Why old nodes reject it |
|---|---|---|
| Larger block size limit | Raising a 1 MB cap to 8 MB | Blocks exceed the old maximum |
| New transaction format | Added fields or new signature types | Old parsers treat data as invalid |
| Consensus mechanism switch | Proof of work to proof of stake | Old rules expect different block proofs |
| Changed issuance or fees | New reward schedule or fee burning | Coinbase or balance checks fail |
| State correction | Reversing a specific exploit | Balances differ from replayed history |
Hard fork versus soft fork
The essential difference is compatibility direction. A soft fork only tightens the rules, so blocks produced under the new rules remain valid for old nodes, and an un-upgraded node keeps following the main chain. A hard fork loosens or changes the rules, so old nodes cannot validate new blocks and must upgrade to stay on the same chain.
| Property | Hard fork | Soft fork |
|---|---|---|
| Rule change | Loosens or alters rules | Tightens rules only |
| Old nodes | Reject new blocks | Still accept new blocks |
| Upgrade requirement | All validating nodes | Mainly block producers |
| Risk of chain split | High if support is divided | Low with majority enforcement |
| Design flexibility | Almost unlimited | Limited to restrictions |
Different communities prefer different tools. Bitcoin developers have historically favored soft forks for conservatism, as with Segregated Witness and Taproot, while Ethereum has used regular scheduled hard forks such as London and the Merge as its normal upgrade path. Neither approach is inherently safer; the risk depends on coordination and community agreement.
Planned upgrades and contentious splits
Hard forks fall into two broad categories depending on whether the community agrees. In a planned upgrade, developers, node operators, exchanges and wallet providers coordinate in advance, so the old chain dies out quickly. In a contentious fork, a significant minority rejects the change and continues to produce blocks under the old rules, creating two persistent chains and two assets.
| Event | Year | Type | Result |
|---|---|---|---|
| Ethereum and Ethereum Classic | 2016 | Contentious | Chain split after the DAO exploit reversal |
| Bitcoin Cash | 2017 | Contentious | Separate chain with larger blocks |
| Ethereum London | 2021 | Planned upgrade | Fee burning introduced, no lasting split |
| The Merge | 2022 | Planned upgrade | Switch from proof of work to proof of stake |
When a split occurs, every address that held coins before the fork holds the same balance on both chains. This is sometimes described as an airdrop, but the market value is not simply added: prices on each chain are set independently, and the combined value can be lower than the pre-fork value if the split damages confidence.
Replay attacks and other technical risks
The most serious technical risk after a chain split is the replay attack. Because both chains share the same history and key formats, a transaction signed for one chain may also be valid on the other, so a user who sends coins on one chain can unintentionally send the same amount on the second chain.
Developers prevent this with replay protection, for example a chain identifier included in signatures, as Ethereum introduced through EIP-155, or a changed signature hashing scheme. Forks without strong replay protection force users to split their coins manually, which is error-prone and has caused real losses.
| Risk | Cause | Mitigation |
|---|---|---|
| Replay attack | Same transaction valid on both chains | Chain ID in signatures |
| Accidental split | Nodes fail to upgrade in time | Long notice and client alerts |
| Low-security minority chain | Little hash power or stake remains | Higher confirmation counts |
| Exchange confusion | Unclear ticker and deposit handling | Pause deposits around activation |
| Consensus bugs | Clients implement rules differently | Testnet forks and client diversity |
A chain with little remaining hash power is also vulnerable to a 51% attack, because renting enough mining capacity to overpower it becomes cheap. Exchanges therefore often raise required confirmations for a newly split chain.
How node operators and users prepare
Preparing for a hard fork means upgrading software before the activation point and checking how services will treat each chain. The responsibilities differ by role.
- Node operators install the new client release, verify the activation height and monitor that their node follows the expected chain.
- Validators and miners upgrade early, because producing blocks on the wrong chain wastes rewards and may incur penalties.
- Exchanges and custodians often pause deposits and withdrawals around activation and announce whether they will support both assets.
- Ordinary holders usually need to do nothing in a planned upgrade, because wallets and exchanges handle the change.
- Holders during a contentious split should keep private keys under their own control and wait for replay protection before moving coins.
A common misconception is that a hard fork creates a new blockchain from scratch. In reality, the forked chain inherits the full history up to the fork point, including the genesis block, all balances and all smart contract states; only the blocks after activation differ.
In simple terms
Imagine a sports league that changes a rule so that a goal scored in a new way now counts. Referees who learned the new rules accept the goal, while referees still using the old rulebook reject it. If everyone switches, the game simply continues under the new rules.
A hard fork works the same way for a blockchain. Every computer checking the network must install the new rules. If some refuse, they keep their own version of the game, and the network splits into two separate chains with their own coins.
A practical example
A small cryptocurrency exchange learns that a major network has scheduled a hard fork at a specific block height roughly six weeks away. The operations team installs the new node release on a test environment, then upgrades its production nodes two weeks before activation to leave time for problems.
On the day of the fork, the exchange pauses deposits and withdrawals for that asset about one hour before the expected block and publishes a notice to customers. After activation, engineers confirm that their nodes follow the chain supported by the large majority of validators, compare block hashes with public block explorers and check that a small test transaction confirms normally.
Because the upgrade was uncontested, no competing chain gains meaningful support. The exchange reopens deposits about three hours later, and customer balances remain unchanged. Had a split occurred, the team would have waited for replay protection before crediting the second asset.