Bitcoin blocks

Every ten minutes or so, a Bitcoin miner finds a number that satisfies a hash puzzle, broadcasts a block, and the network’s ledger advances by one row. As a Thai engineer who has been running Bitcoin Core in a closet in Bangkok since 2017, I have watched my node tick through hundreds of thousands of these. Each one is small. None are dramatic. Cumulatively, they are the entire reason Bitcoin works. If price is the noise, blocks are the signal — the actual unit in which the protocol does its only job.

What a block actually is

A block is a sealed batch of transactions plus a small fixed-size header — up to about 1 megabyte of raw data, or roughly 4 megabytes when measured in weight units (the post-SegWit accounting that discounts witness data). Inside, you will find one mandatory coinbase transaction — the special one that pays the miner — followed by however many regular transactions the miner chose to include from their mempool. The header is 80 bytes, summarizing everything below it.

The chain is a chain because each header references the hash of the previous block. Tamper with any old block and every subsequent header becomes invalid. That is the security model, dressed up.

The block header, field by field

Six fields, 80 bytes, fully specified in the Bitcoin developer reference:

• Version — protocol version bits, occasionally used for soft-fork signaling
• Previous block hash — links this block to its parent
• Merkle root — a single hash that commits to every transaction in the block body
• Timestamp — when the miner claims the block was assembled (with some tolerance)
• Bits — the difficulty target, encoded compactly
• Nonce — the 32-bit number the miner brute-forced to make the hash satisfy the target

The merkle root is the elegant part: change any single transaction byte, and the root changes, the header hash changes, and the proof-of-work is invalidated. The whole body of a block is committed to in those 32 bytes.

Block intervals are not regular — and that is fine

The protocol targets a 10-minute average interval, enforced by difficulty adjustment every 2,016 blocks. But the actual time between blocks is geometrically distributed — memoryless. Sometimes two blocks arrive within 60 seconds of each other. Sometimes you wait 40 minutes. This is not a bug. It is the statistical signature of a Poisson-like process where every hash attempt is independent.

People new to Bitcoin worry when they see a 30-minute gap. They should not. Roughly 5% of the time, the gap exceeds half an hour. The next block does not “owe” us speed to make up. Each block starts the clock fresh.

The tip, forks, and the longest chain

The tip is the most recently accepted block — the current end of the chain. Most of the time, the network agrees on the tip within seconds of propagation. Occasionally two miners find valid blocks at nearly the same height and the network briefly splits — a fork. It almost always resolves within one or two blocks: one branch extends further, the longer chain wins, and transactions in the orphaned block return to the mempool. No funds are lost.

Miners, pools, and who actually finds blocks

The “miner” labeled on a block — Foundry, AntPool, ViaBTC, F2Pool — is almost always a pool. A pool aggregates hashpower from thousands of independent operators running actual mining hardware, then distributes the reward proportionally. The pool name is the entity that coordinated the work and built the block template, not a single company that turned on a warehouse. Pool concentration is a real centralization concern worth tracking, but it sits one layer above the physical machines.

Now and then a block contains only its coinbase transaction — an empty block. The mundane explanation: when a miner has just learned about a new tip, they often start hashing on the previous header before their full transaction list is assembled, to avoid wasting hashpower. As soon as the template is ready, they switch. Small inefficiency, not a flaw.

Block rewards, today

A block reward is the subsidy plus the sum of transaction fees. Today the subsidy is 3.125 BTC, after the April 2024 halving. Fees fluctuate with mempool demand — typically 0.05 to 0.5 BTC per block, more during congestion. The subsidy halves again around 2028, and every four years thereafter, until it rounds to zero around 2140. After that, miners earn fees only — which is why long-term fee economics matter so much for Bitcoin’s security model.

Chain size and what your node stores

At the end of 2025 the full blockchain is around 600 GB, growing roughly 50 to 70 GB per year. A full node stores every byte and verifies every signature from genesis. If you lack the disk space, you can run a pruned node — discard old block data after verifying it, and live on around 20 GB. Pruning loses none of Bitcoin’s security guarantees for your own use; you just cannot serve historical blocks to peers. Bitcoin Core’s docs cover the trade-offs.

Why ten minutes

Why not one minute, or one hour? Ten minutes is a deliberate compromise. Faster intervals mean shorter confirmation waits but more forks, more orphans, and more propagation pressure. Slower means cleaner consensus but agonizing waits. Satoshi’s choice in the whitepaper — defended by the difficulty adjustment — has been remarkably durable. After fifteen years and many alternatives, ten minutes remains the right answer.

Reading the data on this page

What you see live is the current tip height, the most recent block intervals, and a running average. If the average drifts well above 600 seconds, hashpower has dropped — the next retarget will lower difficulty. If well below, hashpower has grown and difficulty will rise. The system self-corrects every 2,016 blocks, on schedule.

Run a node, verify the chain yourself

Everything on this page is summarized from APIs I trust — but the only way to know the chain is to verify it. A modern full node syncs from genesis in 12 to 24 hours, validates every block against consensus rules, and from then on accepts only what is valid. No trust in me, in mempool.space, or in any block explorer required. That is what don’t trust, verify actually means. Read Mastering Bitcoin, run your own node, and the heartbeat on this page becomes something you witness directly — ten minutes at a time.