The Usual Confusion
Most people encounter transaction fees the same way: they're about to send some crypto, and a number appears—sometimes small, sometimes larger than they expected—that nobody explained in advance. The instinct is to treat it like a bank wire fee or a PayPal surcharge: an arbitrary charge someone decided to impose.
That framing leads to real confusion. Why does sending $5 of crypto sometimes cost $20? Why do fees spike at random times? Why can't the network just eliminate them or cap them at something reasonable?
The confusion comes from assuming fees are a pricing decision. They're not. They're an emergent property of a specific structural constraint: public blockchains have finite capacity, but anyone in the world can submit a transaction at any time. Something has to decide whose transaction gets processed first. Transaction fees are that mechanism—how blockchains ration scarce block space through voluntary market signals, and how networks compensate the participants who actually do the work of processing transactions.
How the Mechanism Actually Works
Every block on a blockchain can hold a limited number of transactions. Bitcoin's blocks are capped at roughly 1-4MB of data depending on transaction types. Ethereum's base layer processes around 15-30 transactions per second. Solana handles significantly more but still operates within limits. These aren't arbitrary restrictions—they reflect deliberate trade-offs between throughput, decentralization, and the hardware requirements for running a node. Push blocks too large, and only well-resourced participants can run full nodes, which undermines decentralization.
Because blocks are finite but submission is open, the network frequently receives more transactions than it can immediately process. Those pending transactions sit in the mempool—a shared waiting room where validators and miners can see every candidate transaction.
Fee markets emerged naturally from this dynamic. Users who want their transaction processed quickly attach higher fees. Validators and miners, acting rationally, sequence transactions by fee rate—taking the most profitable ones first. A user paying a low fee might wait through several blocks before their transaction gets picked up. One paying a high fee often lands in the very next block.
Ethereum restructured this model in 2021 with EIP-1559. The upgrade split fees into two parts: a base fee (algorithmically determined by the previous block's fullness, burned and removed from ETH supply) and a priority tip (paid directly to validators as an incentive to include the transaction). When blocks are full, the base fee rises. When they're empty, it falls. The tip lets users signal urgency on top of that baseline. This made fee estimation more legible—though it didn't eliminate fee spikes. It just made the underlying mechanism more transparent.
There are two functions fees serve that often go unmentioned.
The first is spam prevention. Without fees, submitting millions of transactions costs nothing. A permissionless network with zero-cost transactions is trivially attacked—any entity could flood the mempool indefinitely, paralyzing the network at no financial risk to themselves. The fee creates friction that makes mass transaction spam economically irrational.
The second is infrastructure compensation. Miners and validators spend real resources: electricity, hardware, staked capital exposed to slashing risk. Transaction fees are part of how they get paid. As block subsidies (newly issued coins awarded per block) decline over time—Bitcoin's halving mechanism cuts this reward every four years—fee revenue becomes progressively more important to long-term network security.
Where the Constraints Live
The fee mechanism depends on block capacity being genuinely constrained. If blocks had infinite capacity, fee markets would collapse—there'd be no meaningful competition for space. The constraint is partly technical (limits on what nodes can process and propagate before the next block) and partly a deliberate design choice.
Bitcoin's block size was a years-long political and technical debate. Keeping it small was a choice that preserves decentralization at the cost of throughput. Layer 2 networks exist specifically to work around this constraint without changing the base layer limit.
What's Changing
Three structural shifts are in motion.
EIP-1559's burn mechanism means Ethereum's fee dynamics now touch monetary policy. High network activity burns more ETH, reducing circulating supply. Low activity reverses this. The interaction between transaction volume and ETH issuance is now a live feedback loop.
Layer 2 networks are changing base layer fee pressure. As more activity migrates to rollups (Arbitrum, Optimism, Base, and others), Ethereum L1 transaction counts decline, which reduces fee competition on the base layer. EIP-4844, introduced in 2024, created a separate blob data market specifically for rollups to post data cheaply—further separating L2 user costs from L1 gas fees.
Bitcoin's fee market is maturing into something more consequential. As block subsidies continue declining through successive halvings, miner revenue will depend increasingly on transaction fees. Whether Bitcoin's fee market can generate sufficient revenue to sustain long-run network security is an open question, not a settled one.
What Would Confirm This Direction
Watch the ratio of fee revenue to block subsidy in Bitcoin—if that ratio rises consistently across halving cycles, it supports the thesis that fee markets can replace subsidy as the primary miner incentive. On Ethereum, sustained high L2 adoption with L1 fees remaining lower than 2021-era peaks would confirm that the scaling architecture is working as intended.
What Would Break or Invalidate It
The spam-prevention argument weakens if networks develop alternative Sybil resistance mechanisms that don't require monetary cost—none have been deployed at meaningful scale, but the possibility exists. The infrastructure compensation thesis breaks down if MEV (maximal extractable value) extraction becomes so dominant that base transaction fees become economically irrelevant to validators. Bitcoin's fee market thesis breaks if transaction demand on the base layer remains structurally low after each halving, leaving miners with insufficient revenue to justify continued operation.
Timing Perspective
Now: Fee markets are live and functioning. L1 fees reflect real-time demand. L2 fees are structurally lower and represent the practical option for most users doing routine transactions.
Next: Bitcoin's fee sustainability question becomes more empirical with each halving. The 2024 halving was a data point; the 2028 halving will add another.
Later: Whether advances in zero-knowledge proofs, data availability layers, and alternative scaling approaches materially change base layer fee dynamics will become clearer over a 3-5 year horizon.
What This Post Doesn't Cover
This explains why transaction fees exist as a mechanism—not what fees you should expect to pay, which chain has the lowest fees for a given use case, or whether fee structures across networks are fair. Fee levels vary by chain, by network congestion, and by transaction type. The existence and logic of fees is distinct from any claim about optimal fee design.
