The cost of sending a blockchain transaction can vary enormously from one hour to the next, and it bears no relation to the amount being sent. Fees are the outcome of an auction for space.
Block space is a hard constraint
Each block can hold only a limited quantity of transaction data, and blocks arrive at a controlled pace. The supply of space per hour is therefore effectively fixed.
Demand is not fixed. When many users want to transact at once, the pending pool grows beyond what the next several blocks can accommodate, and something has to determine which transactions go first.
That something is the fee attached to each transaction, which converts a queue into a market where position is bought rather than earned by arrival time.
The fee is priced by data, not value
Fees are calculated against how much space or computation a transaction consumes, so a transfer of an enormous sum can cost less than a small one that is structurally complex.
A payment consolidating many small prior receipts occupies more space than a simple transfer, and interacting with a contract consumes computation that must be paid for regardless of the amounts involved.
This is why wallets that quietly accumulate many tiny receipts become expensive to spend from, a cost that appears only when the holder tries to move the funds.
Congestion is driven by events
Fee spikes cluster around moments of intense activity, such as a sudden price move that triggers a wave of exchange deposits, or a heavily anticipated launch that draws simultaneous participation.
Because participants in those moments are competing for a narrow window, they bid aggressively, and ordinary users sending routine payments are priced out for the duration.
Once the event passes, the pending pool drains within hours and fees fall back, which is why patience is often the cheapest available strategy.
Fee design differs across networks
Some networks use a pure auction where users guess what will clear, which tends to produce overpayment and erratic pricing during busy periods.
Others set a base fee algorithmically from recent demand and destroy it, with users adding only a small tip for priority. That approach makes fees more predictable and links supply pressure to network usage.
The mechanism chosen affects both user experience and the coin's issuance, since burning fees offsets new supply while paying them to producers does not.
Layered systems move demand elsewhere
Secondary networks batch many transactions and record only a compressed summary on the base chain, spreading one block-space cost across a large number of payments.
This lowers per-transaction cost substantially while still inheriting settlement assurances from the underlying chain, at the price of added complexity and new assumptions about who can be trusted.
Base-layer fees still matter to these systems, because the cost of posting their summaries rises and falls with the same auction everyone else is bidding in.