Slot ID: blog-post-top
At 3am on a Tuesday, moving ETH across the network costs about 40 cents. By lunchtime the identical transfer can cost $9. Same wallet, same amount, same 21,000 gas. The only thing that changed is how many other people wanted a slot in the next block.
That gap is the whole story of Ethereum gas fees, and it explains why the network so often feels like it's nickel-and-diming you. Fees aren't a fixed toll. They're the clearing price of a queue that resets every 12 seconds.
Every transaction carries a computational price tag measured in gas. Those unit counts are fixed by the protocol and don't move:
The number that moves is the price per unit, quoted in gwei, which is one-billionth of an ETH. At 10 gwei with ETH at $3,000, a 21,000-gas transfer costs 0.00021 ETH, about 63 cents. Push the same transfer through at 120 gwei during a hyped NFT mint and you're staring at $7.56. The job didn't get bigger. The bidding did.
Since the London upgrade in August 2021, every fee splits into two parts. The base fee is set by the protocol based on how full the last block was, and it gets burned, permanently removed from supply. The priority fee is your tip to the validator who includes you. Base fee buys the ticket; the tip decides how fast you get through the gate.
Ethereum aims for blocks that are about half full. When demand pushes past that target, the base fee climbs by as much as 12.5% per block, and a block lasts 12 seconds. Sustain that for a few minutes and fees multiply tenfold, then collapse just as fast once a mint sells out or a claim window shuts.
Roughly 30 million gas fits in a block. When 50,000 people want in during the same 12 seconds, the queue clears by price, not by fairness. The people who absolutely need inclusion that block pay for it. Everyone else waits, which is why an 8-minute delay is often worth more than a 5x tip.
The Dencun upgrade in March 2024 introduced blobspace, cheap temporary data that rollups use to post batches back to Ethereum. Fees on Arbitrum, Base and Optimism dropped from roughly $0.50 to $1.00 per swap down to fractions of a cent within days. If your activity is swapping, lending or trading NFTs, the honest question is no longer how to beat mainnet fees. It's why you're still on mainnet at all.
A 20 gwei transaction at ETH $1,800 costs half what the same 20 gwei costs at $4,000, with zero change in network demand. Before blaming congestion for an expensive week, check what the Ethereum exchange rate did. Wallets always show you the dollar figure and almost never show the gwei behind it, so it's worth knowing which half of the number moved.
The visible cost is the swap you signed. The invisible one is everything around it. A failed transaction still burns gas, and reverts are common on fast-moving pools. Approvals stack up because each protocol wants its own permission, so people end up approving the same token five times. Bridging to a rollup and back can cost more than the position you're moving.
Staking has its own transaction bill, which is part of why staking ETH for that headline 3% yield deserves a closer look than the APR suggests. Add everything up over a year of active trading and fees regularly land between 2% and 8% of the capital deployed. That's not a rounding error. It's the difference between beating the market and donating to it.
Every base fee is destroyed, and on busy days that's millions of dollars of ETH removed from circulation. It's a real mechanism, though it's easy to overweight. Burned fees matter far less than issuance changes or demand shifts, which is roughly the argument in what actually drives ETH and why most price targets miss. High fees are a symptom of demand for blockspace. They aren't a price catalyst on their own.
Open a gas tracker and look at the base fee for the last ten blocks. Rising base fee plus a hyped event means wait. Flat or falling means sign now, because your transaction will likely clear below your max fee and the difference gets refunded.
Then ask two questions. Does this action need mainnet, or will a rollup do the same job for a hundredth of the cost? And is the fee small relative to the position, or is it a meaningful slice of it? A $6 fee on a $40,000 transfer is noise. The same $6 on a $70 trade is a bad trade, no matter how good the token looks. That ratio is the number worth carrying around, and it matters just as much when you're closing out, which is where the details of selling Ethereum without leaving money on the table start to compound.
Slot ID: blog-post-bottom