You're watching a popular NFT mint tear through Ethereum. Gas prices are climbing by the second, your transaction is stuck in the mempool (the waiting room for unconfirmed transactions), and somewhere in the back of your mind a question forms that has nothing to do with whether you'll get the mint: where is all this fee money actually going? Since August 2021, a meaningful chunk of it has been going nowhere. It gets destroyed.

That's the core of EIP-1559, and the supply implications compound in ways that aren't obvious until you run the numbers.

The Mechanism Behind the Burn

Before EIP-1559, every fee a user paid went entirely to the miner (now validator, post-Merge) who included the transaction. Simple auction: you bid, they pocket it. The problem was wild fee volatility and a perverse incentive for validators to artificially congest the network.

EIP-1559 split the fee into two parts. The base fee, a protocol-set price per unit of gas that every user in the same block pays equally, and the priority fee (the tip), a voluntary top-up paid directly to the validator as an inclusion incentive. The base fee does not go to the validator. It gets burned, meaning that ETH is permanently removed from circulation.

The base fee isn't static, either. The protocol targets 50% block utilization. If a block exceeds that, the next block's base fee rises by up to 12.5%. Below 50%, it falls by up to 12.5%. This is an exponential curve, not a linear one, and that distinction matters enormously. Ten consecutive 100%-full blocks would multiply the base fee by roughly 3.3x from its starting point. That multiplication drives how much ETH gets destroyed per unit of time.

What Sustained Congestion Actually Does to Supply

A worked scenario illustrates the stakes. Imagine Ethereum runs at consistent 80% block capacity for a prolonged stretch, the kind of thing that happens during DeFi liquidation cascades or major token launches. At 80% utilization, the base fee rises every block. Burn accelerates. Meanwhile, validators issue new ETH as staking rewards at a rate tied to total ETH staked: roughly 1,700 new ETH per day when around 30 million ETH is staked (the exact figure shifts with stake size, but the order of magnitude is well-established).

During a congestion spike where the base fee climbs to 80 gwei and the network processes around 15 transactions per second, the burn rate can exceed issuance. Net supply contracts. ETH becomes, in that window, a deflationary asset.

Then congestion eases. The base fee falls. Issuance outpaces burn again.

The long-run supply trajectory isn't a straight line in either direction. It's a sawtooth pattern that tilts toward deflation precisely when the network is most in demand. That's not an accident. It's an emergent property of the design, and it's more elegant than most people give it credit for.

Two Stakers, One Outcome Neither Expected

Consider Priya and Marcus, who both started staking 32 ETH on the same day. Priya has followed the issuance numbers closely and expects slow, steady accumulation. Marcus read that EIP-1559 made ETH "deflationary" and assumed his holdings would always be appreciating in real ETH terms.

Neither is right. During a quiet six-month stretch with low network activity, both see ETH supply grow modestly: issuance outpaces the relatively small burn, and their percentage of total supply dilutes slightly. Marcus is surprised. During the following three months of intense congestion, the burn rate flips the equation and total supply contracts. Priya is surprised in the other direction.

The actual outcome depends entirely on network usage. Staking rewards are nearly fixed, varying slowly with total stake. The burn is variable and usage-driven. That asymmetry is what makes EIP-1559's supply effect genuinely different from any simple "X% inflation" model, and it's why anyone quoting you a single long-run ETH supply figure deserves skepticism.

The Ceiling Nobody Talks About

One honest complication: the base fee burn has a hard ceiling, because the block gas limit is fixed (30 million gas, with a 15 million gas target). No matter how high demand climbs, you cannot burn more than one full block's worth of base fee per block. Physical throughput caps the destruction.

This is where the long-term picture gets genuinely uncertain, and I think most commentary glosses over it too quickly. If Ethereum scales primarily through Layer 2 rollups, as its current roadmap intends, most transaction activity moves off the base layer. Users on Arbitrum or Optimism pay fees on L2, with only occasional settlement transactions touching L1. Those L1 settlements do burn ETH, but far less than if every swap and transfer happened on mainnet. A highly successful L2 ecosystem could, counterintuitively, reduce the base-layer burn rate even as overall Ethereum usage grows.

The opposing force is EIP-4844, also called Proto-Danksharding. It introduced a separate fee market for blob data (chunks of temporary data that rollups use to post transaction records cheaply to L1), and those blob fees are also burned. L2 scaling doesn't entirely escape the burn mechanic. It redirects it into a new fee type, like water finding a different pipe rather than disappearing.

Are the long-term effects of this three-way interaction between staking issuance, base-fee burn, and blob-fee burn fully modeled? No. The on-chain data site ultrasound.money tracks cumulative burn in real time and is worth bookmarking if you want to watch the numbers move. But projections beyond a few years require assumptions about usage patterns that nobody can pin down.

The Honest Caveat

A lot of commentary treats EIP-1559 as a guaranteed deflationary engine. It isn't. It's a conditionally deflationary mechanism, the condition being sustained high demand. If Ethereum usage plateaus, or if L2 migration siphons enough activity from the base layer, issuance can comfortably outpace burn for extended periods. Supply would grow, slowly.

So what did EIP-1559 actually accomplish, structurally and permanently? It tied monetary policy to product demand. ETH's supply trajectory now has a direct feedback loop with how much people want to use the network. That's a genuinely novel monetary design, and it's one I find more intellectually defensible than fixed-schedule issuance, because it doesn't pretend that demand is irrelevant to money supply.

Whether it's better depends on what you think Ethereum's usage looks like over the next decade. Anyone claiming certainty there is selling something. Check the current issuance-versus-burn ratio yourself before forming a strong opinion: the data is public, updated by the block, and more informative than any pundit's projection, including this one.