EIP-8311 Raising the calldata floor cost to 96 (Increase Calldata Floor Cost to 96)
No call decision · Fees & gas · EIP status Draft · created 2026-06-20
Raises the calldata floor cost applied to data-heavy transactions from 64 gas per byte to 96 gas per byte. This creates room to raise the gas limit by aligning the byte density of calldata-heavy blocks with the ETH-transfer block baseline.
Why it is needed
The gas limit of an Ethereum block is tied to the volume of data the network can handle. When setting the limit, the worst-case block that packs the most bytes for the same gas must be used as the reference. A block filled entirely with ETH transfers is charged 21,000 gas per transfer, so it cannot be made more expensive and serves as a fixed baseline. Right now, however, a block filled entirely with calldata packs about 1.49 times more bytes per gas than that baseline. So even if we want to raise the gas limit, it must be kept low enough that the worst-case calldata block does not exceed the network's bandwidth. Raising the calldata floor cost to 96 brings the density of calldata blocks down to or below the level of transfer blocks, removing calldata as a bottleneck and creating room to raise the gas limit. Ordinary user transactions rarely hit the floor cost, so the impact on them is minimal.
What changes
Today
- The calldata floor cost for data-heavy transactions is 64 gas per byte
- A block filled entirely with calldata packs about 1.49 times more bytes per gas than an ETH-transfer block
- This worst case forces the block gas limit to be kept low
After EIP-8311
- The calldata floor cost rises to 96 gas per byte
- The byte density of calldata blocks drops to at or slightly below the transfer-block baseline
- Calldata is removed as the bottleneck constraining the gas limit, creating room to raise it
- Ordinary transactions with heavy EVM computation keep the existing 4/16 gas pricing
Who is affected
- Ordinary users: Ordinary transactions with heavy computation, such as DeFi interactions or token transfers, do not hit the floor cost, so their fees stay the same.
- Wallet and node developers: eth_estimateGas must be updated to reflect the new value of 24. Failing to update it can cause gas to be underestimated and transactions to fail.
- Services that post data via calldata: The cost of data-heavy transactions rises from 64 to 96 gas per byte, creating a stronger incentive to move to blobs.
- The whole network: The worst-case bandwidth scenario is lowered to the level of transfer blocks, creating room to raise the block gas limit.
In depth
Parameters that change
STANDARD_TOKEN_COST stays at 4, while TOTAL_COST_FLOOR_PER_TOKEN rises from 16 to 24. Since floor_tokens_in_calldata = (zero_bytes_in_calldata + nonzero_bytes_in_calldata) * 4, one byte of calldata is 4 tokens, and the floor cost becomes TOTAL_COST_FLOOR_PER_TOKEN * 4 = 96 gas. The same 96 gas applies equally to zero and non-zero bytes. The token abstraction is retained only for consistency with EIP-7623 and EIP-7976, and aside from this single value, EIP-7976's gasUsed formula and transaction validity rules remain unchanged.
Deriving the anchor density
A 21,000-gas ETH transfer is the reference point that pins down both the execution and bandwidth sides of a block. A block filled entirely with transfers is itself a data payload: a single transfer's envelope and signature contribute a worst case of about 130 bytes, and its worst-case marginal contribution to the block access list is about 91 bytes. Combined, this is roughly 221 bytes per transfer carried in 21,000 gas, giving a density of about 0.0105 bytes/gas. Unless the cost of a transfer is raised above 21,000 gas, this density cannot be lowered further, making it the baseline for the maximum bytes-per-gas the protocol must be able to handle.
Why 96
At the floor value of 64 set by EIP-7976, a pure-calldata block has a density of 1/64 ≈ 0.0156 bytes/gas, about 1.49 times that of a transfer block. The smallest integer floor value that closes this gap is F* = ceil(1 / 0.01052) = ceil(95.06) = 96. At 96, the density is 1/96 ≈ 0.0104 bytes/gas, at or slightly below the anchor density. In other words, 96 is not an arbitrary number but the smallest integer such that 1/F does not exceed the transfer-block density of about 0.0105 bytes/gas, without overcharging beyond what the anchor requires.
Effect in absolute cost terms
Carrying an uncompressed 10 MiB execution payload as calldata costs about 105 million gas at EIP-7623's 10/40 pricing, about 671 million gas at EIP-7976's 64/64, and about 1.01 billion gas at this proposal's 96/96. Since the floor value only applies to data-heavy transactions, transactions with heavy EVM computation continue to pay the standard 4/16 pricing, while demand for data-availability use cases faces an even stronger incentive to move to blobs.
What wallets and nodes need to do
A non-backwards-compatible gas repricing that requires a scheduled network upgrade. Wallets using eth_estimateGas must accurately reflect the change of TOTAL_COST_FLOOR_PER_TOKEN to 24; failing to do so can cause gas to be underestimated and transactions to fail. Node software RPC methods such as eth_estimateGas must also compute gas using the formula with the new floor value. End users can continue using their existing workflow once wallets and RPCs are updated.
Limits of security and scope
Since the maximum block size shrinks even further than under EIP-7976, there are no security concerns beyond those already addressed in that proposal. Bundling multiple transactions into one was possible before and remains possible now, but it does not undermine the goal of reducing block size, and practical constraints such as trust and coordination costs remain. Also, this EIP only addresses the pure-calldata case. Combinations that mix cheap standard-rate calldata with byte-dense state access do not fall to the anchor level through a floor increase alone, and are left for a separate proposal.
Progress status
This EIP is in Draft status, so its figures and design may still change. It assumes EIP-3860, EIP-7623, and EIP-7976, and no core developer call decision has yet been made on which network upgrade it would be included in.