EIP-8254 Per-Block Deposit Request Limit (Cap Deposit Requests Per Block)

No call decision · Consensus & staking · EIP status Draft · created 2026-05-06

Limits the number of validator deposit requests that can be included in a single execution layer block to 8192. A block exceeding this count is invalid.

Why it is needed

To become a validator, one sends ETH to the deposit contract, and this deposit request is read directly from the block under EIP-6110 and forwarded to the consensus layer. However, the Execution Layer has no per-block limit on deposit requests, and the only limit is the Consensus Layer's list limit of 8192. As the gas limit rises toward the 193-200M range, a batching contract could be used to pack more than 8192 deposits into a single block. In that case the execution layer produces the block without any problem, but the consensus layer cannot interpret that data and block proposal fails. Subsequent proposals keep failing for the same reason until the backlog of deposits clears, potentially halting chain progress. To prevent this latent flaw, the same limit is added to the execution layer as a consensus rule.

What changes

Today

After EIP-8254

Who is affected

In depth

New block validity rule and constant

This introduces the constant MAX_DEPOSIT_REQUESTS_PER_BLOCK = 8192. The block's deposit-request count is the number of DepositEvent logs emitted by DEPOSIT_CONTRACT_ADDRESS that match DEPOSIT_EVENT_SIGNATURE_HASH, counted across all of the block's receipts, and under EIP-6110 encoding equals len(get_deposit_request_data(block.receipts)) // 192 (192 bytes per entry). From FORK_BLOCK onward, any block whose count exceeds 8192 is invalid and must be rejected at the same level as any other consensus rule violation. Descendant blocks built on top of such a block are also invalid.

Mandatory enforcement at the block construction stage

The limit must be respected not only during validation but also during block construction. Across all block construction paths, including engine_getPayload and local mining/validation, clients MUST track the deposit-request count of the block being assembled and MUST NOT include any transaction that would push the count past the limit. Subsequent transactions that do not generate a deposit can continue to be included under the normal rules. Excluded deposit transactions remain eligible for inclusion in the next block. The same applies to the fallback path used when no external builder is available, and enforcement of this limit must not be delegated to the builder. If only validation is enforced, an honest builder unaware of the limit would produce an invalid block and miss its slot.

Engine API and the consensus layer

Engine API responses follow the validity rule above directly. engine_getPayloadV* must not return a payload that violates the limit, and engine_newPayloadV* must report such a payload as INVALID. There are no consensus layer changes. The existing SSZ limit MAX_DEPOSIT_REQUESTS_PER_PAYLOAD = 8192 remains unchanged, and because the execution layer limit matches it exactly, the execution layer can no longer produce a deposit list that the consensus layer fails to decode.

Why the execution layer, and why 8192

The alternative of raising the consensus layer's SSZ limit was rejected. The consensus layer's processing cost is dominated by roughly one BLS signature verification per deposit, so raising the limit would only raise the ceiling on worst-case processing time without bounding it, and the same problem would recur every time the gas limit rises. Since deposit list size is inherently a property of execution layer execution, it is natural to place the limit at the layer that constructs the list. 8192 = 2^13 matches the existing SSZ limit exactly. Up to 8192 is always decodable, 8193 is the first value that breaks, only a single value needs to be managed across both layers, and since EIP-6110 already analyzed 8192 as the upper bound, no security re-analysis is needed. Lower values such as 8191 were rejected because they would only create a gap with no functional benefit while adding a second constant. A count-based criterion was used instead of a gas-reservation approach because it is independent of changes to deposit gas costs and matches the consensus layer's cost unit (number of signature verifications).

Demonstrated attack scenario and liveness

It has been demonstrated that at a 193-200M gas limit, a batching contract can pack roughly 410 deposits per 10M-gas transaction, filling a 200M-gas block with about 8200 deposits. The failure observed on Lighthouse was a decoding error, "VariableList of 12300 items exceeds maximum of 8192," and since the SSZ limit is common to all consensus clients, this is not an issue specific to one client. Leaving it to the builder is also not safe, since an attacker can flood deposit transactions through the public mempool while simultaneously triggering the builder's circuit breaker, forcing a fallback to local block construction that is unaware of the limit. The reference attack scenario in the test cases is spamoor's deposit-contract-batch configuration, which reliably reproduces more than 8192 deposits per block at around 193M gas.

The cost of builder packing becoming a two-dimensional knapsack problem

Without the limit, block construction is a one-dimensional problem of maximizing fees under the single constraint that total gas stay under the limit, and a greedy strategy based on fee per gas is close to optimal. Adding the limit turns it into a two-dimensional knapsack problem, where the sum of deposit counts must also stay at or below 8192, which is NP-hard in general. Builders are expected to use heuristics such as reserving a budget of deposit slots, which is not optimal and can be exploited. For example, a transaction that looks attractive by fee per gas but generates many deposits can be used to grief the builder by exhausting its budget early, crowding out more profitable deposit transactions. However, filling the limit requires at least 8192 x 1 ETH ~= 8192 ETH at the 1 ETH minimum deposit, which acts as a deterrent on cost. This added complexity was accepted as the price of the liveness guarantee.

Backwards compatibility, data size, and relationship to prior analysis

Since this is a new block validity rule, it requires a hard fork and a coordinated client upgrade. The limit cannot be reached at the current gas limit. By EIP-6110's analysis, a 30M-gas block holds about 1,271 deposits, and even a 200M-gas block only reaches the limit through adversarial batching, so no past block is affected. Applications that consume deposit data are also unaffected, since the value stays below the SSZ limit they already expect. A block's deposit list filled to the limit is 8192 x 192 bytes, about 1.5 MB, which is within the scope of EIP-6110's existing analysis. The worst-case bound for optimistic sync and the weak subjectivity analysis also rely on the same value derived from the existing SSZ limit, so their conclusions are unchanged. The highest daily deposit volume on record, about 12,000 (roughly 2 per block on average) as cited by EIP-6110, leaves the 8192-per-block limit with substantial headroom over normal activity.

Progress status

This EIP is in Draft status, created on May 6, 2026. It is still in an early review stage, so details of the specification may change. No decision to include it in a specific fork has been confirmed in core developer calls so far.

Original EIP · ← EIPs