ckUSDT’s Archive Gets Capacity Before It Needs It

The Internet Computer’s ckUSDT archive is being provisioned for a future workload, not an immediate one. NNS proposal 144040 proposes changing canister yhujl-liaaa-aaaar-qaiha-cai from a zero-byte, best-effort memory allocation to 2,147,483,648 bytes—2 GiB.

That distinction matters operationally. ICP documentation describes memory_allocation as a pre-allocated pool: memory growth within the allocation is covered by the reservation, while growth beyond it can still depend on available subnet capacity. With a best-effort setting, growth on a heavily used subnet can instead require additional reserved cycles and may fail when the canister’s limit is exhausted.

The proposal says the ckUSDT ledger is not currently archiving blocks, so the archive is not growing today. The allocation is consequently a readiness measure for re-enabling archiving. The canister reportedly uses 196 MiB, leaving about 1.81 GiB of headroom. The proposal also says the 2 GiB target mirrors the allocation already assigned to the ckBTC archive.

The safety margin is deliberately temporary. A full archive node can require more than 2 GiB: the proposal cites 3 GiB for block data alone through node_max_memory_size_bytes. That means the change does not establish a final capacity target. It establishes a predictable starting point that will need review if archiving resumes and the archive approaches its ceiling.

There is also a cost trade-off. The proposal estimates roughly 52.4 trillion cycles per year for the full 2 GiB on the 34-node fiduciary subnet, while saying the current subnet position would require no immediate reservation charge for making the change. Official ICP documentation confirms that memory allocation is rented whether or not the allocated capacity is fully used, and that storage-reservation behavior depends on subnet utilization. The proposal’s annual figure should therefore be read as a proposal-time estimate, not a permanent tariff.

The broader Chain Fusion lesson is that archive reliability is being handled as capacity planning. The governance action does not restart ckUSDT archiving or prove that 2 GiB will be sufficient. It removes one predictable failure mode before historical data growth returns, while leaving the larger question—how much capacity a live archive ultimately needs—for a later operational decision.

Caveat: the live dashboard page was not rendered by the search tool, so the proposal-specific figures and current archiving status above are reported from the newsroom-fetched primary material. The annual cycle figure is also an estimate and can change with subnet conditions and pricing parameters.