Launch budgets buy a mainnet. Almost nobody budgets the decade after — and the gap between those two numbers is where chains die. In October 2025 the industry got a clean public demonstration, and Ethereum's own maintenance economics put numbers on what “keeping a chain alive” actually costs. This post is the taxonomy, from public data only.
- Kadena — a top-tier L1 of the 2021 cycle, $4B peak — announced in October 2025 that its foundation was ceasing “all business activity and active maintenance.” The chain keeps producing blocks; the maintenance layer is what shut down.
- Ethereum funds client maintenance socially: Protocol Guild has distributed $33M+ since May 2022 across ~30 teams and 181 contributors — median member ~$66k over the last 12 months. Public estimates put the sustainable need near $30M per year.
- Post-genesis costs split into five lines: client upgrades & hard forks, validator/node operations, RPC + indexer infrastructure, security response, and tooling rot — and they scale differently.
- The structural mismatch: maintenance is perpetual opex, budgeted against a token treasury that was priced for capex at launch and repriced downward since. Kadena's token was down over 99% from its 2021 high when the foundation folded.
- When an in-house team shrinks under this load, the pattern is always the same sequence: upgrade cadence slips first, pinned client forks accumulate, indexers rot, and then the chain faces a takeover-or-sunset decision.
Protocol Guild — membership & funding data
The Defiant — Protocol Guild distributions
What a running chain actually costs
Five cost lines survive genesis. They are not equal, and they do not scale alike.
| Cost line | What it covers | How it scales |
|---|---|---|
| Client upgrades & hard forks | tracking upstream, consensus changes, emergency patches | with ecosystem velocity — faster upstream, higher bill |
| Validator & node operations | infra, monitoring, incident response, key ceremonies | roughly linear with node count and SLA |
| RPC & indexer infrastructure | public endpoints, explorers, archive nodes, data pipelines | with state growth — the quiet compounder |
| Security response | disclosure handling, war rooms, audits of every change | step function — cheap until the day it is everything |
| Tooling rot | SDKs, wallets, docs, CI decaying against moving dependencies | with time, independent of usage |
Launch budgets fund none of these past year one. They fund the genesis event, the audit before it, and the marketing around it — capex. Everything in the table is opex, and it runs whether or not anyone is using the chain.
The public evidence
Kadena is the clean case because the chain did not die — the maintenance did. On 21 October 2025 the foundation announced it was “ceasing all business activity and active maintenance of the Kadena blockchain,” citing market conditions. Blocks kept coming: the miner network runs on, and a small transition team shipped one upgraded node binary so operators could continue unattended. A $4B-peak L1 reduced its entire maintenance layer to a farewell binary. The token fell more than 60% on the announcement and sat over 99% below its 2021 high — which is the budget story, not a separate story.
Ethereum shows the other end: maintenance funded as a social institution. Protocol Guild — 181 contributors across ~30 client and research teams — has distributed over $33M since May 2022, with another $33M vesting over four years, funded by 1%-of-supply pledges from projects like ether.fi, Taiko, and EigenLayer. The median member received about $66k over the last twelve months — competent-engineer money, not mercenary money — and public analyses frame the sustainable need at roughly $30M per year for the client ecosystem. The most valuable chain in the industry treats client maintenance as a permanent public-goods bill. Chains without Ethereum's gravity have the same bill and none of the pledge base.
Capex treasuries, opex realities
The mismatch is financial, not technical. A launch treasury is denominated in the chain's own token and sized against launch-cycle prices; maintenance is a perpetual expense that arrives in engineer-salaries and cloud invoices. When the token reprices down 90%+ — the ordinary fate of an altcycle L1 — the treasury's purchasing power evaporates exactly when the maintenance bill is at its steadiest. Foundations then face the sequence Kadena made public: cut ecosystem grants, cut the core team, cease active maintenance. The chain's liveness outlasts its stewardship, which is the most dangerous configuration — users still aboard, nobody on watch.
The pattern when teams shrink
From the takeover side of the table, the decay has a recognizable order. Upgrade cadence slips first — upstream clients move, the fork stays pinned. Then the pinned fork accumulates patches nobody upstreams. Then the indexers and RPC layer rot as state grows past what the original sizing assumed. By the time the symptom is user-visible — explorers lagging, RPC flaking, a security advisory with no one to receive it — the fix is no longer a hire; it is a handover. The economic decision at that fork is honest: either the chain's remaining value justifies a professional maintenance takeover, or it justifies a planned sunset. Both are respectable. Drift is the only wrong answer.
Limits
Three caveats on the data. Public cases skew visible-failure — chains that quietly fund maintenance well don't make headlines, so the sample is survivorship-inverted. Foundation budgets are coarse: grants, salaries, and infrastructure blur inside annual reports, so per-line numbers above stay qualitative by design. And per-stack variance is real — a Cosmos-SDK chain, an EVM L1, and an L3 on shared infrastructure carry very different versions of the same five lines. The taxonomy travels; the invoice amounts do not.
If you're working on something similar, we'd like to hear about it. Talk to a founder — 30 minutes.