What is Built, and One Open Question
News
Lava Foundation
Aug 6, 2026
What is Built, and One Open Question
News
Lava Foundation
Aug 6, 2026
Why we are publishing this?
Lava Foundation is opening a question for the community to decide: should Lava continue to run a chain of its own?
This post follows a review the Foundation completed in July, looking at the infrastructure any Cosmos network depends on: the wallets people hold their tokens in, the explorers and indexers that make a chain legible, and the maintenance behind the framework itself. Nothing in it points at a problem with Lava, which is running well. What it points at is a support layer that is thinner and less predictable than it was two years ago, and that is a question about the next several years rather than about today.
Whether to keep running our own chain is the whole question in this post. It is not an announcement and not a decision. A change this size has to pass a governance vote, and we would rather argue it before the vote than after. If the case is not made, Lava carries on running its own chain and we carry on maintaining it. That is a legitimate outcome.
What Lava set out to build, and why a chain was right in 2022?
Lava is a decentralized network for RPC access: the connections applications use to read and write blockchain data. It routes an application's traffic to independent providers, meters what each serves, and pays them for it, with the rules enforced by a ledger rather than a company's database. Almost every application depends on that access; Lava exists so it need not run through one company.
In 2022, running your own ledger was the credible way to get rules like those enforced by one at all: the alternatives were immature and the tooling a network like Lava needed had not settled. So Lava was born an app-chain, from its first commit in February 2022, built for one application rather than adapted into one.
This was not a contrarian bet. In June 2022 dYdX announced it was leaving Ethereum for a chain of its own, on grounds (throughput, and control over transaction ordering) that were correct on 2022 information. We made the same call, on the same information, and we would make it again.
What the network does today?
The chain went live on 17 January 2024 and has run to schedule since. Every monthly installment of the providers' reward pool has landed, thirty-one as of 2 August 2026, leaving 23,375,000 of the 66,000,000 LAVA allocated at genesis. Lava works, and it works at scale.
In the nineteen days from 16 July to 4 August 2026, the chain settled payment for 133 billion compute units, the protocol's unit of metered work, to 47 independent providers across 49 networks, twenty-six of them mainnets. That is roughly seven billion a day, reproducible with one public query. Demand reaches the chain aggregated through Lava's gateway rather than as many separate on-chain accounts. That is how the product is built, and why this post counts traffic rather than buyers.
Behind that traffic is a supply side the network attracted on its merits. Sixty-two provider operators are staked and reachable (the 47 above are those that earned paid traffic in the window), and they run 441 distinct endpoints across 62 networks, operated by more than twenty separate domains. All seven of the protocol's geographic regions are served, 35 networks from every one of them, and the supported-network list is still growing, with four added in May 2026 alone. The chain carries around 48,000 settlement transactions a day, uninterrupted across the whole 49-day window a public node retains. More than 720 addresses that are not providers themselves have delegated 76 million LAVA to providers, about a fifth of all bonded stake.
What has been learned from running it?
Running the network at that volume taught us where its value concentrates: in the service layer, meaning routing, metering, quality, and the provider relationship. The traffic settled what the design could only assume. That is product-market fit showing up in the data, and over the past year we have concentrated engineering there, consolidating around the paths that serve traffic.
The chain's role in that picture is settlement, and it does the job well. We built its on-chain surface with room for much more: 24 kinds of transaction and 72 kinds of query. One message type is broadcast automatically by the software running the network as providers settle their work, and those 48,000 daily transactions are almost all that one message. Every other write is triggered by a person, either an operator or governance. Reads far outnumber writes.
A public node retains 49 days of history, 16 June to 4 August 2026. We hold the full chain and cite the window anyone can reproduce without trusting us. Across it, nobody needed to file a dispute on-chain. The process is live and the market has not called on it.
What changed in the wider world?
Lava's 2022 bet was an industry-wide one, and the industry has since re-priced it: not refuted, re-specified. A dedicated chain was the default for a serious protocol in 2021 and 2022. By 2026 it is a conclusion teams reach for specific reasons, not a premise they start from.
The clearest evidence is structural. Polkadot, whose design was organised around auctioning multi-year chain slots to applications, discontinued those auctions in September 2024, replacing them with on-demand purchase of the same capacity. The standing-slot premise went away. Moonbeam, which won one of that ecosystem's first slots in November 2021, announced in July 2026 that it was ending parachain operations. Ronin, one of the most prominent independent launches of 2021, reversed that decision in May 2026. Polkadot and Moonbeam hosted other people's applications rather than running one of their own, which is the point: what moved is the default. The cohort unwinding now is the same cohort that was right to commit in 2020 and 2022, on the information available then.
The second shift is cost and speed. Sustained capacity work across the industry through 2024 and 2025 moved the cost base by orders of magnitude, and throughput, which in 2022 a team built for itself, is now unremarkable.
Both shifts land on the same assumption, that a network like Lava had to run a chain of its own. In 2022, that was true.
And the honest counter: purpose-built chains are not over. Some of the industry's most sophisticated teams launched new ones in 2025 and 2026, and where the chain and the product are the same thing, a chain of one's own is still the right answer. That is the distinction Lava has to make about itself. We have a view; the point of this post is that we would rather put the test to you than the conclusion.
What changed around Lava's own stack?
Everything in this section is about direction, not about where Lava stands today. Lava's footing is solid.
The Cosmos stack Lava runs on is not one dependency but many: wallets, validator coordination, maintenance of the SDK, of CometBFT and of IBC, explorers, indexers, staking interfaces. A project depends on all of it and controls almost none of it. One part weakening is something you adapt to. The concern is several weakening together, until ecosystem support falls below what Lava needs.
We measured one part of this ourselves. Through the first half of 2026, non-merge commits to the SDK, CometBFT and IBC-Go were down 26% year over year, though up 30% on the second half of 2025, while the layer around them (clients, relayers, testing and chain tooling) fell considerably faster. Commit counts are a rough proxy; the divergence is the point, and we will publish the basket and the counting rule on request.
To be precise: the Cosmos SDK is actively maintained, still shipping releases, still chosen by new projects, and core maintenance is not the worry. The trend runs in the layer around it, the one a project controls least. Everything Lava depends on there works today; we are watching the slope, not the level.
Wallet support is the clearest case, and it works today. Keplr covers Lava and is well-built software. What we watch is the direction. Leap Wallet sunset its products in May 2026, and fewer independent teams carry Cosmos wallet support than two years ago.
Projects have also been leaving or consolidating. Stargaze voted in November 2025 to migrate into the Cosmos Hub, Nillion moved off its chain in March 2026, and Evmos halted block production by governance decision in May 2026. These do not share a single cause, and none says anything about any other project's health, including ours. The wider market has been down across the board, and none of this is Cosmos-only. What they change is how confidently anyone can forecast third-party support five years out.
What it adds up to, and what comes next?
Lava funds a validator set, a reward schedule, an upgrade pipeline, and custom software whose only job is keeping validator nodes upgraded in step. On top sits a maintenance bill. We maintain our own build of the Cosmos SDK, with CometBFT and IBC held at matched versions, a deliberate arrangement we keep current. It also means upstream and our build diverge, and closing that divergence is a version migration that produces no new product. We would absorb it once on any path; on this one we absorb it again every upstream cycle, at a cadence we do not set. We are not publishing a headline figure. The honest cost is engineering time, not spend, and we would rather give the components than a number we picked.
This is why the question has to be asked. Lava pays in full for a chain of its own, at a price set by running one rather than by how much of the product runs on it; the support that chain depends on is trending in a direction we neither control nor influence; and the assumption that made a chain of one's own obvious in 2022 has been re-priced by nearly everyone who made it. The obvious reply is that a chain can be run more cheaply than we run ours, and that is one of the options part two examines. Most of what we describe dates from the last eighteen months, and slopes are easier to act on early than late. Nothing forces the timing; we would rather ask now.
Here is what we would want out of a change, and part two measures its candidates against exactly this. First, less of Lava's future should rest on support that Lava neither sets nor influences. Second, we want to build more for the people who use Lava, and sooner; what we can offer them is shaped in part by what the stack around us supports. Third, the network should cost less to run and maintain, with security stronger for the change rather than merely preserved. Those are objectives for a change, not a description of any answer.
Part two takes the next step: what we examined against them, including staying exactly where we are, and what we concluded. It goes out within two weeks, and nothing is being asked of anyone in the meantime.
What a change would mean for holders, delegators, and the operators who run the chain today will be set out in full before there is anything to vote on, as part of the proposal rather than an afterthought. Then a technical session, a forum temperature check, and a formal governance proposal. None of these is a formality. If the temperature check does not show clear support, it does not go to a vote.
So argue with this. If you think we have mis-priced what the chain costs, overstated what has changed around it, or understated what we would give up, say so on the forum or bring it to the call. The strongest version of this proposal is whatever comes out the other side; if the case does not hold up, that is an answer too.
