Why aren’t you running an RPC service on Horizon already? Take my money!
I am not an indexer. I am the other thing: a developer who runs indexing infrastructure and pays for RPC. This post is a receipt, a repo link, and a question.
The receipt
Since 12 May I have paid Alchemy $617.50 across seven invoices, and the account has accrued $674.79 as of this morning. That is 1.26 billion compute units, all plain HTTP JSON-RPC apart from a two-month Solana gRPC experiment that I will come back to.
Here is what I was actually buying, August and September combined, by method:
| Method | Calls | Cost | Share |
|---|---|---|---|
eth_getBlockByNumber |
36.4M | $311.81 | 70.8% |
eth_getLogs |
4.0M | $102.63 | 23.2% |
eth_blockNumber |
5.1M | $21.95 | 5.0% |
eth_call |
0.4M | $4.10 | 0.9% |
| everything else | $0.69 | 0.2% |
And by chain:
| Network | Cost | Share |
|---|---|---|
| Arbitrum One | $305.86 | 69.1% |
| Monad | $91.57 | 21.0% |
| Ethereum mainnet | $32.02 | 7.3% |
| BNB, Base, Polygon, Gnosis, OP, testnets | $11.72 | 2.6% |
Look at that first table. Three quarters of my bill is block headers and log ranges. No traces, no debug namespace, no archive-state gymnastics, no MEV bundles. Headers, logs, and a tip poll. This is the most commoditised workload in the entire EVM stack, and it is exactly the workload every one of you already serves to your own graph-node all day long.
The shape of the spend matters too. Steady-state tip following costs me $1–2 a day. The money is in backfills: a 54.7M CU day on Arbitrum on 19 August, a Monad backfill that cost $91.57 in three days in early September, a 60M CU Ethereum mainnet backfill on 9–10 September. September on its own is heading for $265–300. When I point a new indexer deployment at a new chain, someone gets paid a few hundred dollars for the privilege of handing me headers. Right now that someone is Alchemy.
(The Solana bit, for completeness: 1.72 TB of Yellowstone gRPC egress in June and July, $128.74 at $75/TB. Briefly the most expensive thing on the account. Different product, different post, but the same question applies — why is nobody on the network selling me a firehose?)
This already exists, sort of
Dispatch is a JSON-RPC data service on Horizon (see GRC-005). Indexer runs an RPC node behind it, registers a provision on the DispatchService contract, serves requests, gets paid through TAP receipts and RAVs like any other Horizon data service. The gateway routes, the escrow settles, GraphTally does the tally. Same rails as subgraphs, different payload.
I will be honest about where it is: MVP, rough around the edges. I did a gap analysis against dRPC across roughly 25 dimensions, and the two hard problems are real — response verification (there is no canonical on-chain truth to slash against for an RPC answer) and cold-start (a data service with zero providers is not a data service). There are MUST-level items on the roadmap that are not done. It is not a drop-in Alchemy replacement today.
But it is not vapour either. The contracts are there, the protocol mapping is there, and the thing nobody can write in a repo is the part I am asking for in this post: providers.
Why you should care
You already run the node. An archive Arbitrum One node, an Ethereum mainnet node, probably a few L2s. They sit at some fraction of capacity because graph-node’s demand is bursty and you provisioned for the peak. That idle capacity is currently worth zero.
The RPC market is not small. Alchemy, Infura, QuickNode and dRPC exist because developers will pay real money — my money, above — for boring, reliable eth_getBlockByNumber. dRPC in particular has shown that a decentralised provider network can win real traffic. There is no structural reason that provider network should not be The Graph’s indexer set, on The Graph’s payment rails, with The Graph’s QoS scoring deciding who gets the query.
Horizon was built precisely so that data services beyond subgraphs could be added without a governance war every time. Substreams is one. RPC is the obvious next one, because the supply already exists and the demand is sitting in a Stripe dashboard in Palo Alto.
And from my side of the table, the pitch is not even about decentralisation. It is about who I would rather pay. I would rather pay an indexer I can see on-chain, whose provision I can inspect, whose QoS history is public, than a black box that bills at a tier rate its own CLI does not know about. I am already in the ecosystem. My GRT is already here. Let me spend it.
The ask
Concretely:
- Indexers with spare RPC capacity on Arbitrum One, Ethereum mainnet or Monad - reply here or open an issue on the repo. I want to know what it would take for you to register a provision. Hardware you already have? Ops burden? Payment expectations? Something in the contract design you would want changed first?
- Anyone who has looked at Dispatch and thinks it is wrong - say so, specifically. The gap analysis and roadmap are open. I would rather be told the verification model is broken now than after providers have staked against it.
I have a paying workload, a public usage ledger, and no loyalty whatsoever to my current provider. I will move the Arbitrum tip cursor first, then the backfills, and I will publish what it costs me on the network versus what it costs me on Alchemy, month by month.
Why aren’t you running an RPC service on Horizon already?
Take my money.
Usage figures are from Alchemy’s Orb invoices and dashboard export for 12 May – 14 September 2026. Happy to share the full ledger, including daily burn and per-method reconciliation, with anyone who wants to size a node against it.