GRT is a security budget with a steering wheel

The Graph is a decentralized indexing and query network. It turns “who serves data” into a market, and it uses GRT as the bond that decides who gets paid and who gets punished. Indexers stake GRT to run infra and serve queries, Curators signal GRT to steer indexing attention toward specific subgraphs, and Delegators park GRT behind Indexers to scale their stake-weight, as described in the indexing and query overview.

That sounds like neutral market structure. The real design center is control. The contracts are upgradeable, and the protocol is configured by a Governor role that sits above the contract registry (“Controller”). The Graph’s own contracts documentation describes the Controller as owning the protocol contracts, and states that “the owner of the Controller is The Governor,” adding that “The Governor is The Graph Council.” This matters more than any single incentive parameter. Whoever holds Governor authority holds the ability to rewrite the economic constitution.

Operationally, the principal network deployment has moved to Arbitrum One. The protocol contract list calls Arbitrum One “the principal deployment of The Graph Network.” Tokenomics analysis that treats Ethereum mainnet as the center of gravity is now structurally stale.

Supply, issuance, and initial distribution

GRT supply is designed to be elastic. The protocol’s tokenomics documentation states an initial token supply of 10 billion GRT and describes a target of 3% new issuance annually (minted as indexing rewards), alongside recurring burn mechanisms. This is the lens we apply in our tokenomics methodology.

On March 6, 2026, current supply figures show 10,739,019,282 GRT circulating and 10,800,262,816 GRT total supply. CoinGecko also shows a “max supply” equal to the same ~10.800B range. That field conflicts with The Graph’s own framing of ongoing issuance. Treat it as a data-label artifact, not a monetary cap.

The earliest public distribution event is well documented. The Graph’s public sale announcement says 4% of total token supply (400M tokens) was allocated in the sale, and the sale post describing mechanics specifies 400,000,000 tokens for sale at $0.03 per preGRT, with an initial total supply of 10,000,000,000 GRT.

Initial release distribution (as documented by The Graph Academy) is below. These buckets matter because governance power does not just come from “one token, one vote” narratives. It comes from who held meaningful size early, who vested, and who had the runway to delegate or stake at scale during low-fee periods. For a baseline comparison, see how another major network frames distribution and issuance in our ATOM tokenomics review.

Fees, burns, and who gets paid

The fee path is simple in principle and messy in implementation details. Developers and data consumers pay query fees in GRT, and the protocol distributes those fees to network participants “based on their contributions.” Indexers earn “query fee rebates” and indexing rewards. The docs describe query-fee payments as mediated via GraphTally between an Indexer and a gateway, with signed receipts.

On the sink side, The Graph’s docs specify three routine burn channels:

Delegation tax burn: the tokenomics docs state a 0.5% delegation tax that is burned when a Delegator delegates. Note that the Graph Horizon documentation describes a future/upgrade state where “the 0.5% delegation tax has been completely removed.” If you model long-term burn, you cannot assume this tax persists unchanged.

Curation tax burn: Curators incur a 1% curation tax when signaling on a subgraph, and the docs state this fee is burned.

Query fee burn: the tokenomics docs state 1% of query fees paid to the network are burned.

For Curators, the revenue share is explicit. The curating docs state 10% of all query fees go to Curators, pro-rata to curation shares. That is not just “yield.” It is a steering incentive. A protocol that pays Curators to point Indexers is admitting that indexing attention is scarce and must be budgeted.

The burn side also includes enforcement burns. If an Indexer is slashed, the tokenomics docs say 50% of indexing rewards for the epoch are burned (with the other half going to the Fisherman), and the Indexer’s self-stake is slashed by 2.5%, with half of that slashed amount burned. The glossary reiterates that the slashing percentage is a protocol parameter “currently set to 2.5%,” with a 50/50 split to Fisherman and burn.

Staking roles and incentive geometry

Indexers sit at the center. They choose what to index, how to price, and how to share revenue with Delegators. Indexing rewards are distributed across subgraphs based on curation signal, then to Indexers based on allocated stake, and eligibility depends on closing allocations with a valid Proof of Indexing (POI) that meets arbitration standards.

Indexers also set their own “cuts” for sharing with Delegators. The docs show Indexers can configure queryFeeCut and indexingRewardCut via setDelegationParameters. This is a governance surface that is not token-holder governance at all. It is local, operator-level policy that shapes who can compete for stake.

Delegation is constrained by design. The protocol enforces a delegation capacity, and the docs state Indexers can accept up to 16x their self-stake in delegated stake. This is a decentralization lever. It blocks pure “brand capture” where an Indexer could absorb infinite delegated stake without bonding meaningful self-risk.

Withdrawal friction is also intentional. Delegators must wait roughly 28 epochs to withdraw after unstaking. The undelegating docs summarize it as approximately 28 days. The tokenomics page adds more precision, describing a 28-epoch unbonding period and giving an epoch size of 6,646 blocks, estimating about 26 days.

Slashing exposure is asymmetrical, and this shapes who bears tail risk. The Explorer docs explicitly state: “Delegated” stake can be allocated by the Indexer but “cannot be slashed.” The earlier protocol design blog is blunt that delegated stake “will not be slashable,” explicitly to avoid trust relationships that could produce winner-take-all dynamics.

Then Horizon re-opens the question. The Horizon documentation says delegation is currently not slashable, but the new system “includes the capability for delegation to become slashable in the future,” with indexer stake slashed first as a buffer. From a governance-power perspective, this is the pattern to watch. Economic risk can be shifted between roles by upgrade, even if your “base token utility” story never changes.

Governance and parameter control: where the real power sits

The Graph’s public governance narrative is council-led, not token-holder-led. The governance page states: “Protocol upgrades and the community treasury are overseen by The Graph Council, with support from The Graph Foundation, core contributors and the community.” It also describes a GIP flow where community votes are “non-binding” signals, followed by council approval.

The explicit threshold is not subtle. The governance page states 6 of 10 votes are required for council approval. That is a small committee with veto power over upgrades and treasury direction. You get speed, coordination, and a single accountable surface for execution. You also get concentrated control over the parameter layer that most directly determines GRT’s value capture profile.

Onchain, the contracts documentation makes the control stack legible. The protocol contracts README describes the Controller as the registry of protocol contract addresses and “the owner of all the contracts,” and states that the Governor can configure the whole protocol. It also notes that epoch configuration is configurable by the Governor.

Even without enumerating every setter function, the control implication is clear. Monetary policy (issuance rate), market microstructure (rebate logic, taxes, burn rates), and enforcement parameters (slashing) are not immutable. They are governance surfaces. The same documentation bundle that lists deployed addresses also shows protocol-level parameters like an issuanceRate value in initialization arguments and references a specific governor address in contract initialization.

Trade-off, stated plainly. The Graph optimizes for shipping a complex, multi-role marketplace with a coherent economic policy. The price is credible neutrality. You can call this “progressive decentralization.” You can also call it what it is structurally: a governance stack where a small group can ratify upgrades, and where community sentiment is advisory until it is operationalized by that group.

Risk register

The Graph’s mechanism design is coherent. The risk comes from where the cashflows are thin, the rules are adjustable, and the role boundaries can be redrawn by upgrade. We track these governance and cashflow risks across protocols in our crypto research.

Dominant risk: governance-driven parameter instability

The dominant risk is not that a parameter is “bad” today. It is that the parameters you think you own are not sovereign. The protocol is explicitly configured by a Governor role, and the contracts documentation ties that role to the Graph Council. The governance process itself is council-ratified, with non-binding community votes and a 6-of-10 council threshold to approve implementation.

Mechanically, this creates a live political economy around monetary policy and market design. Issuance is a protocol-level dial, described as “3% annual issuance” for indexing rewards in the docs. Burn levers exist and are already subject to redesign. The Horizon documentation explicitly removes the delegation tax and introduces a pathway to slashable delegation. That is not cosmetic. It is a redistribution of who pays protocol friction and who bears tail risk.

From a power-distribution lens, “decentralized governance” claims should be discounted until control over upgrades and monetary parameters is credibly dispersed. Today, the documents emphasize council oversight of upgrades and treasury, not token-holder execution. For participants, this means expected returns and risk profiles are contingent on a small-group decision process, even if those decisions are socially constrained by community legitimacy.

  1. Governance capture / upgrade-key risk, Trigger: a contentious proposal affecting issuance, burns, delegation slashability, or fee routing gets pushed through a council ratification cycle. Mechanism: Governor-level configuration and upgrade authority concentrates economic rule changes behind a council process where community votes are non-binding signals and council approval is decisive (6 of 10). Who bears it: all passive holders and role-participants whose strategies assume stable taxes, stable risk, and stable issuance. Indicators: frequency of GIPs that modify “economic parameters,” visible governance milestones requiring council ratification, and any documented shifts like Horizon’s explicit removal of delegation tax or introduction of delegation slashability capability. For a contrasting approach to upgrade governance, see our Optimism tokenomics review.

  2. Security budget outruns organic fee demand, Trigger: query fees remain low relative to indexing rewards over extended periods. Mechanism: indexing rewards are funded by ongoing issuance described as 3% annual issuance, while query-fee burns are a fixed fraction (1% of query fees burned) and do not automatically scale to offset issuance. Who bears it: unstaked holders via dilution, plus ecosystem participants paid in GRT if sell pressure increases. Indicators: circulating supply trend, protocol-level issuance versus burn disclosures, and sustained divergence between “query fees” and “indexing rewards” visible in Explorer dashboards and indexer revenue summaries.

  3. Curation incentive distortion, Trigger: curation becomes dominated by self-signaling and capital-driven placement rather than third-party discovery. Mechanism: indexing rewards are allocated across subgraphs based on curation signal, and Curators receive 10% of query fees, creating a feedback loop where capital can steer both attention and rewards. Who bears it: smaller developers and independent curators who cannot compete for signal placement, plus the network if indexing attention is misallocated. Indicators: concentration of curation signal on a small set of subgraphs, high rates of subgraph version migrations (which incur curation taxes), and persistent “pay-to-be-indexed” dynamics implied by developer-led curation behavior.

If you are building a protocol or treasury strategy around GRT exposures, treat governance and parameter mutability as first-class modeling inputs, not footnotes. This is where tokenomics consulting becomes less about “utility narratives” and more about mapping who can change which dials, under what process constraints, and with what credible commitment.



This article is part of our Tokenomics Deep Dive series.