RAY in the Raydium machine

Raydium is less “a DEX with a token” and more liquidity infrastructure with multiple execution surfaces. It runs swaps and liquidity across several pool designs (CLMM, CPMM, and legacy AMM v4) and routes order flow through a single app and APIs on Solana. Raydium’s own docs frame it as on-chain liquidity primitives and position it as broadly integrated liquidity infrastructure on Solana.

In that system, RAY is primarily the protocol’s incentive + fee-recycling asset, not the token you must hold to use the exchange. The two load-bearing mechanics are straightforward:

1) A fixed-cap token supply whose remaining distribution is largely a controlled emissions budget.

2) A protocol-level rule that routes 12% of trading fees into programmatic market buys of RAY.

Everything else about RAY’s value capture depends on who controls the accumulated RAY and what constraints exist on changing the rules. That’s where the mechanism design gets interesting. And where the modelability gets brittle.

Supply, allocation, emissions

RAY has a fixed maximum supply of 555,000,000.

The canonical mint on Solana is 4k3Dyjzvzp8eMZWUXbBCjEvwSkkk59S5iCNLY3QrkX6R.

Raydium also publishes a live circulating supply endpoint. As of March 6, 2026, it returns 268,599,298.675857 RAY.

The allocation (including on-chain addresses) is specified in current protocol docs.

Lockups were defined crisply for the most governance-sensitive buckets. Raydium states that team and seed allocations (25.9% of supply) were fully locked for the first 12 months after TGE, then unlocked linearly on a daily basis from months 13 through 36, with vesting ending on February 21, 2024.

On emissions, Raydium’s current token page is intentionally high-level. It says the protocol “currently emits approximately 1.9 million RAY per year from the mining reserve.”

From a mechanism design standpoint, that line matters because it implies the “inflation rate” is not a mint function beyond the cap. It is a release schedule from a pre-allocated reserve. The constraint is real. The exact emission function, though, is not specified in the same place. That makes forward modeling dependent on off-chain commitments and admin behavior.

Fee flows and buybacks

Raydium’s most legible value-accrual rule is also its most deterministic. The protocol defines explicit fee splits by pool type, routing fees to LPs, a buyback stream, and sometimes a treasury stream.

At the policy layer, the key invariant is simple: 12% of trading fees go to RAY buybacks.

At the implementation layer, Raydium documents concrete splits:

CLMM pools: LPs earn 84% of trading fees, 12% is allocated to buybacks, and 4% to the treasury.

CPMM pools: same 84% / 12% / 4% split.

Standard AMM v4 pools: LPs get 88%, and 12% goes to buybacks (no treasury share listed on that page).

Raydium also publishes the bps breakdown for “standard AMM (OpenBook)” swaps: a 25 bps fee where 22 bps goes to LPs and 3 bps goes to RAY buybacks.

The buyback process is automated but not “burn-enforced.” Buybacks trigger automatically when accumulated fees reach $10, and the bought RAY is held by the protocol.

Raydium even discloses collection addresses by pool type and the custody address where bought-back RAY is held.

This is where I draw a hard line between “predictable” and “discretionary” tokenomics.

Predictable: the fee split is parameterized and documented, and the buyback is programmatic with a defined trigger threshold.

Discretionary: the outcome of the buyback is not fixed. Bought RAY is held, and the docs do not state a mandatory burn, mandatory redistribution, or immutable escrow constraint. Control of the relevant treasury and fee-collection addresses is explicitly attributed to a protocol multisig.

So the system gives you a deterministic bid, not deterministic permanent supply reduction. For contrast, the RUNE tokenomics review covers a different value-capture design space.

Staking and incentives

Raydium’s current documentation treats staking as an emissions-based reward mechanism: “Stake RAY to earn additional RAY rewards.”

Critically, the up-to-date English docs that specify protocol fee splits do not route a portion of swap fees to stakers. They route fees to LPs, buybacks, and sometimes the treasury.

That puts staking in a clean bucket, economically:

Staking is a claim on emissions (distributed as RAY), not a direct claim on fees (distributed as USDC or base/quote assets). For a checklist of what to model, see token design components.

From a mechanism design lens, emissions-based staking is simple to implement and easy to reason about at the contract level. It also has an inherent reflexive dependency: staking APR depends on the token price and the emission rate, and it competes with LPing yields that come from fee flow. Raydium’s design tries to offset this by making buybacks a persistent sink for fees, which can counterbalance emissions on net.

The missing piece, for rigorous modeling, is the exact control loop that sets emissions over time. Current docs provide a point-in-time rate (“currently emits”) rather than a formal schedule. That is workable if you trust governance. It is weaker if you demand immutable constraints.

Governance and parameter control

Raydium’s regulatory-style token white paper describes RAY as “the native utility and governance token” and states that token holders can participate in governance by voting on “community proposals and protocol parameters.”

In practice, the enforceable governance that matters for tokenomics is the access-control layer over programs and fee wallets.

Raydium is explicit that its programs are upgradeable programs owned by the BPF Upgradeable Loader, with an Upgrade Authority that can push updates. It further states that upgrade and admin authority is held under a Squads multi-sig.

Two lines in the Security docs should be treated as tokenomics primitives, not security footnotes:

1) “Raydium does not currently employ a timelock for upgrades.”

2) “All treasury and fee collection addresses are controlled by the protocol multisig.”

Those statements bound your confidence in parameter stability. Fee splits can be changed by upgrading programs. Buyback custody can be redirected or repurposed by multisig-controlled addresses. Even if the project intends to decentralize later, today’s tokenomics have a discretionary actuator. For a Solana-native comparison, the JUP tokenomics review discusses a different governance and incentive setup.

Raydium does say timelocks are planned “as code moves toward open-sourcing and community governance.” That is a direction, not a constraint. If you’re valuing RAY on fee capture, the presence or absence of a timelock is not cosmetic. It’s a commitment device.

Structural changes and documentation drift

Raydium has evolved materially since its early “hybrid AMM” identity. Even the current “About Raydium” page notes that AMM v4 pools were originally hybrid, sharing idle liquidity with central limit order books (Serum, then OpenBook), but that today “all v4 pools function as traditional AMMs.”

That evolution matters for RAY because it changes what fees exist, where they can be routed, and what portion of the stack Raydium controls end-to-end.

Raydium’s own Security page also explicitly references “Order-book AMM & OpenBook migration” as an audited program change.

There’s also a clean, dated tokenomics milestone: vesting for team and seed concluded on February 21, 2024.

Finally, the project has documented an exploit remediation flow that explicitly used RAY from vested team reserves to compensate LP losses. Raydium’s Claim Portal page describes remediation “for the December 16th exploit” and states that RAY used for certain compensation categories was distributed from vested team token reserves.

The associated Medium post is dated January 4, 2023 and reiterates the same structure, including the 1:1.2 compensation ratio in RAY for some uncovered losses and that RAY came from vested team reserves.

From a deterministic-governance perspective, the remediation is a double signal:

It is positive because the project executed a concrete, written plan and documented the mechanics publicly.

It is also a reminder that “token allocations” are not just a distribution chart. They can be mobilized as an insurance backstop by whoever holds effective control. That is a governance feature. It is also a centralization dependency. If you track these changes across protocols, our crypto research collects related work in one place.

Risk register

Raydium’s tokenomics are unusually readable on fee splits and unusually under-specified on governance constraints. That combination creates a clear dominant risk: parameter and custody centralization around buybacks, treasury, and upgrades. The mechanism works. The question is whether it stays the same mechanism.

Top 3 risks

  1. Dominant risk: Upgrade-and-custody centralization (no timelock). Trigger: a multisig-approved program upgrade or treasury action that changes fee splits, buyback routing, or the disposition of bought-back RAY. Mechanism: Raydium programs are upgradeable under an Upgrade Authority, controlled via a Squads multisig, and Raydium explicitly states it does not currently employ a timelock for upgrades; treasury and fee-collection addresses are controlled by the protocol multisig. Who bears it: RAY holders (value capture assumptions), LPs (net APR via fee splits), and integrators (routing assumptions). Measurable indicators: changes to published protocol-fee split tables, changes to disclosed treasury/buyback addresses, on-chain program upgrade events, and any introduction or removal of timelock controls in the Security docs.

    The mechanism-level reason this is dominant is that it subsumes most other risks. If emissions rise, you can model dilution. If volumes drop, you can model fee pressure. If upgrade authority is unconstrained, then the mapping from activity to value capture is not purely algorithmic. It is a policy choice, exercised through code upgrades and multisig-controlled wallets.

    Even the buyback design, which is often treated as “automatic,” terminates in custody. Raydium states bought-back RAY is held by the protocol, and it discloses the holding address. Without a stated burn rule or a non-upgradeable escrow constraint, “buybacks” are best modeled as treasury accumulation funded by fees. That can still support price. It can also be redeployed later in ways that dilute the holder’s mental model.

    If you want deterministic credibility, you look for at least one of these: an on-chain timelock, immutable fee split parameters, or a hard-coded disposition of bought-back tokens. Raydium’s docs acknowledge the timelock gap and treat timelocks as future work. Until that changes, your confidence interval on “future tokenomics” should be wide.

  2. Emissions-policy opacity. Trigger: a change in effective emissions rate from the mining reserve relative to current expectations. Mechanism: Raydium states it “currently emits approximately 1.9 million RAY per year” but does not define the immutable emission function or a schedule on the same page; emissions-based staking incentives depend on that policy remaining stable. Who bears it: stakers (APR variability), RAY holders (dilution vs buyback balance), and LPs (relative competitiveness of incentives). Measurable indicators: changes to the “currently emits” statement, on-chain transfers from the mining reserve address, and any new published emission schedule.

  3. Smart contract and operational shock (including exploit remediation costs). Trigger: a contract vulnerability, an integration migration event, or an incident requiring treasury-backed compensation. Mechanism: Raydium operates multiple pool programs and has documented exploit remediation via a Claim Portal process, including compensation partly denominated in RAY sourced from vested team reserves. Who bears it: LPs and traders first (fund safety), then RAY holders indirectly (treasury depletion, token reserve usage), and the broader ecosystem (liquidity fragmentation during migrations). Measurable indicators: bug bounty activity, disclosures of new incidents, changes to audit coverage, and net outflows from treasury-related addresses.

If you’re doing tokenomics consulting on RAY for a treasury, market maker, or DeFi protocol integration, the work is less about “what the token does” and more about which parameters are credibly constrained. The fee splits are clear. The custody and upgrade constraints are not yet hard-bounded by a timelock. That gap should be explicit in any token economy design model you publish, and it’s the kind of work our tokenomics design services are built around.



This article is part of our Tokenomics Deep Dive series.