Tokenomics

You can only optimize for two out of the three points in the diagram above.

We choose to remain permissionless and let rewards for each node float based on the current levels of inference demand.

Nodes earn POD tokens based on inference work completed.

If more people are using the network, utilization of each node will increase which naturally incentivises more supply to come online (with a bit of a delay).

We like this because the node count grows naturally with demand, compared to other networks that overly incentivise the supply side with predictable rewards – resulting in a vast network of mostly idle compute & heavy token emissions.

Node operators earn emissions from treasury based on inference tokens processed.

Currently on datagen.dphn.ai we pay $0.5 per 1m tokens worth of POD in emissions.

In prod, we may set pricing so that protocol revenue targets net buy pressure, or set pricing lower than emissions so that we subsidise growth with emissions.

The peer-to-pool design is different to some other peer-to-peer networks that link buyers of compute into a "session" with a node for X period of time, and settle on-chain at the end.

On the demand side, API users buy credit from the protocol by depositing tokens to a smart contract. Their requests get sent to any node on the network depending on what is currently available.

On the supply side, nodes running the same model form a 'pool' that processes requests sent to that model. Nodes are assigned jobs based on availability and performance, and no direct link between the request sender and node provider is made. Nodes are rewarded for their contributions with POD tokens from the node rewards pool.

The decoupling of the buy and sell sides of the network means that more or less POD can be distributed to nodes than what is received from revenue.

This allows the network to initially operate at a net loss to grow the number of available nodes and offer larger amounts of free inference until the network stabilizes around a natural equilibrium of inference demand and node supply.

The network is designed to have protocol revenue outpace emissions when nodes are close to full utilization, which would mean that the network is operating at a profit resulting in net positive buy pressure on POD.

Our token design is inspired by the best parts of other projects that we found to compliment our distributed inference & training network:

• ETH style slashable bond deposits for node operators & validators
• CRV style reward boosting for node operators (up to 1.3x boost for bonding staked POD - up to 1.2x is linear with 3 months of bond, and 1.2x to 1.3x is competitive based on what other node operators are doing plus an absolute 6 month bond requirement for the max boost)
• xSUSHI / yCRV / yYFI style auto-compounding staking, no need to claim rewards means that staked POD can be deposited as a bond to boost node rewards. Tax efficient as no claims have to be made and all appreciation is in the staked POD position
• stAAVE style 3 month cooldown windows for staking & bonding + stAAVE & vlCVX constant non-decaying rewards

POD can be staked into a transferable auto-compounding vault similar to xSUSHI / yCRV / yYFI.

Staked POD wallet balance stays the same as rewards are distributed - the redemption ratio increases as POD rewards are sent to the contract.

eg. Stake 10 POD on day one. We add 10% dividend. Redemption ratio is now 1.1. Withdraw your staked position - receive 11 POD.

Staked POD is for investors or inference users who receive:

1. A portion or all of network revenue, in POD (auto-compounding balance)
2. Access to different levels of product subscriptions without having to pay directly
3. Daily free inference allowance

Staked POD has a delegation function for both governance & daily inference allocations.

Meaning that we could have a CRV / CVX style votemarket where people bid to receive delegations from stakers who don't use their daily inference.

Extra yield for stakers who don't use their allocation & cheap inference below the market rate for those who bid for delegations.

vlCVX "constant power" refers to the mechanism within Convex Finance where the voting power & staking APY derived from locking CVX tokens for 16 weeks (vlCVX) remains constant throughout that lock-up period, unlike Curve Finance / other veTokens where voting power decays over time and it forces you to perpetually re-lock with capital becoming a sunk cost.

Many consider this an improvement over veTokenomics as the asset is more attractive to buy & stake.

This takes the benefits of vlCVX (consistent full APY without having to re-lock to maintain 100% of rewards) while having users stay constantly staked by default instead of all stakers becoming unlocked every 16 weeks.

Our node bonding contract takes staked POD as deposit - this becomes bonded POD.

Bonded POD is non-transferable with slashing / freeze & some other features.

Node operators that are bonded can be slashed if they are caught doing anything malicious such as trying to cheat by modifying the node software or model that is being run.

Hot wallets on validator nodes can freeze bonds if they detect model mismatch - this disables exit during the 3 month cooldown, and can be cancelled if a false flag.

Multisig can slash or unlock these frozen wallets upon review - this process can be automated further in the future.

The POD reward multiplier determines how much additional yield an operator earns on top of their base node rewards. It is inspired by Curve's LP boost mechanism, but adapted for a decentralized AI network with usage-based rewards, account-wide bonding, epoch-based accounting, and slashable operator bonds.

At a high level:

• Nodes earn base rewards from completed inference, validation, and related protocol work.
• Accounts that bond staked POD receive a multiplier on eligible node rewards owned by that account.
• Boosts activate only for registered, unsuspended accounts with enough recorded earnings history and at least roughly $5 of total lifetime base earnings at launch. This is implemented as a governance-set POD threshold, not as a live USD oracle and not as a $5/day requirement.
• The earnings denominator is based on a trailing multi-epoch average of base rewards, then updated with asymmetric smoothing: it rises quickly when activity increases, but falls slowly when activity decreases.
• Reward multipliers are computed only at epoch boundaries and then frozen for the next reward epoch. Reward workers use the frozen account snapshot instead of recalculating from live state.
• Bonding the equivalent of 3 months of earnings (13 weeks) gives a linear pathway up to a 1.2× multiplier on boostable rewards once the account is boost-eligible.
• The maximum multiplier is 1.3×. The pathway from 1.2× to 1.3× is competitive, based on relative over-bonding versus other operators, and also requires at least 6 months of earnings bonded as an absolute requirement for the max boost.
• Accounts that maintain at least ~3 months of earnings bonded (13 weeks) can become validator-eligible. Validators check the logprobs of other nodes to detect and report deviations from the expected model fingerprint. Eligibility uses persistent epoch-based hysteresis to prevent flapping.
• Providers exclusively earn bonded POD for the first month of operation so that all nodes have a strong economic incentive not to cheat.
• Node providers have two options when claiming their weekly rewards from the contract: claim as bonded POD as the standard option, or claim as liquid POD with a 20% fee that goes directly to stakers by being sent to the staking contract.
• Until an operator has 4 weeks worth of their expected earnings as bond, they can only claim bonded POD. After this point, they can choose between both options.
• Bonds that activate the cooldown timer lose boost + the ability to claim liquid POD & validator status.

All calculations are done in POD units only, without any price oracle inside the reward engine. The bond is per account (wallet), and the resulting multiplier applies across all nodes owned by that account. Adding more nodes increases the account's base earnings and therefore requires proportionally more active bond to maintain the same weeks-bonded ratio.

Our seed round was $886,000 on 20th June 2024 for 117,600,000 tokens with 1 year cliff vesting.

All seed tokens are fully unlocked.

As of 1st May 2026, seed investors are still holding 67.4m POD in total, equal to 13.49% of total supply and 57.36% of the seed allocation.

Of this, 15.5m POD is liquid and 51.9m POD is staked.

The team bought back 38m POD from seed investors at $0.01 during Q1 2026.

The treasury tokens & LP is held in a 3/4 Gnosis Safe multisig.

Treasury tokens will only be used for node emissions or future fundraising.

Node emissions are paid solely based on how much inference is processed by node operators.

This avoids over incentivising supply in the absence of demand.

Team tokens are not fully distributed yet. All core team members will be vested for a minimum 2 year cliff + 5 years linear starting from June 2026.

4.25m tokens were also sold otc across 3 deals in 2025.

1 year cliff + 2 year linear vesting was applied to these sales - all of them were sent as staked tokens.

Right now staking and bonding both have a 3 month stAAVE style cooldown period - with a 1 week withdraw window to claim at the end.

Bonding withdrawals can exit directly to POD, as the bonding contract is whitelisted to bypass a double cooldown.

The exit-queue from these contracts can be watched live at https://exitqueue.dphn.ai.