Litepaper
Turning curtailed energy into compute you can own.
A short, honest description of what the protocol does, how the accounting works, and what is still undecided.
This document describes a system that is not yet deployed. It explains mechanisms, not achievements. Where a number is undetermined it is marked TBD rather than estimated.
The mismatch
A solar farm produces on the sun's schedule, not the grid's. When local demand and transmission capacity cannot absorb what a site generates, the operator is instructed to curtail, to throttle down or disconnect. That energy is not stored, not sold, and not deferred. It never exists.
At the same time, the constraint on AI infrastructure has moved. It is no longer primarily the supply of accelerators; it is the availability of power and the interconnect queue to deliver it. One industry is discarding energy while the other cannot get enough.
These are the same problem separated by geography. Moving electricity across that distance is expensive and slow, it means new transmission, new substations, and years of permitting. Moving a workload is nearly free. So Watt moves the computers.
The mechanism
Watt places modular data centres behind the meter at generation sites, sized to the curtailment profile rather than the nameplate capacity. Surplus is absorbed at the instant it is produced, which is the only instant it is available.
The protocol turns that physical absorption into a claim anyone can hold:
- Stake $WATT. Staking registers a claim on the network's surplus. It is a position in physical capacity, not a subscription to an emissions schedule.
- Earn Energy Points. At the close of each epoch the network attests how much surplus it actually absorbed, then divides it pro-rata across everything staked while that energy was being produced. Points are denominated in kilowatt-hours.
- Mint compute vouchers. Burn points to mint vouchers in units a buyer recognises, an H100-GPU-hour, a GB-month of storage, a CPU-hour.
- Redeem or sell. Run a job and the voucher burns on delivery. Or list it, and let someone who needs the cycles more pay for the privilege.
Why the budget is published afterwards
Most staking systems commit to an emission rate in advance. That cannot work here, because it would require forecasting how much sun a site will waste. A forecast that is too high mints points against energy that was never absorbed; one that is too low under-rewards the people who financed the capacity. Either way the backing claim breaks.
So Watt inverts it. The oracle publishes an epoch's budget only after that epoch has closed, from measured output. Points are then split by time-weighted stake across the window that already happened.
Two properties fall out of this, and both matter:
- There is nothing to front-run. You cannot position ahead of a number that is only knowable after the window it applies to has ended.
- Flash loans earn nothing. Accrual is a time integral over stake, not a balance snapshot. A position opened and closed in the same block is integrated over zero seconds and earns exactly zero.
Every division rounds down and the truncation residue rolls into the next epoch, so the total points ever issued cannot exceed the total energy ever attested.
Vouchers
Vouchers are ERC-1155. The token id packs the SKU and an expiry bucket together, which makes two vouchers interchangeable if and only if they are the same product expiring at the same time, exactly the right equivalence class for a commodity, and it costs no extra storage.
Compute is perishable. A GPU-hour in a datacentre that has already run is not a thing you can sell, so vouchers expire. There is deliberately no refund on expiry: refunding would turn a voucher into a free option on compute prices, redeem when spot is expensive, expire and reclaim when it is cheap, which drains the treasury by construction.
Redemption
Redemption escrows the voucher, dispatches the job off-chain, and settles when a quorum of attestors signs off on what was actually consumed. Unused units are returned. Nobody can verify from on-chain data that a GPU ran, so the attestors are trusted for correctness, but not for liveness: a timeout lets the user unilaterally reclaim an escrowed voucher if settlement never arrives.
Marketplace
The secondary market is a signed-order book rather than an AMM. A voucher decays deterministically toward its expiry, so a liquidity provider in one would be short theta with no premium, a guaranteed loss. Orders are signed off-chain, filled on-chain, and the maker keeps custody until the moment of the fill.
The token
$WATT launches on Robinhood Chain as a fixed-supply ERC-20 with no mint function. This is a deliberate constraint with a real consequence worth stating plainly: because the launch venue owns the liquidity pool, the protocol earns nothing from swap fees. Revenue comes from the contracts we do control: a fee on voucher minting and a fee on marketplace fills.
Supply, fee levels, epoch length and the unbonding period are TBD and will be published before deployment.
Status
Contracts are written and tested but not audited and not deployed. No data centre is operating. No energy partner is contracted. Anything on this site describing physical infrastructure describes an intended operating model, not an existing one.
$WATT is a utility token for accessing compute capacity within the Watt network. Nothing in this document is an offer of securities or financial products, or investment advice. Access and participation may be restricted by law in your jurisdiction. You are responsible for determining whether you are permitted to participate.