A Standard Proving Unit is one cryptographic proof, built to a stated type, circuit size, and speed. Proving networks list their raw hardware capacity for sale. Routing agents buy SPUs on the fly to keep payloads private. Every time one sells, it creates a Tre’gent receipt. That receipt uses Ed25519 signing, you can check it offline, and SHOD has to clear it first.
Zero-knowledge proving is what makes something checkable as real and private at the same time, without showing anyone the underlying data. It takes heavy math and dedicated GPU or ASIC hardware to run. As regulated payment rails, identity proofs, and policy checks scale up, demand for this kind of proof starts to look a lot like demand for electricity. It comes in bursts, it's time-sensitive, and it gets paid for one unit at a time. The SPU is that unit. The proving network is the seller. The agent is the buyer.
GPU and ASIC fleets running Groth16, PlonK, Halo2, and whatever comes after them. Operators list their capacity. Every sale settles on its own.
Routing, payment, identity, and policy-checked workflows. Each call buys exactly the proof type it needs, nothing more.
The spot price moves with how busy the network is, how complex the circuit is, and how fast you need it. A forward curve builds up as fleets list their future capacity.
Every sale creates a Tre’gent receipt. That gives the agent proof the work was actually done to spec. No need to trust the provider's word for it.
Crude oil trades in different grades, like light sweet and heavy sour. SPU works the same way. It trades as different technical specs. Two of them are specified today on the Manuka Exchange, which is at sim stage with live mainnet routing not yet enabled.
This is one Groth16 proof on the BN254 curve, with a circuit size up to 2^20 constraints, targeting a 5-second p99 speed SLA. It's the default proving unit for keeping transactions private on Base and EVM Layer-2 networks. Routing agents use one SPU-GROTH16-BASE to keep a single payment payload private.
| Symbol | SPU-GROTH16-BASE |
|---|---|
| Proof system | Groth16, BN254 curve |
| Circuit ceiling | 2^20 constraints |
| Latency SLA | p99 ≤ 5,000 ms |
| Verification cost | ~3 pairings, fixed-time |
| Unit | 1 SPU = 1 valid proof meeting spec |
| Receipt | Tre’gent · Ed25519 · offline-verifiable |
This is one PlonK-family proof with a universal trusted setup, built on hardware-accelerated provers tuned for Aleo and ZK-EVM systems. It costs more to generate than Groth16, but it skips the per-circuit setup ceremony, so it's a better fit for general-purpose private computing. Routing agents buy SPU-PLONK-ALEO when the payload shape changes often, or when the audit system requires a universal setup.
| Symbol | SPU-PLONK-ALEO |
|---|---|
| Proof system | PlonK-family, universal setup |
| Circuit ceiling | 2^22 constraints |
| Latency SLA | p99 ≤ 12,000 ms |
| Verification cost | Constant in proof size, succinct |
| Unit | 1 SPU = 1 valid proof meeting spec |
| Receipt | Tre’gent · Ed25519 · offline-verifiable |
Here's how it works. A proving operator lists capacity for a time window. A routing agent submits a payload that needs a specific proof type. The exchange matches them up. The operator generates the proof. That proof gets hashed and tied to a Tre’gent receipt. Then SHOD runs six checks on the operator: allowlist, daily cap, recipient cap, price window, trust tier, and anomaly detection. If all six pass, the receipt gets signed and the SIU-equivalent USDC settles right away. If even one check fails, the receipt never gets signed and nothing settles. The buyer ends up holding a receipt they can check offline in seconds against the published issuer key.