First SGP-0003 Simulation Quantifies Protocol Costs: DeFi Routers and CLOB Market Makers Most Exposed

The governance debate over SGP-0003 has run largely on broad economic arguments about resource pricing and validator incentives. A simulation published August 24 by analyst Mostly Data supplies what neither camp has yet produced: protocol-specific cost figures, a quantified SOL burn range, and a systematic look at where the fee change would actually land. With the vote closing around August 27-29, the analysis arrives at the last window it could plausibly influence validator ballots.

SGP-0003 would replace the current flat 5,000-lamport base fee with two components: a fixed 2,500-lamport inclusion fee paid to the block leader, and a variable resource fee set at 0.1 lamports per requested compute unit (CU), entirely burned. For background on how Solana SOL$114.76-3.3% Solana's governance vote was triggered and the procedural dispute over quorum display, see prior Compass coverage of the proposal clearing its stake threshold and the 60% hardcoded quorum discrepancy.

SOL Burn Projections Across Three Behavioral Scenarios

The largest macro effect the simulation identifies is an increase in daily SOL destruction. CryptoBriefing reports that the network currently burns approximately 648 SOL per day through signature fees, and translates Mostly Data's per-epoch projections to a daily range of 1,500-9,000 SOL under the proposal, a 2.3x to 14x increase over the baseline. Mostly Data models three scenarios, explicitly described as "behavioral counterfactuals, not forecasts":

At the optimistic end, 10% resource optimization by network participants produces an estimated 6,000-8,000 SOL per epoch. Matching current actual consumption with perfectly calibrated compute requests pushes that to 8,000-10,000 SOL per epoch. The p75 residual-mispricing scenario, which reflects realistic ongoing compute over-requesting, reaches 30,000-35,000 SOL per epoch.

Jupiter, Titan, and DFlow: Average Per-Transaction Fee Increases

The simulation isolates average per-transaction cost increases for three Solana trading protocols:

Jupiter JUP$0.376+14.5% routes a large share of Solana DeFi volume, so even a small per-transaction figure compounds at scale. Titan and DFlow face proportionally higher per-transaction costs relative to their respective routing and order-flow models. The simulation adds that higher fees raise the minimum profitable alpha threshold for atomic arbitrage, a headwind for the activity that underpins price discovery on central limit order books.

The Compute Over-Requesting Gap

The simulation's most structurally significant finding concerns the relationship between what transactions request and what they actually consume. On aggregate, consumed compute units represent approximately 20% of requested compute units, according to Mostly Data's data. Even transactions that explicitly configure both compute unit limits over-request by around 95,000 CU on average, with the 90th percentile reaching 280,000 CU unused, according to the simulation.

This gap carries no financial consequence under the flat base fee, which charges per transaction regardless of compute footprint. Under SGP-0003, the resource fee is assessed on requested CUs, not consumed ones. A transaction reserving 500,000 CU and using 100,000 would pay the resource fee on all 500,000. Applications that do not recalibrate their compute requests before activation would face a larger cost increase than the simulation's base projections assume, because those projections model at least partial optimization.

CLOB Market Makers and On-Chain Routing Face Compounding Pressure

The simulation identifies CLOB market makers and on-chain routers as the most exposed workloads. Both categories combine high transaction volume with significant compute requirements, the two variables that multiply under a resource fee model. Atomic arbitrage, which depends on narrow cost margins and precise compute budgeting, faces a structural squeeze from the higher minimum profitable alpha threshold.

Transfers, by contrast, are relatively unaffected or marginally cheaper under the simulation's assumptions, because they carry low compute requirements and would benefit from the reduced 2,500-lamport inclusion fee relative to the current 5,000-lamport base.

Validator Economics: Vote Transactions Cheaper, Multisig Revenue Lower

For validators, the simulation finds mixed outcomes. Optimized vote transactions become approximately 12.3% cheaper under the dual-fee model, according to Mostly Data's simulation. Leader protocol-fee revenue holds flat for single-signer transactions but falls for multisig traffic, with an estimated average loss of approximately 0.00015 SOL per block, per the simulation.

A Counterargument From the Simulation's Own Author

In a follow-up post on August 25, Mostly Data raised a point the simulation leaves open: the cost projections do not account for rational priority-fee repricing by market participants. Under SGP-0003, transaction ordering still depends on priority fees. An application could theoretically offset the new resource fee by reducing its priority fee by the equivalent amount, holding total cost flat while maintaining the same effective fee-to-compute ratio and queue position.

Whether trading firms and protocol routers would coordinate this repricing consistently in practice is a separate question. The simulation leaves it unresolved.

What the simulation does not leave ambiguous is the compute inefficiency finding. With network transactions consuming roughly one-fifth of their requested compute on average, SGP-0003's resource fee would create an immediate cost for applications that treat the current flat-fee model as a signal to request liberally. Validator ballots close this week.