← Proposals

DIP-001

Hold & stake Venice DIEM to offset dLLM compute cost

Draft   Status: For review — not approved, not executed.   2026-06-15

1 · Background

The Venice API is USD pay-as-you-go, metered by per-model rate limits — not a subscription. (Venice's consumer "unlimited text" plans are for their chat app, single-user, and can't be resold via the API.) Every token a dLLM user spends on a Venice-served model costs the Treasury real USD. At any volume, that compute cost is our largest variable expense on the resale tier.

2 · The mechanism

• DIEM = a staked compute credit. 1 DIEM, once staked, grants $1/day of Venice API inference, refreshing daily, forever.
• You acquire DIEM by buying it on-market (then staking, 1-day cooldown) or minting it from locked staked-VVV.
• VVV is Venice's governance token; staking it yields a DIEM allocation (~12% APR in credit value).
• Why this fits dLLM specifically: DIEM's yield is denominated in Venice-compute credit. Most holders must sell DIEM to realize value. dLLM consumes it at full $1/day face value by reselling it — so we capture the full yield where a passive holder cannot. This is our edge.

3 · The decisive constraint — DIEM is use-it-or-lose-it

Each staked DIEM gives $1/day that does not roll over. Effective yield = headline yield × daily utilization. Idle DIEM on a low-traffic day is value burned that day. Therefore DIEM must be sized to the reliable daily floor of our Venice consumption — not the peak. Days above the allocation overflow to USD pay-as-you-go (fine); days below waste credit (the only real loss). This is why the program must be measured and volume-led, never a speculative stack.

4 · Economics (snapshot — verify live before any action)

VVV ≈ $17; staking APR ≈ 12%. DIEM ≈ $1,629 and trades volatile (~$1,150–$1,900 range). At $1,629, $365/yr of credit = ~22.4% perpetual yield; bought nearer the range-low (~$1,200) that rises to ~30%. Minting is cheapest historically but the window is effectively closed (~597 DIEM mintable network-wide, ~17 days) — so for us it's buy-and-stake.

Sizing at full daily utilization, buy-DIEM route:

The annual offset is the freed margin. Payback ≈ 4.5 years at full utilization, after which the compute is perpetual and we still hold the (tradeable) DIEM. Underutilization extends payback proportionally.

5 · Acquisition strategy

• Buy + stake DIEM (preferred) — directly the compute credit, no VVV price risk, no mint lockup. ~$4.50 of DIEM per $1/yr offset.
• Accumulate below the 100-day moving average. Since we buy-to-consume, our entry price sets our locked-in compute yield; the daily $1 draw is unaffected by later price swings. Place laddered limit bids at or below the DIEM 100DMA, sized to our growing consumption — this systematically captures a better-than-average entry (higher locked yield) and avoids paying the spikes.
• Optionally stake VVV only if the DAO also wants governance + token upside (an ecosystem-alignment bet). ~$8.40 per $1/yr offset, full VVV volatility, 7-day unstake cooldown. Not required for the offset.
• Never buy ahead of consumption. Cap total DIEM at the proven daily floor; cap any VVV position at ~1× annual compute.

6 · Use of the widened spread

Offsetting Venice cost lifts the spread on resale traffic from the 10–20% markup toward ~100% of retail. Proposed allocation of that surplus, in priority:

1. Worker subsidies — funds our own decentralized open-model supply, the asset the network owns and the path to not depending on Venice. First call on the surplus.
2. More DIEM/VVV — only to keep coverage matched to growing Venice volume.
3. dLLM burn — once the network is standalone-profitable, return value to the token. Not before (burning early just to pump is the pattern we reject).

7 · Risks & honest limits

• Utilization risk (primary): idle DIEM is wasted daily. Mitigated by sizing to the daily floor and the measure-first trigger below.
• DIEM price volatility: DIEM is not a stablecoin — it has swung ~$1,150–$1,900. Mark-to-market swings don't affect the compute we draw (we hold to consume), but principal is at risk if we ever need to exit. Buying below the 100DMA cushions entry.
• VVV volatility: if we also stake VVV, principal risk is larger (−23% in 10 days has occurred). Capped at ~1× annual compute.
• Preliminary system: Venice labels DIEM economics "actively refined"; mint cap, rates, and capacity can change. Re-pull live data before each tranche.
• Counterparty/alignment: we'd hold a competitor's token. At our scale it's immaterial to Venice and net-aligned (same uncensored-AI thesis) — but it is exposure to Venice's survival. The core business must pencil out without VVV/DIEM; this is a bootstrap accelerant, not the foundation.

8 · Staged execution (the actual ask)

1. Now — measure. Run Venice resale on USD pay-as-you-go; track daily Venice supplier cost (live on the admin dashboard). No capital deployed.
2. Trigger. When the 7-day-average daily Venice cost reliably clears ~$10–20/day for 2–3 weeks, authorize a first small tranche sized to the p25 daily floor (~5–10 DIEM), acquired via laddered bids ≤ 100DMA.
3. Measure utilization (target >70%); scale the tranche toward p50 only as overflow-to-USD shows we're under-provisioned.
4. Review quarterly as a treasury position.

9 · Decision points for the DAO

• Authorize the staged DIEM acquisition program (buy + stake), gated on the volume trigger above.
• Approve the 100DMA-laddered acquisition method and the consumption-matched size cap.
• Confirm the surplus-allocation priority (worker subsidies → coverage top-ups → burn).
• Decide whether to additionally stake VVV for governance/upside, or DIEM-only.

Token figures are a snapshot (source: VeniceStats) and move intraday — they must be re-verified at execution time. This document is a draft for review and confers no commitment until formally approved by the DAO. Not financial advice.