Dinference.
Same hardware playbook, different revenue. We buy and operate GPU machines that serve LLM inference tokens ~25–35% below OpenRouter. You earn a stable share of the output, paid monthly in USDT at 20% APY, amortized over 24 months. No difficulty adjustments.
Example terms for a $100,000 SAFE note into the Dinference hardware capex round. Payouts are a target, not guaranteed, and depend on utilization and token economics.
Returns Calculator
Projection assumes Dinference captures 10% of current OpenRouter token volume at the cost/price spreads below. Payouts are settled monthly in USDT and amortize principal + yield over 24 months — not a guarantee.
Hardware Allocation
Each Dinference GPU server costs $4,000. Tickets are offered in whole-server blocks so every dollar is matched to a physical machine in the next batch.
Hardware is held by Dinference and serves inference across all investors pro-rata. All shown tiers are fully matched to whole servers.
Model Economics
- GLM-5.229B tok/day · $3.30 cost+$30,030/d25% under OROR $4.40us $4.35
- GLM-5.135B tok/day · $2.84 cost+$24,430/d25% under OROR $3.81us $3.54
- GLM-550B tok/day · $1.80 cost+$19,617/d25% under OROR $2.41us $2.19
- Kimi-K2.610B tok/day · $2.84 cost+$6,666/d22% under OROR $3.62us $3.50
- DeepSeek-V3.217B tok/day · $0.27 cost+$1,980/d39% under OROR $0.45us $0.39
- gpt-oss-120b5B tok/day · $0.28 cost+$572/d34% under OROR $0.43us $0.39
FAQ & Terms
- A SAFE note (Simple Agreement for Future Equity) into Dinference. It entitles you to monthly USDT payouts that amortize your principal plus a 20% annualized yield over 24 months. No conversion to equity is required for the yield stream.