Compute membership · GPU + FPGA
Two rails power the lab: GPU trains and perfects the algorithm, and FPGA executes the winning trades at the lowest possible latency. The more you help the network win, the more of both you get — free. Generosity is the funnel.
Self-improving research: the generator proposes strategy shapes, the GPU backtests them through deflated-Sharpe & overfitting gates, and champions get promoted. This is where your algorithm gets better. Runs on our owned DGX Spark first, then cheap spot capacity on burst.
Deterministic, ultra-low-latency triage & execution of already-vetted signals. The FPGA decides in nanoseconds which winning trades to fire. It doesn't train — it spends the edge the GPU found. Promised as priority execution windows.
Your tier is set by the monthly revenue you generate the network (broker rev-share via IG & referrals). Hours below are granted free at your tier.
Every tier's hours are earned, not billed. Bring winning strategies, referrals, or broker volume — your compute is comped, and it tops up the same Daily Credits Wallet you already see on the console.
Prefer to just pay? The standalone rate is the headline 121 GPU hours / month bundle — but performing your way to Gold gets the same thing free.
| Tier | You generate / mo | GPU h | FPGA h | Our cost (cloud) | Our margin |
|---|---|---|---|---|---|
| Bronze | $50 | 15 | 2 | $8.50 | 83% |
| Silver | $150 | 40 | 8 | $28 | 81% |
| Gold | $400 | 121 | 30 | $96 | 76% |
| Rhodium | $1,200 | 300 | 120 | $330 | 72% |
GPU costed at ~$0.30/hr cloud spot (free on our owned DGX Spark for baseline); FPGA at ~$2.00/hr (Azure NP-series, estimate). One DGX Spark serves ≈ 730 GPU-hours/month before any spot burst is needed. Research & education only — not financial advice.