🚨 Nearly 7 in 10 GPU rentals on $TAO Lium SN51 are now launched by code and agents, not humans. And the numbers keep getting better. Lium, last 30 days (10 Sep to 9 Oct), per @lium_io: • GPU spend: $1.40M (+45%) 🚀 • Annualized run rate: $19.3M 💰 • Rentals: 17,497 (+106%) 📈 • Paying customers: 1,379 (+16%) ⬆️ • New paying customers: 1,060 (+32%) • New accounts: 6,329 (3.4x) ⬆️ • Launched by code and agents: 68.7% (up from 63.0%) ⬆️ • Rental failures: -32% ⬇️ • Providers: +15% ⬆️ Zoom out. Monthly spend went from about $709k to $1.40M in two months. That is roughly 2x. Net dollar retention is 112%. Customers who were already there spent 12% MORE than the month before. People are not just trying it, they are scaling on it. Right now: • 1,722 GPUs on the network • 1,340 rented (77.8%) • 13 GPU types at 100%, including every B300 • 380 RTX 5090s at 97.6% Lium says it is fully rented out of most GPU types and is calling for providers. They pay 95%+ of rental fees, even while unrented. Supply is the bottleneck, and that is the best problem a marketplace can have. Overall utilization is 77.8%, so not everything is sold out $1.4M a month is small next to the hyperscalers But look at the direction. Real customers are paying real dollars, with a public Grafana to check it. Every subnet gets underestimated, and the market keeps asking: Where's The Revenue? This is the answer. Sources: @lium_io month in review, grafana.lium.io GPU demand dashboard $TAO / Not financial advice, always DYOR.