Silicon's Cluster 00 represents $2.7M in GPU server infrastructure deployed on RunPod Secure Cloud through our datacenter partner FarmGPU. Over the past eight months, this cluster has generated verified, auditable performance data that has directly informed the design of our pool token and the investment thesis behind Cluster 01.
The cluster currently holds a net asset value of $2,705,991, reflecting server value after depreciation plus total returns earned. Across all 33 servers in the portfolio, key metrics include:
These figures are net of depreciation, operating expenses, taxes, and platform fees. The full breakdown is available on our public dashboard at cluster00.silicon.net.
Diversification drives resilient returns. Cluster 00 spans multiple NVIDIA architectures, from A100s through B200s. Rather than concentrating in a single GPU model, the blended portfolio approach smooths out variance in utilization rates and rental pricing across hardware generations. This finding directly shaped our pool token structure, which is designed around diversified baskets of GPU infrastructure rather than single-asset exposure.
Newer hardware does not guarantee better unit economics. One of the more significant takeaways from the data is that the latest NVIDIA architectures do not consistently produce the highest returns on a cost-adjusted basis. The H100 and H200 server classes occupy a favorable position in the market right now: used servers still under warranty can be acquired at meaningful discounts to new pricing, while hourly rental rates on RunPod Secure Cloud and real-world utilization remain strong. The ratio of acquisition cost to revenue generation is currently more attractive than newer hardware like B200 servers, where purchase prices have not yet depreciated enough to justify the capital outlay relative to yield.
Eight months of operating data replaces assumptions with evidence. Financial models for GPU infrastructure tend to rely on estimated utilization rates and projected rental pricing. Cluster 00 provides actual operating history across market conditions, hardware types, and utilization cycles, giving us a materially different foundation for forecasting than theoretical models alone.
Pre-deposits for Cluster 01 are now open. The portfolio strategy for Cluster 01 is informed directly by Cluster 00 performance data and concentrates on the highest-returning asset classes from the founding cluster: H100, H200, and RTX Pro 6000 Ada servers. Hardware with less favorable cost-to-yield profiles at current market pricing, including B200 servers, will be reconsidered as acquisition costs decline.
Cluster 00 was built to validate the model. Cluster 01 is built to optimize it.
Full performance data: cluster00.silicon.net
Disclaimer: Past performance does not guarantee future results.
