RUN: A Neocloud in Your Neighborhood? Early Days, but TAM Could Be Massive

研报英文原文证据摘录

RUN: A Neocloud in Your Neighborhood? Early Days, but TAM Could Be Massive

Clean Energy Equity Research

Numerous Questions Remain Including Long-Term Scalability

While the opportunity appears compelling, several practical challenges remain. Earlier NVIDIA GPU

generations can be air cooled, but future platforms such as Blackwell Ultra and Rubin are expected

to require liquid cooling and, eventually, immersion cooling, both of which are likely impractical in a

residential setting. As a result, we view this opportunity as more of a medium-term than a perpetual

growth driver.

That said, RUN does not need access to the latest GPUs to create value. H100 rental rates have

remained resilient, with pricing in some cases increasing as newer architectures improve efficiency and

free up demand for prior-generation hardware. RUN could also migrate to Blackwell-based systems

while remaining within air-cooling constraints, providing a potentially long runway before more

intensive cooling requirements become a limiting factor.

GPU Obsolescence Is A Key Risk

GPUs depreciate rapidly, with average useful lives typically around five years. Because RUN would

likely utilize prior-generation hardware, economic lives could be even shorter. Our analysis excludes

DD&A, but we would expect depreciation expense and replacement capex to be significant. That said,

with estimated paybacks of just 2-3 years, and potentially lower under more favorable utilization

assumptions, we believe much of the obsolescence risk is mitigated by the speed of capital recovery.

The Opportunity Could Be Massive

While we recognize it is early and there are still a lot of unknowns, we believe the TAM for RUN is

significant. For context, exporting power back to the California grid typically generates about $0.07/