Henry

CoreWeave $CRWV results look pretty strong and thought management did a good job addressing credit concerns on the call. No plans to raise unsecured paper in 2027 and instead relying on prepayments and borrowing against the GPUs to stand up compute. Amperes and Hoppers are pricing 30-50% higher QoQ with short-term capacity selling above $40M/MW/yr (note $SPCX is at $50M/MW/yr on later gen GPUs). 130,000 GPUs are renewing in 2027 and expected to generate an incremental $4B in annualized revenue w/ no additional CapEx. 170,000 GPUs are renewing in 2028 and are expected to contribute an incremental $6B in annualized revenue w/no additional CapEx. That's a total of $10B in incremental revenue on the existing fleet by YE 2028 which would double LTM revenue alone. High visibility there as management claims line of sight on 10x as much qualified demand as available capacity through 2027 which is also 'several times larger' with respect to 2028. No expectation of any change in business due to the mid-term election results which featured a Democrat sweep. View no disruption to existing, permitted pipeline and describe 'building responsibly' as a long-standing company objective. Management did note they have an additional 1.3GW outside the United States in Indonesia, India, Spain, Denmark, Canada. As the industry shifts towards prepayments investors should have a lot more certainty into which neoclouds have the serious backing and confidence of the hyperscalers / AI labs. For example, Anthropic has agreed to buy $13.7B of compute from Rumble $RUM over the course of 6-years for ~150MW which is about $15M/MW/yr (depends on total size). That looks like a steal versus the $40M/MW/yr CoreWeave is charging. The SpaceX strategy has been to always sell at spot demand prices vs long-term contracts. But can Rumble actually deliver? I think they can, but this deal will be financed by Rumble w/ debt & equity rated against Anthropic's credit rating. CoreWeave expects to receive ~$9B in prepayments over the next three months. The real test to determine whether Rumble and other neoclouds are treated as serious counterparties in their ability to stand up compute fast now centers on prepayments. Neoclouds that receive complete prepayments from hyperscalers / AI labs investors should now implicitly assume are deemed as extremely capable operators. @GavinSBaker suggests investors can measure this via Operating Cash Flow (OCF) to EBITDA. Prepayments boost your operating flow (cash in the door) that will scale ahead of the profits that come with standing the compute online. Overall, supply remains extremely short, and spot pricing continues to rise. CapEx at the hyperscalers / AI labs is going to continue to surge as more prepayments are made. In my opinion, this change in funding structure is an incredible tailwind for the neoclouds and it shows you how all in on AI the biggest companies in the world truly are.