A scaled AI compute specialist can still compound if it converts contracted demand into delivered capacity and proves platform attach beyond raw GPU-hours. Leverage, build timing, and hyperscaler substitution are the gating risks.

CoreWeave can compound into a scaled “AI compute utility + platform attach” by converting its contracted demand/backlog into delivered capacity, then defending pricing with higher-level inference, security, and regulated workloads; the swing factor is whether leverage/capex cadence stays financeable through inevitable build and power bottlenecks.

It sells the scarcest input (reliable AI compute) and can build moats via distribution, trust, and predictable delivery—if it secures power, supply, and financing better than general clouds.

The bull case is not “more GPUs”; it’s higher certainty-of-delivery per MW plus software/platform layers that reduce churn and protect pricing as raw capacity becomes less scarce. CoreWeave already operates at meaningful scale with multi-year customer commitments, so the key to 2030 is execution: on-time campus bring-up, better power/site control, and reduced customer concentration. If those hit, the company can grow into a lower-but-still-premium infrastructure multiple versus pure data-center real assets.

CoreWeave’s upside is real, but it is balance-sheet-and-build-schedule sensitive: power/site delays, rising chip/power costs, or tighter credit can force dilution and compress pricing before software attach matures; hyperscalers are the ever-present substitute if differentiation slips.

No change: still a compute-scaling story where delivery reliability and financing terms determine equity outcomes; federal channel is incremental upside.

Upsized convert and higher perceived credit risk tilt value capture toward debtholders, slightly reducing equity upside despite strong demand signals.

$126.84