The Rise of the Neoclouds—Updating Capacity Model (48GW by 2030) + Addressing Recent Debates

Research evidence excerpt

The Rise of the Neoclouds—Updating Capacity Model (48GW by 2030) + Addressing Recent Debates

Industry Update Equity Research

GPU Monetization Rising as Shortages Intensify

CoreWeave: older/existing GPU fleets continuing to see improving monetization as time goes on:

As highlighted on its 1Q earnings call, CoreWeave disclosed that average GPU hourly-rate pricing

increased in the March quarter relative to the December quarter across A100s, H100s, H200s, and

L40s. This marked an improvement vs prior trends, where the company had largely highlighted stable

pricing (not necessarily improving/increasing). On its 4Q25 call, the company noted that H100 pricing

was being recontracted at the end of the year within 10% of where it started 2025 and A100s were

actually up from where they started 2025, while on its 3Q25 call the company noted signing its first

H100 contract that rolled off within 5% of original contracted pricing.

Nebius: still sold out + duration lengthening despite raising GPU prices across older generations:

On its 1Q26 earnings call, Nebius also highlighted continued improvements in pricing across older

hardware as well as longer-duration deals despite the higher pricing. This has led to 1) overall larger

deals for the co, and 2) even greater prepays than originally seen as demand continues to outpace

supply. Similarly, Nebius called out stable pricing on its 4Q25 call (previous generation pricing "didn't

fall") as well as lengthening duration.

Ornn, SemiAnalysis data shows recent spike in contract values: More recent data including data

from 3p vendor Ornn suggests that spot rate pricing and short-term contract pricing (directionally

helpful indicators, but not necessarily a true read into underlying contracted econs) continue to go up,