Do you really believe these companies are raising debt capital at 2% to 8%…?! Please learn how to value the debt component imbedded in convertible deals. It might surprise (and enlighten) you. $CRWV$IREN$NBIS
And why the company essential to the entire neocloud business model, $NVDA, trades at a substantial discount to the neoclouds themselves. $CRWV$IREN$NBIS
Bernstein models $IREN scaling to ~5.8 GW longer term with cloud revenue reaching ~$17B and adjusted EBITDA ~$13B by 2030.
Interesting that they assume Childress stays flat at 750 MW so almost all of that upside comes from proving Sweetwater design can be repeated across Oklahoma, Australia and Europe.
Thats where the opportunity gets really large if IREN can execute with three-year contract pricing already up 125% even as the note also models ~$65B of incremental CapEx before the full revenue ramp.
1. You are quoting a colo rate, which does not include GPU’s, and, 2. My point is that if Anthropic can’t make money paying $20M per MW on the margin, then the neoclouds have some ‘splainin to do.
So you are comparing ARR at EOY with average compute cost for the past. Got it. You might want to keep in mind that “ARR” is a marginal revenue metric, which should be compared to marginal costs.
Anthropic and OpenAI are now chasing 20 to 30 MW compute deals alongside gigawatt projects as “speed to usable capacity” becomes increasingly important.
That puts already energized sites in a way stronger position as labs look to take whatever capacity can come online fastest.
Yes, but you seem to be confusing business models. Anthropic is renting the “production machines” that are already in place. They rent compute, and turn that compute into tokens-which they sell. The hyperscalers/neoclouds make the “investment”.
The “vast majority” of their compute has just been signed, or will be in the 4th quarter. According to the hyperscaler/neocloud bulls it is not at $10-15M per MW anymore. Someone is getting this narrative wrong.