why Finland halted Google data centers over cleared forest, and why permitting is now as big a bottleneck as power, pushing builders behind the meter #datacenters #google #ai
collide. retweeted
I’m getting an award. Lifetime Achievement from the @EnergyCouncil_
Join me. energycouncil.com/event-even…
[email protected]
Play the video below.
how one social media report on a Pakistani Iran proposal moved crude, and why Dan Pickering compared it to reading chicken entrails #oilprices #crudeoil #iran @pickeringenergy
Iran Mines Hormuz, Trump Turns on Zelenskyy & Venture Global Goes AI Utility | BDE 10.20.26 xitter.cf/i/broadcasts/1jxXgBNQV…
Somewhere a utility is quietly redoing its 2030 load forecast.
🦔Gartner figures the owners of AI data centers need about $8.5 trillion in revenue through 2030 to cover their spending and earn a 12% return. Its forecast for the AI revenue tied to those data centers tops out at $3.6 trillion. The research firm puts more than $3.5 trillion of that spending at risk of write-downs and expects over 80% of neoclouds, the companies that rent out AI computing power, to lose their independence by 2030.
My Take
I think we should focus on Gartner’s point that less than half of AI revenue reaches the companies that own the hardware. Customers mostly pay the AI apps, which rent computing power as cheaply as they can. The chip owners end up with the debt and a thin slice of the money, and CoreWeave’s bonds yielding around 16% show us how lenders already see that setup.
Hyperscalers can take a write-down because their other businesses keep paying the bills. Neoclouds have nothing like that to fall back on. In a slow unwind, a hyperscaler or chip supplier probably buys the bigger ones cheaply and their shareholders eat most of the loss. A sudden stop in funding would be uglier, because some would go bankrupt and the losses would spread to bondholders and the private credit funds that paid for their data centers.
Hedgie🤗








