Here are recent rough notes from the site.
Notes from recent Odd Lots discussion with CreditSights analyst on utility issues in AI buildout:
3/ Utilities' inventive is to claim higher demand: transmission, substations, and rate base all boost allowed earnings. That drives overbuild risk. if real power use disappoints, utilities lose the AI growth narrative and go back to being slow-growth, regulated, interest-rate-sensitive businesses.
https://www.bloomberg.com/news/articles/2026-02-02/are-we-building-more-ai-datacenters-than-we-need
Notes from recent Odd Lots discussion with CreditSights analyst on utility issues in AI buildout:
2/ Most third-party forecasts count compute only. Grid interconnects include cooling, lighting, and overhead. Once you adjust, the apparent shortage vanishes. Apples-to-apples, the capacity gap that underpins the “AI power constraint” narrative disappears.
Notes from recent Odd Lots discussion with CreditSights analyst on utility issues in AI buildout:
1/ Utilities claim AI-driven power scarcity, but math points the other way: ~45 GW data-center load today, ~90–95 GW by 2030, versus ~140 GW of firm, contracted utility capacity already queued. After de-duping adjustment, committed supply likely exceeds plausible demand growth.
This is a good musing on charisma and why it's ... bullshit.
tl;dr: Being charismatic means being good at social paradoxes.
Charisma Is Bullshit—Or: A Theory of Charisma
https://www.everythingisbullshit.blog/p/charisma-is-bullshit