Pinned Tweet
The most important thing I’ve ever learned about wholesale power is that energy is a commodity, capacity is not.
Hyperscalers have the greatest social license to allocate physical resources because their resource allocation yields maximum productivity gains with the least amount of surprise
We continuously move in the direction of humanities best risk-adjusted future when we pursue great economic growth potential through low entropy market frameworks producing maximally predictable outcomes
Replying to @chamath
Pretty much.
As of the past few weeks, debt markets have almost entirely shut off access to money for any company that doesn’t have substantial revenue independent of AI data centers.
SpaceX closed $40B of debt for GPUs, but we’re investment grade and have Starlink & launch revenue.
That means the 200+ neoclouds are going to see guppy-level death and consolidation. Some will make it through, but not many.
Yeah but who split the G?
Introducing The Cheeky Pint Benchmark.
We made our intern watch every episode of Cheeky Pint to bring you this breakdown of pints drank by guests to the nearest sixth.
Congratulations to @elonmusk who is currently in the lead with three full pints of Guinness.
How utility PMs look at you when you’re explaining you’ve purchased 2028 HV CBs for 100 kajillion dollars per unit and their next available outage slot is 2045
The core issue for ERCOT (or any bulk power system) is that the legacy transmission planning process becomes a fundamentally intractable/unverifiable problem for the modern scale of load growth
You cannot produce firm solutions for service to single point loads exceeding 1GW each if the total quantity of load evaluated would double the size of the system
It’s not that we are arriving at the wrong solutions, it’s that we are asking the wrong questions
Sam retweeted
Complex systems and processes are essentially encrypted such that one needs a key in the form of an information structure to productively interact with it
Twitter is the trading pit and open outcry is still the method of exchange for compute
We are so early
Our ability to discern which socioeconomic protocol (i.e. capitalism, socialism, etc) yields the optimal results relies on a Goldilocks zone of stable productivity
Not enough growth and no design is sufficient to sustain the system
Too much growth and any design is sufficient - short run surplus is so significant that ambition to evaluate system sustainability declines and ability to differentiate consequences becomes intractable
Margin from sticky customers (rent seeking from market inefficiency) will soon be a bygone concept
The markets of the future will be extraordinary efficient by today’s standards
A business will create value or die
Many business today exist at a level subordinate to actual value creation where they extract margin from a structural inefficiency in how market rules or norms insufficiently map the complexity of a territory. Market rule “maps” will have a vanishing thin gap in their representation of their territories as we march forward in time
The most difficult part of acknowledging the Ghost in the Machine will be acknowledging that it was always there, and we are an artifact of its self assembly
Tectonic shifts in the infra landscape are currently rewriting the most fundamental tenets of how civilization will harness thermal and information entropy.
We are converging on the terminal networks.















