Working on a graph for an idea about “reindustrializing” via a new kind of AI-company-town based on DC energy density trends, plus a wrinkle separating real and solipsistic economies (based on some mix of speculation and structural separation — applies to crypto internal economies, private market fiat IRRs, and AI expected value).

Related: a 3d tokenomics model we’re starting to work on at the Protocol Institute — energy tokenomics (AI), trust tokenomics (fiat —> crypto) and auth tokenomics (access keys ecology basically: identity, API access, commit authority…). Share of GDP, energy density, and gap between real and solipsistic valuations.

This graph is guesstimated but I believe it can be made rigorous. The big interesting features are: AI training is at core industrial levels of energy intensity, and the real/solipsistic split is an ontologically new feature of what’s going on. One might say: microeconomics can stay solipsistic longer than you can stay solvent. Crypto is the clearest case. Half the world still thinks it isn’t real, while a solipsistic fraction of it has people finishing longish careers in it, not even starting. That’s not a bubble, it’s a parallel reality. Increasingly, more and more economic sectors have this quality. They are sustainable escaped realities that are weakly coupled to less escaped realities, which view them as bubbles that will collapse any minute now. They aren’t bubbles. They are undeclared reality shows we misunderstand because there is no Hollywood simultaneously claiming authorship and validating your presumption that it can be safely dismissed as fiction.

The 2 tables are my rough starter analyses of DC energy density diffs with an eye on the narratives that each can sustain. If you can read between the lines, we’re looking at the preconditions for the birth of a new kind of “company town.”

Teaser: a company town can basically be defined in terms of human-to-infrastructure ratio of energy and water use in a given spatially defined population center (call it R). Both intensive measures (watts/rack, gallons of water per household) and full macro measures (gdp, unemployment rate) are the wrong level of analysis. You need the spatial agglomeration meso-level. Company towns basically died in the US (drive through Gary, IN to see a ghost of one), and have to be reinvented. In the west, we are used to, and attached to, low-R agglomerations because all the energy intensive economy was exported to China in pursuit of green theater and quality of life. This is untenable with AI for various reasons, which means company towns (high R) have to be reinvented in the west. Sure some DCs will be put in space or platforms in the ocean or the Sahara desert or in friendly poorer offshore nations, but I suspect needed human capabilities and infrastructure needs are entangled enough that new kinds of company towns will emerge to align political and security needs to economics and livelihoods. But the economies around these towns will be closer to Hollywood than Jamshedpur, the steel-auto town I grew up in. Fundamentally new and weird and new energy-intensive cultural capital production.

The reason “low-R” AI company towns won’t work in the west btw is because you’d need to put them next to big blue-red transition metro-scale exurbs to get R low enough. And you’d either need to densify American lifestyles OR make “industrialized life” appeal to the current exurbs. You do see many major DCs in outer exurbs, but the narrative tension developing is a powder keg.

I’m going to post a protocol fiction bounty around this 🤔

Cc Rohit Krishnan rafa Jasmine Sun curious about your takes on this line of thinking