Source

The foundational primary source under this workspace's AI-compute-buildout thesis cluster, archived here with full text. Aschenbrenner published this 165-page essay series on 2024-06-04 (updated 2024-06-06); it is the document the ai-power-bottleneck perspective, the ai-power watchlist, and the aschenbrenner-13f source profile all descend from. Until now only the announcement tweet and the downstream 13F were in the corpus — the essay body itself was never archived. This event is that archive.

Why this is primary research for us

Aschenbrenner's aschenbrenner-13f voice patterns — "watts per rack," "the trillion-dollar cluster," "compute scaling laws," "AGI timeline," "energy bottleneck" — are lifted almost verbatim from this essay, principally Chapter IIIa, "Racing to the Trillion-Dollar Cluster." That chapter is the literal origin of our two-clocks frame ("you can order 100,000 GPUs in 6 months; you cannot add 500MW to the grid in 6 months"). When we say the ai-power watchlist was "built from his 13F," the 13F was itself the trade expression of the thesis written here. This document is the root; the 13F, the NuttyCLD AI-power-crisis series, and the Crusoe operator transcript are all downstream corroboration.

The lo

Takeaway

Status: primary source, full text in body below. This is a 2024 forecast — read predictions as the author's dated priors, not established fact. still comes only from precompute summaries, never from this essay. Canonical PDF (165pp, re-fetchable from pdf_url) is kept locally at pdf_local per the repo's "raw heavy bodies stay local" convention; the durable, committed primary research is this full-text event.