one year on
Leopold Aschenbrenner publishes 'Situational Awareness' series arguing AGI by 2027 is 'strikingly plausible'
The former OpenAI superalignment researcher releases a multi-part essay series predicting trillion-dollar compute clusters, a national security mobilization, and superintelligence within the decade, alongside a four-hour podcast with Dwarkesh Patel.
Leopold Aschenbrenner, the former OpenAI superalignment researcher, today published ‘Situational Awareness’, a sweeping multi-part essay series arguing that AGI by 2027 is ‘strikingly plausible’. The document, released at situational-awareness.ai, traces trendlines in compute scaling and algorithmic efficiencies to predict another preschooler-to-high-schooler-sized qualitative jump in AI capability within three years, leading to superintelligence by the end of the decade.
Aschenbrenner details a techno-capital acceleration already underway: clusters that recently cost $500 million are now being planned at $100 billion scales, and he projects trillion-dollar clusters consuming over 20% of US electricity by 2030. He warns that current AI lab security is inadequate against state-actor espionage, and predicts the national security state will launch a government AGI project by 2027 or 2028. The series also argues that unsolved superalignment problems could lead to catastrophe during a rapid intelligence explosion.
Simultaneously, Aschenbrenner appeared on a four-hour episode of the Dwarkesh Podcast, where he elaborated on his thesis and announced the launch of an investment firm backed by Patrick and John Collison, Daniel Gross, and Nat Friedman.
The record
Hosts Aschenbrenner on a four-hour podcast episode, calling him one of the 'most significant' sources for his understanding of AI, and discussing the trillion-dollar cluster and unhobbling.
One year later — open only if you can handle spoilers
Over the following two years, many of Aschenbrenner's specific predictions—such as AGI by 2027—were not realized, but his framing of AI as a national security imperative influenced policy discussions in Washington and prompted increased attention to compute governance and export controls.