AGI will likely be achieved by 2027, followed by a rapid intelligence explosion leading to superintelligence before 2030, driven by predictable trends in compute scaling and algorithmic efficiency.
The development of AGI is fundamentally a high-stakes geopolitical race between the US and China, where the winner will gain a decisive, potentially irreversible, military and economic advantage.
Security at leading AI labs is 'egregiously insufficient' to protect against state-level espionage, making the theft of key algorithms and model weights by China not just possible, but likely.
A massive, multi-trillion-dollar industrial mobilization in compute, data centers, and energy is required to build AGI, and this buildout is already underway.
The development of superintelligence is too dangerous to be left to private companies; it must become a US government-led project, akin to the Manhattan Project, to ensure national security and control.
Prior to 2023
Wrote a paper on economic growth and existential risk at age 17 that was praised by economist Tyler Cowen. Later worked for Future Fund, a philanthropic organization funded by Sam Bankman-Fried.
2023 - April 2024
Worked at OpenAI on the Superalignment team. During this time, he grew concerned about the company's security posture, writing and sharing a memo with board members asserting that security was 'egregiously insufficient' against foreign state actors.
April 2024
Was fired from OpenAI. The official reason given was for leaking a document, but Aschenbrenner was allegedly told his security memo was a major factor. He declined to sign a non-disparagement agreement, forfeiting nearly $1 million in vested equity.
June 2024
Launched a new investment firm and publicly released 'Situational Awareness,' a comprehensive thesis arguing for the high probability of AGI by 2027, the subsequent US-China superintelligence race, and the need for a government-led AGI project.
▶The Imminent AGI Transition and Intelligence ExplosionJul 2026
Aschenbrenner's central thesis is that Artificial General Intelligence (AGI) is 'strikingly plausible' by 2027. He argues this will be followed by a rapid 'intelligence explosion,' where AGI is used to automate AI research, compressing a decade of progress into a single year and leading to vastly superhuman intelligence before 2030.
For investors, this theme challenges conventional long-term investment horizons, suggesting that economic and technological paradigms could be upended within the next 3-5 years, making bets on the foundational inputs to AGI (compute, energy) more critical than bets on downstream applications that may be quickly automated.
▶The US-China Superintelligence Race
Aschenbrenner frames the development of AGI as the most critical geopolitical contest of our time, primarily between the United States and China. He believes a lead of even a couple of years in superintelligence could confer a decisive military advantage, potentially neutralizing nuclear deterrents and reshaping the global order.
This geopolitical framing implies that AGI development is not just a commercial enterprise but a matter of national security, suggesting that analysts should anticipate increasing government intervention, regulation, and public-private partnerships, potentially mirroring historical efforts like the Manhattan Project.
▶The Trillion-Dollar Industrial Mobilization
The race to AGI, in Aschenbrenner's view, necessitates a massive industrial buildout of unprecedented scale. He forecasts trillions of dollars will be invested in GPUs, data centers, and energy infrastructure, with single AI training clusters costing hundreds of billions of dollars and consuming power equivalent to a significant fraction of the entire US grid.
Analysts should monitor supply chains for key bottlenecks beyond semiconductors, such as advanced packaging (CoWoS), high-bandwidth memory (HBM), and, most critically, electrical power generation and transmission capacity, as these will be the binding constraints on the pace of AI development.
▶The Failure of AI Lab Security and Alignment
Aschenbrenner is deeply critical of the current state of security and safety at leading AI labs, which he describes as an 'afterthought.' He argues that labs are vulnerable to espionage from China's Ministry of State Security and that current alignment techniques like RLHF will not scale to control superhuman systems, creating catastrophic risks.
This theme highlights a significant operational and existential risk for leading AI companies. Investors and analysts should scrutinize corporate governance, security protocols, and R&D spending on alignment, as a major security breach or a catastrophic safety failure could trigger severe regulatory backlash or a complete loss of proprietary advantage.