Leopold Aschenbrenner is someone who combines technical understanding and actionable macro foresight. Born in Germany in 2001, Aschenbrenner’s trajectory is quite extraordinary. He graduated as valedictorian from Columbia University at the age of 19, armed with dual degrees in economics and mathematical statistics. By age 22, he was a key researcher on OpenAI’s elite Superalignment team, working at the very frontier of AGI .
Aschenbrenner’s departure from OpenAI in early 2024, reportedly after raising safety concerns, freed him to share his unvarnished perspective. In June 2024, he published a 165-page manifesto titled Situational Awareness: The Decade Ahead. The document argued that while the masses view AI as a mere technological progression, the reality is a sprint toward AGI that will reorder the global economy and geopolitical balance of power.
But Aschenbrenner is not merely a theoretician; he has put his capital where his convictions lie. He launched Situational Awareness LP, a hedge fund seeded by Silicon Valley heavyweights including Nat Friedman (Meta AI product lead), Daniel Gross (co-lead of Meta Compute), and Stripe co-founders Patrick and John Collison.
The fund’s performance has been staggering. Launched in September 2024 with roughly $225 million, the fund grew its disclosed U.S. equity book to $5.52 billion by the end of 2025, a roughly 22-fold increase in twelve months. As of early 2026, the portfolio’s total return exceeded 100% year-to-date, generating more than 96 percentage points of alpha over the S&P 500.
Crucially, Aschenbrenner is not betting on the AI models themselves. Instead, his portfolio, featuring massive positions in power producers like Vistra and Constellation Energy, and infrastructure plays like Bloom Energy and CoreWeave, reflects a singular thesis: the real bottleneck to AGI is not algorithmic, but physical. He is betting that electricity generation and computing capacity will be the most valuable assets of the coming decade.
Dwarkesh and Leopold’s Conversation in 2024:
The “Situational Awareness” Thesis and the Trillion-Dollar Cluster
At the core of Aschenbrenner’s Situational Awareness is the concept of the “Intelligence Explosion.” He argues that the scaling laws governing LLMs dictate that as we pour exponentially more data and compute into these systems, their capabilities will increase predictably and dramatically. This is not a plateauing technology; it is an accelerating one.
To facilitate this explosion, the physical infrastructure required is almost unfathomable. Aschenbrenner envisions the construction of “Trillion-Dollar Clusters,” massive, gigawatt-scale datacenters that will consume more electricity than small nations. This necessitates a complete overhaul and expansion of the global power grid, a renaissance in nuclear and natural gas power generation, and a supercycle for the specialized equipment that cools, connects, and powers these facilities.
Based on his manifesto, we identify several key investment themes that align with this physical bottleneck thesis:
Power Generation (Nuclear, Natural Gas, Renewables): The energy density required by next-generation AI clusters cannot be met by the existing grid. We need baseload, reliable power at huge scale.
Power Grid & Electrical Infrastructure: Getting power from generation to the datacenter requires massive upgrades in transformers, switchgear, and high-voltage transmission lines.
Datacenter Construction & Cooling Technology: The thermal management of gigawatt clusters requires advanced liquid cooling and specialized engineering.
Energy Infrastructure Equipment: The picks and shovels of the energy renaissance, from natural gas compressors to pipeline components.
Copper & Power Grid Materials: The electrification of everything, compounded by AI, creates a structural deficit in critical conductive materials.
Networking & Connectivity: Moving terabits of data between thousands of GPUs requires advanced optical networking and telecommunications infrastructure.
Hunting for Value in the AI Infrastructure Supercycle
While the market has recognized the AI infrastructure theme, bidding up darlings like NVIDIA and Vertiv to astronomical multiples, we believe it is important to remain disciplined with regard to the price we are willing to pay.
We set out to answer this question: Can we find companies that are direct beneficiaries of the Aschenbrenner thesis, but which are still trading at deep value multiples?
To accomplish this, we cast a global net, screening thousands of companies across the micro-cap ($20M to $500M) and small-cap ($500M to $5B) universes. We filtered these universes through a strict value lens, requiring at least one of the following criteria to be met:
Price to Book (P/B) ≤ 1.5x (and ≤ 1.0x for micro-caps)
Trailing P/E < 15x (and < 10x for micro-caps)
Price to Sales (P/S) ≤ 2.0x (and ≤ 1.0x for micro-caps)
Dividend Yield ≥ 4.0% (and ≥ 6.0% for micro-caps)
This quantitative screen yielded hundreds of discounted candidates. However, cheapness alone is insufficient; we also looked for evidence of fundamental momentum. We deployed an automated research process to analyze the most recent earnings reports, press releases, and earnings call transcripts for every passing company. We specifically searched for management commentary indicating strong demand growth acceleration, record backlogs, and explicit mentions of tailwinds from AI, datacenters, or power infrastructure.
The results were illuminating. We found dozens of obscure, overlooked companies where management is explicitly telling the market that their businesses are inflecting due to the forces Aschenbrenner described, yet their stock prices still reflect stagnation or decline.
Selected Global AI Infrastructure Value Plays
We present nine stocks from a much larger initial list. These companies represent a diverse cross-section of the AI infrastructure themes, trade at compelling valuations based on our stated criteria, and score highly on our demand growth acceleration metrics.
We also include six in-depth PDF documents generated along the way as we conducted our research. These documents contain a larger array of companies, along with relevant direct quotes from recent earnings calls, substantiating the demand acceleration thesis at selected companies.
