Leopold Aschenbrenner's investment thesis has been decisively validated by market movements, propelling his hedge fund, Situational Awareness LP, to a total market value of $5.517 billion as of the end of the fourth quarter of 2025. This figure represents a staggering 14-fold increase from the $383 million in assets under management recorded just 12 months prior, a performance that dwarfs the single-digit growth of the S&P 500 during the same period. The fund's portfolio has recently seen significant appreciation in holdings such as Bloom Energy, Cipher Mining, Intel, Applied Digital, SanDisk, and IREN, with many assets posting daily gains exceeding 10%. Data compiled by Woofun AI shows that this surge has forced the market to re-examine the fund's 13F filings, revealing a deliberate departure from the crowded AI narrative focused on NVIDIA, Microsoft, and OpenAI. Instead, Aschenbrenner has constructed a portfolio centered on the physical bottlenecks required to sustain AI growth: electricity, data center real estate, and semiconductor manufacturing capacity.

The core of this strategy lies in the recognition that the AI narrative is shifting from algorithmic superiority to infrastructural capacity. Training and inference require GPUs, which demand data centers, which in turn require massive amounts of electricity, land, cooling systems, fiber optics, and long-term power contracts. Aschenbrenner's largest holding is Bloom Energy, valued at $876 million or 15.87% of the total portfolio. This company specializes in solid oxide fuel cells capable of converting natural gas into electricity with high efficiency, bypassing the aging U.S. power grid which has an average component age exceeding 25 years. According to Gartner projections, global electricity consumption for AI-optimized servers is set to surge from 93 terawatt-hours in 2025 to 432 terawatt-hours in 2030, a nearly fivefold increase. Bloom Energy's ability to generate power on-site for data centers, evidenced by a 2025 contract with CoreWeave for an Illinois facility, positions it as a critical solution to this impending energy crisis.

Complementing the power generation bet is a massive allocation to CoreWeave, Aschenbrenner's second-largest holding. The fund holds $7.74 billion in CoreWeave call options alongside $4.37 billion in common stock, totaling over $12 billion or 22% of the portfolio. CoreWeave, originally a Bitcoin mining operation founded in 2017, pivoted to become a GPU cloud provider after the 2018 crypto market crash rendered mining unprofitable. The company went public on Nasdaq on March 27, 2025, raising $1.5 billion at $40 per share. Woofun AI notes that Aschenbrenner's conviction in CoreWeave stems from its extensive GPU inventory and deep ties with NVIDIA, effectively making it an arms dealer for AI computing power in an era where compute is the primary driver of productivity. This pivot from cryptocurrency mining to AI infrastructure highlights the fluidity of asset utility in the current technological landscape.

Perhaps the most controversial aspect of the portfolio is the $7.47 billion position in Intel call options, representing 13.54% of the fund. In 2025, Intel was widely viewed as a failing entity, with its stock price halved from 2024 peaks and market share eroded by AMD and NVIDIA.

However, Aschenbrenner's bet is not on Intel's technology but on the geopolitical imperative of U.S. chip sovereignty. In November 2024, the U.S. Department of Commerce announced up to $7.86 billion in direct funding for Intel under the Chip and Science Act, aiming to establish a domestic foundry to compete with TSMC. This move underscores a strategic wager on national will rather than corporate performance, anticipating that the U.S. government will prioritize local manufacturing capabilities amidst ongoing tech decoupling with China.

The portfolio further extends into the Bitcoin mining sector, with significant holdings in Core Scientific ($419 million), IREN ($329 million), Cipher Mining ($155 million), Riot Platforms ($78 million), and Hut 8 ($39.5 million). These companies are not held for their Bitcoin production but for their access to cheap electricity and vast data center spaces. Core Scientific, which emerged from Chapter 11 bankruptcy in January 2024 after shedding $1 billion in debt, secured a 12-year, $10.2 billion deal with CoreWeave to convert its mining facilities into AI data centers. Similarly, IREN signed a $9.7 billion AI deal with Microsoft, Cipher Mining inked a 15-year lease with Amazon, and Riot Platforms agreed to a 10-year, $311 million deal with AMD. Woofun AI analysis suggests that these Bitcoin miners have effectively become the landlords of the AI era, possessing the power purchase agreements and physical infrastructure that new entrants cannot easily replicate.

Aschenbrenner's logic is completed by liquidating positions in the previous AI darlings, including NVIDIA, Broadcom, and Vistra, while shorting Infosys. This contrarian move signals a belief that the bottleneck for AI is no longer in software or chip design but in the physical world. The fund also holds Lumentum ($479 million) for optical components, SanDisk ($250 million) for data storage, and EQT Corp ($133 million) for natural gas fuel. This comprehensive supply chain approach covers every link from power generation to transmission, chip manufacturing, and interconnection. The strategy was stress-tested on January 27, 2025, when the release of DeepSeek's DeepSeek-R1 model caused a market panic, wiping out nearly $1 trillion in Nasdaq 100 value in a single day. While the market feared a collapse in GPU demand due to algorithmic efficiency, Aschenbrenner remained calm, arguing that cost reductions would only accelerate demand for compute power, a view that was quickly vindicated as the sector rebounded.

The intellectual foundation for this portfolio is Aschenbrenner's 165-page manifesto, "Situational Awareness: The Decade Ahead," published in June 2024 after his dismissal from OpenAI's Superalignment team. The document predicts that Artificial General Intelligence (AGI) will be achieved by 2027, driven by a 100,000-fold increase in effective computation. Aschenbrenner argues that this growth will necessitate trillion-dollar computing clusters and a fundamental restructuring of the U.S. energy grid. His time at OpenAI, where he warned of inadequate security measures and was subsequently fired for "leaking confidential information," solidified his belief that the race for AGI is a geopolitical struggle for survival. This perspective led him to view the physical infrastructure supporting AI not merely as a business opportunity but as a critical national asset.

Ultimately, Aschenbrenner's success stems from his unique ability to bridge the gap between high-level technological forecasting and granular physical reality. While technologists understand AI algorithms and finance professionals understand market dynamics, Aschenbrenner possesses both, allowing him to identify price discrepancies in a future that has already arrived but is not yet fully priced in. By focusing on the smokestacks of power plants, the substations of mines, and the fiber optic cables spanning continents, he has capitalized on the divergence between the exponential growth of AI and the linear constraints of the physical world. This methodology, reminiscent of the 19th-century gold rush where shovel sellers outperformed miners, has transformed a $383 million fund into a $5.517 billion powerhouse, proving that in the AI era, the most valuable assets may well be those that keep the lights on.