Leopold Aschenbrenner's portfolio has surged again and again, and as a rising star in the hedge fund world, his investment logic is being validated in reverse by the market trends.

In the past few days, multiple stocks in Leopold's Situational Awareness LP publicly disclosed holdings have collectively risen: stocks such as Bloom Energy, Cipher Mining, Intel, Applied Digital, SanDisk, and IREN saw daily gains exceeding 10%, prompting the market to revisit his 13F report from the end of last year, trying to understand why this former OpenAI researcher had accurately predicted the AI infrastructure sector in advance.

What makes him worthy of attention is not the "young" or "get-rich-quick" narrative, but that he has provided a framework different from mainstream AI trading. Most people equate AI investment with Nvidia, Microsoft, OpenAI, and model capabilities, but Leopold's portfolio has avoided the most crowded star assets, turning instead to Bloom Energy, CoreWeave, Core Scientific, Lumentum, Intel, Bitcoin mining companies, and power-related firms.

The AI narrative is shifting from "whose model is stronger" to "who can support the continued expansion of the model." Training and inference require GPUs, GPUs require data centers, data centers need power, land, cooling, fiber optics, permits, and long-term power supply contracts. Leopold bets on the physical bottlenecks that must be passed for AI to continue growing. Fortune summarized his latest holdings as: this former OpenAI researcher is translating his AGI thesis into multi-billion dollar bets on power, AI infrastructure, and crypto mining companies.

Earlier in March, Beating conducted a deep dive into Leopold and his fund's holdings and investment logic, sharing his vision for the future of AI competition. All of this is now being validated in reality: the AI narrative is retreating from the models on screens back to the land and power grids beneath our feet. The most expensive thing in the future may not be algorithms, but the physical world that supports the continued expansion of algorithms.

The following is the original content from Beating:

In February 2026, the hedge fund Situational Awareness LP submitted its quarterly holdings report, which revealed that as of the end of the fourth quarter of 2025, the total market value of its U.S. stock holdings was $5.517 billion.

Wall Street manages trillions of dollars in assets, and $5.5 billion is just a drop in the ocean. However, this fund's management scale was less than $400 million just 12 months ago, and its founder and chief investment officer is a young man born in 1999.

His name is Leopold Aschenbrenner. 27 years old.

In 12 months, he grew this fund from $383 million to $5.517 billion, an increase of over 14 times. During the same period, the S&P 500 only saw single-digit growth.

What’s more surprising is his holdings. Opening the quarterly holdings report, you won't find any AI star companies you always see in financial news headlines. Instead, there are companies making fuel cells, Bitcoin miners just climbing back from the edge of bankruptcy, and chip giants being abandoned by the entire market.

He says his fund invests in AI, but it doesn't look at all like an AI fund's holdings; it looks more like a madman's shopping list.

But this madman happens to be one of the earliest and most insightful people in the world who understands how AI will change the world. Before joining Wall Street, he was a researcher at OpenAI, responsible for pondering how to ensure that when AI is smarter than humans, it doesn't go out of control; later, he was ousted for saying things he shouldn't have, writing a 165-page manifesto foretelling a future that most people would find absurd.

Later, he gambled his entire fortune on it.

Deciphering the $5.5 Billion: What Did He Buy?

To understand how brilliant Leopold Aschenbrenner is in investing, the most direct way is to open his holdings report and read it line by line.

His largest position is Bloom Energy, with a holding value of $876 million, accounting for 15.87% of the total position.

This company makes fuel cells. More specifically, it produces a type of device called "solid oxide fuel cells," which can directly convert natural gas into electricity with extremely high efficiency. Founder KR Sridhar was once an engineer for NASA's Mars exploration program and has been called "one of the top five futurists creating the future today" by Fortune magazine.

An AI fund placed its biggest bet on a power generation company.

According to Gartner, global power consumption of AI-optimized servers is projected to soar from 93 terawatt-hours in 2025 to 432 terawatt-hours by 2030, nearly a fivefold increase in five years. The power demand from U.S. data centers is expected to triple by 2030, reaching 134.4 gigawatts. Meanwhile, the average age of U.S. power infrastructure is already over 25 years, with many components between 40 to 70 years old, far exceeding their designed lifespan.

In other words, the power required by AI is greater than what the entire power grid can supply, while the grid itself is aging and nearly falling apart.

The scarcest resource in the AI era is not chips, but electricity.

Bloom Energy's fuel cells can precisely bypass this bottleneck. They do not need to connect to the grid; they generate power next to data centers, 24 hours non-stop. In 2025, Bloom Energy secured a contract from CoreWeave to supply fuel cells for its AI data center located in Illinois.

Speaking of CoreWeave, it just happens to be Leopold's second largest position.

He holds $774 million in CoreWeave call options, along with $437 million in common stock, totaling over $1.2 billion, which constitutes 22% of his total position. CoreWeave is a GPU cloud service provider that transitioned from cryptocurrency mining.

In 2017, Mike Intrator, Brian Venturo, and a few others banded together to mine Bitcoin. When the crypto market crashed in 2018, they could no longer mine. However, they had a pile of GPUs. In 2019, they had a lightbulb moment: GPUs could not only mine but also run AI.

Thus, the company transformed from a mining operation to an arms dealer for AI computing power. On March 27, 2025, CoreWeave went public on Nasdaq, raising $1.5 billion at a price of $40 per share. A company that crawled out of a mining operation became a core supplier of AI infrastructure.

Leopold is keen on the vast number of GPUs that CoreWeave has and its deep binding relationship with Nvidia. In an era where computing power equates to productivity, whoever has GPUs holds the power.

But the truly bewildering part is his third largest position: Intel. It has a holding value of $747 million, all in call options, accounting for 13.54% of his total position.

In 2025, Intel is one of the least favored companies on Wall Street. Its stock price has halved from its peak in 2024, with market share being eaten up by AMD and Nvidia, and the CEO has changed repeatedly. Almost all analysts are declaring Intel's demise.

Yet Leopold chose this moment to load up on call options aggressively. It is an extremely aggressive move; if he's right, it soars; if wrong, it zeros out.

What is he betting on? Just two words: foundry.

In November 2024, the U.S. Department of Commerce announced that Intel would receive up to $7.86 billion in direct funding through the CHIPS and Science Act. The purpose of this funding is clear: to make Intel a domestic chip foundry that competes with TSMC.

Against the backdrop of decoupling tech between China and the U.S., America needs a "homegrown" chip builder. Although Intel is lagging, it is the only choice available. Leopold is not betting on Intel's technology but on the national will of the United States.

The following holdings are even more interesting. Core Scientific, holding $419 million; IREN, $329 million; Cipher Mining, $155 million; Riot Platforms, $78 million; Hut 8, $39.5 million.

These companies share a common trait: they are all Bitcoin mining firms.

Why would an AI fund invest in a bunch of Bitcoin miners?

It's simple: because Bitcoin miners have the cheapest electricity and the largest data center spaces in the entire U.S.

Core Scientific has over 1,300 megawatts of power capacity. IREN plans to expand a capacity of 1.6 gigawatts in Oklahoma. These miners have long locked in the cheapest electricity resources globally and signed long-term power purchase agreements to survive in the fierce competition for computing power.

And right now, what AI data centers lack most is exactly power and space.

In 2022, Core Scientific filed for bankruptcy due to the crash in the crypto market. It completed its restructuring in January 2024, reducing approximately $1 billion in debt, and was relisted on Nasdaq. Then, it signed a 12-year, over $10.2 billion contract with CoreWeave to convert its mining site into an AI data center. To fully pivot, Core Scientific even plans to sell all its Bitcoin holdings.

IREN (formerly Iris Energy) signed an AI contract worth $9.7 billion with Microsoft, receiving a $1.9 billion advance payment. Cipher Mining signed a 15-year lease agreement with Amazon. Riot Platforms signed a 10-year, $311 million contract with AMD.

Overnight, Bitcoin miners transformed into landlords of the AI era.

Now, let’s put this puzzle together.

Bloom Energy provides power, CoreWeave provides GPU computing power, Bitcoin mining companies offer space and cheap power, and Intel provides domestic chip manufacturing capabilities. Along with the fourth largest position, Lumentum ($479 million, which makes optical components, a core component of interconnecting AI data centers), the ninth largest position, SanDisk ($250 million, data storage), and the eleventh largest, EQT Corp ($133 million, natural gas producer, supplies fuel for fuel cells).

This is a complete supply chain for AI infrastructure.

From power generation, to transmission, to chip manufacturing, to GPU computing, to data storage, to fiber optic interconnection. He has bought a stake in every link.

At the same time, he did another remarkable thing that made this logic even clearer. In the fourth quarter of 2025, he completely liquidated his positions in Nvidia, Broadcom, and Vistra. These three companies happened to be the stars that saw the largest gains in the AI market of 2024.

He also shorted Infosys, one of India’s largest IT outsourcing companies.

Selling the hottest AI chip stocks while buying unwanted power plants and mining operations. Shorting traditional IT outsourcing because AI programming tools are making programmers more efficient, compressing the demand for outsourcing.

Every trade points to the same judgment: the bottleneck of AI isn’t in software, it’s in hardware; it's not in algorithms, it’s in electricity; it’s not in cloud models, it’s in the physical world.

So the question arises: how did a 27-year-old young man form this understanding?

From the Son of an East German Doctor to an OpenAI Rebel

Leopold Aschenbrenner was born in Germany, where both of his parents were doctors. His mother grew up in East Germany, while his father came from West Germany, and they met after the fall of the Berlin Wall. This family carries a historical mark of rupture—Cold War, division, reunion. His obsession with geopolitical competition may find its earliest seeds here.

But Germany couldn’t hold him. In a later interview, he said, "I really wanted to leave Germany. If you are the kid in the class with the most curiosity, wanting to learn more things, the teachers won't encourage you, they will be jealous and try to suppress you."

He calls this phenomenon the "high poppy syndrome," where those who grow tallest are cut down.

At 15, he convinced his parents to let him fly alone to the U.S. and enter Columbia University.

Attending university at 15 is an oddity anywhere. But Leopold's performance at Columbia turned "oddity" into "legend." He majored in economics and math-statistics dual degrees, winning every award he could, such as the Albert Asher Green Memorial Award, the Romine Economic Award, and was a member of the Junior Phi Beta Kappa Honor Society.

At 17, he wrote a paper on economic growth and existential risks. The renowned economist Tyler Cowen remarked after reading it, "When I read it, I couldn’t believe it was written by a 17-year-old. If this were an MIT PhD thesis, I would be impressed."

At 19, he graduated from Columbia University as the valedictorian. This is the highest honor for undergraduates at that institution. In 2021, while the world was still under the shadow of the pandemic, a 19-year-old German boy stood at the Columbia graduation ceremony, representing all graduates to speak.

Tyler Cowen gave him a piece of advice: do not pursue a PhD in economics.

Cowen felt that the academic field of economics had become somewhat "decadent" and encouraged him to pursue something bigger. Cowen also introduced him to the "Twitter weirdos" culture in Silicon Valley—people fascinated by AI, effective altruism, and humanity’s long-term fate.

After graduating, Leopold first went to the Forethought Foundation, studying long-term economic growth and existential risks. Then he joined the future fund founded by SBF, working with core figures in the effective altruism movement, Nick Beckstead and William MacAskill. His title was "Economist affiliated with the Global Priorities Institute at Oxford University."

This experience is significant. It means that before entering the AI industry, Aschenbrenner had spent years systematically pondering one question: what kind of events could fundamentally change the direction of human civilization?

Then, he joined OpenAI.

The specific timing is unclear, but he became part of a special team—the "Superalignment" team. This team was established on July 5, 2023, co-led by OpenAI co-founder Ilya Sutskever and alignment team head Jan Leike. The goal was to solve the alignment problem of superintelligence within four years, ensuring that an AI much smarter than humans would still heed human directives.

OpenAI had promised to allocate 20% of its computing power to this team. However, a chasm existed between promises and reality.

Leopold saw some unsettling things within OpenAI. He submitted a safety memo to the board, warning that the company’s safety measures were "seriously inadequate" to prevent foreign governments from stealing critical algorithmic secrets. The company's reaction was unexpected. The human resources department spoke to him, stating that his concerns about espionage were "racist" and "unconstructive." Company lawyers interrogated him about his views on AGI and the loyalty of his team.

In April 2024, OpenAI fired him under the pretext of "leaking confidential information."

The alleged "leak" was that he shared a brainstorming document about AGI safety measures with three external researchers. Leopold stated that the document contained no sensitive information and that sharing such documents internally to gain feedback is standard practice.

A month later, Ilya Sutskever left OpenAI. Three days later, Jan Leike also departed. The Superalignment team was dissolved, and the promised 20% computing power allocation was never fulfilled.

A team researching "how to control superintelligence" was disbanded by the very company that was producing superintelligence.

The irony of this situation cannot be overstated. But for Leopold, being fired became a kind of liberation. He was no longer employed by anyone and no longer needed to phrase his thoughts carefully in internal memos. He could say what he wanted to say to the world.

On June 4, 2024, he published a 165-page article on a site called situational-awareness.ai. The title was "Situational Awareness: The Decade Ahead."

165 Pages of Prophecy

To understand Leopold's investment logic, you must first read this manifesto because that $5.5 billion portfolio is the financial translation of these 165 pages of text.

The core argument of the manifesto can be summarized in one sentence: AGI (Artificial General Intelligence) has a very high chance of being achieved by 2027.

This judgment sounded like madness in June 2024. But Leopold's mode of argumentation is direct: quantity and scale.

From GPT-2 to GPT-4, AI's capabilities have experienced a qualitative leap, transitioning from that of a preschool child to a smart high school student. Behind this leap is approximately a 100,000-fold (five orders of magnitude) effective computational growth. This growth arises from the stacking of physical computational power, improvements in algorithm efficiency, and capabilities released through "unbinding" of models.

His prediction is that by 2027, a similar magnitude of growth will occur again. In terms of physical computational power, the computational resources used to train the most advanced models will exceed those of GPT-4 by 100 times. In terms of algorithm efficiency, it improves by about 0.5 orders of magnitude each year, accumulating to about 100 times over four years. Coupled with the gains from "unbinding," allowing AI to evolve from chatbots into intelligent agents capable of using tools and acting autonomously, there's another order of magnitude jump.

Three hundreds multiplied together results in another 100,000 times, another qualitative leap—from a smart high school student to surpassing human intelligence.

What makes this article truly compelling is the series of consequences derived from this prediction.

The first consequence: trillion-dollar level computing clusters.

He writes that in the past year, the topic in Silicon Valley has shifted from $10 billion computing clusters to $100 billion clusters, and recently to trillion-dollar clusters. Every six months, there's an additional zero added to the plans in the boardroom. By the end of this decade, there will be hundreds of millions of GPUs put into operation.

This prediction sounded exaggerated in June 2024. But in January 2025, the Trump administration announced the Stargate project, jointly invested by SoftBank, OpenAI, Oracle, and MGX, planning to invest $500 billion in AI infrastructure within four years in the U.S. The initial funding deployment announced was $100 billion. Construction work has already begun in Texas.

The "trillion-dollar cluster" he wrote in his manifesto became the official plan of the White House six months later.

The second consequence: an electricity crisis.

How much electricity do hundreds of millions of GPUs require? Leopold’s answer is that America’s power production capacity needs to increase by several dozen percentage points.

Data confirms his judgment. In 2024, the total capital expenditure of Amazon, Microsoft, Google, and Meta exceeded $200 billion, a 62% increase from 2023. Among them, Amazon alone spent $85.8 billion, a 78% year-on-year increase. In 2025, Amazon’s capital expenditures are expected to exceed $100 billion.

Most of this money is spent on data centers and power infrastructure.

Microsoft even did something unimaginable a decade ago: it signed a 20-year power purchase agreement with Constellation Energy to restart the Three Mile Island nuclear plant.

Yes, it’s the same Three Mile Island that experienced the worst nuclear accident in U.S. history in 1979.

This nuclear plant is set to reopen in 2028, renamed the Crane Clean Energy Center, and will exclusively supply power to Microsoft's data centers. Joe Dominguez, CEO of Constellation Energy, stated, "Providing power to critical sectors, including data centers, requires sufficient, carbon-free, and reliable energy every hour of every day, and nuclear plants are the only type of energy that can continuously deliver on that promise."

When a software company starts to restart a nuclear plant, you know power has shifted from being an infrastructure issue to a strategic resource issue.

The third consequence: a geopolitical competition.

The most controversial part of the manifesto is where Leopold defines the AGI race in terms of a struggle for the survival of the "free world" using nearly Cold War rhetoric. He harshly criticizes the safety measures of America's top AI labs as being utterly inadequate. He urges that AI algorithms and model weights must be treated as the nation's top secrets.

He even predicts that the U.S. government will eventually have to initiate a national-level AGI project similar to the "Manhattan Project."

These statements sparked intense debate. Critics argue that he oversimplifies the complexities of geopolitics, justifying unrestrained acceleration with a panic narrative.

But some believe he spoke the truth. Both Dario Amodei from Anthropic and Sam Altman from OpenAI share his belief that AGI will soon become a reality.

The true value of the manifesto lies not in whether its predictions are 100% accurate, but in the fact that it provides a complete, actionable framework for thinking.

If AGI really does arrive around 2027, then before that,

what does the world need? A massive amount of computing power.

What does computing power need? GPUs.

What do GPUs need? Electricity.

Where does electricity come from? From power plants, nuclear power stations, and Bitcoin mines with cheap electricity.

Where are chips made? At TSMC.

But what if there's a decoupling between China and the U.S.? Then we'd need Intel.

How do data centers interconnect? We need optical components—Lumentum.

Where is data stored? We need storage—SanDisk.

See, that’s the logic behind his holdings report.

The manifesto is the map, and the holdings are the route. Leopold translated this 165-page macro prediction into an investment portfolio that could bet real money. Every purchase corresponds to a point made in the manifesto. Every sell corresponds to an assumption he believes the market mispriced.

But just having a map is not enough. In the real market, you also need one thing: to continue believing you are right when everyone says you are wrong.

This capability was tested rigorously on January 27, 2025.

DeepSeek Shock

On January 27, 2025, the launch of DeepSeek’s DeepSeek-R1 model sent shockwaves of panic across Wall Street. This model's performance is close to OpenAI's o1, but with costs reduced by 20 to 50 times. Even more shocking is that the training cost of its predecessor model, DeepSeek-V3, was reportedly less than $6 million, using restricted, sanctioned Nvidia H800 chips.

The market logic collapsed in an instant.

If the Chinese can train top-notch models with just $6 million and a stripped-down chip, what are the thousands of billions that U.S. tech giants are pouring every year? Do those trillion-dollar computing cluster plans still make sense? Is demand for GPUs about to plummet?

Panic spread like a plague. Nvidia's stock price plummeted nearly 17%, evaporating $593 billion in a single day, the largest single-day market loss in Wall Street history. The Philadelphia Semiconductor Index dropped 9.2%, marking the largest single-day drop since the pandemic panic in March 2020. Broadcom fell 17.4%, Marvell dropped 19.1%, and Oracle declined 13.8%.

The downward trend started in Asia, spread to Europe, and eventually exploded in the U.S. Within just one day, nearly a trillion dollars in market value evaporated from the Nasdaq 100 constituents.

Silicon Valley venture capital godfather Marc Andreessen referred to DeepSeek as AI's "Sputnik moment," stating, "This is one of the most stunning and impressive breakthroughs I've ever seen, and as an open-source project, it's a gift to the world."

For Leopold's fund, this day should have been catastrophic. His holdings were all in AI infrastructure stocks, and the market was questioning all the logic behind AI infrastructure.

But according to Fortune magazine, one investor in Situational Awareness LP revealed that on that day, during the market's panic sell-off, a large tech fund called in to inquire about the situation. They received a five-word response:

“Leopold says it's fine.”

Why was Leopold so calm? Because in his view, the emergence of DeepSeek not only didn’t overturn his logic, it validated it.

In his manifesto, he has a core argument: the progress of AI will not slow down; it will only accelerate.

The improvement in algorithm efficiency is one of the three engines driving AI development. DeepSeek trained a stronger model with less money and weaker chips, which precisely proves that algorithm efficiency is rapidly increasing. The higher the algorithm efficiency, the stronger the AI that can be produced with the same amount of computing power, which will stimulate further demand for computing power rather than diminish it.

To use his framework from the manifesto: DeepSeek does not prove that "we don't need that many GPUs," but rather proves that "every GPU has become more valuable." When you can train better models with less money, you won’t stop; you will train more, larger, and stronger models.

Panic stems from the fear that "demand will disappear." But those who truly understand AI know that a drop in costs never extinguishes demand; it only creates greater demand.

Leopold bought in against the tide of panic. The market quickly proved him right. Nvidia and the entire AI sector rapidly rebounded in the following weeks, returning to levels even higher than before the crash.

In the world of investment, belief is the most scarce asset. Not because it's hard to form beliefs, but because maintaining belief when everyone says you're wrong is almost counterintuitive.

The End of the Physical World

The story of Leopold Aschenbrenner can certainly be simplified to that of a genius youth getting rich quickly. But focusing solely on the money wastes the true value of this story.

What he did right was shift his gaze to the chimneys of power plants, the substations of mining sites, and the fiber optic cables spanning continents while everyone was focused on the code and model parameters on their screens.

In 2024, the whole world was discussing how strong GPT-5 would be, how realistic Sora could produce videos, and when AI would replace programmers. These discussions are certainly important. But Leopold probed a more fundamental question: how much electricity do these things need? Where does the power come from?

This question may sound too simple, but it precisely points to the biggest investment opportunity in the AI era.

AI is growing at an exponential rate, while the physical infrastructure supporting it remains rooted in the last century. Leopold saw this gap. Then, following this gap, he traced back to the end of the physical world. Every step begins from a physical bottleneck, identifying companies that can solve these bottlenecks, then placing bets.

This methodology isn’t fundamentally new. During the California gold rush in the 19th century, the people who made the most money were not the gold miners, but those selling shovels and jeans. Levi Strauss rose to prominence at that time.

However, understanding this principle is one thing; executing it in the age of AI is another.

To execute it, you need to possess two complementary abilities: a deep understanding of technological trends, knowing the pathways and resource demands of AI development; and specific awareness of the physical world, understanding where electricity comes from, how to build data centers, and how to lay fiber optics.

The former necessitates experience in OpenAI’s lab, while the latter requires the willingness to dive deep into researching the power contracts of bankrupt mining companies.

Technologists understand AI but not the electricity market. Financial people understand markets but do not grasp the physical constraints of AI. Leopold happens to have both.

But even more important than ability is perspective.

One often-quoted line in his manifesto is: "You can see the future first in San Francisco." The subtext of this statement is that the future is not evenly distributed.

The essence of investing is finding price mismatches in a future that has arrived but has not yet been evenly distributed.

Leopold has seen the AI capability curve firsthand in OpenAI’s laboratory; he knows GPT-4 is not the endpoint but the starting point, and he knows there will be larger models, more computing power, and crazier capital investments to come while the market continues to discuss whether "AI is a bubble."

This represents the mismatch. What he has done is convert that mismatch into $5.5 billion.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。