Author: Curry, Shenchao TechFlow
Every time someone makes a big profit in the U.S. stock market, the first thing onlookers do is always the same: check their portfolio holdings to find the next stock to buy.
The most frequently translated report recently belongs to a 24-year-old German named Leopold Aschenbrenner.
In March this year, domestic media extensively covered him, with headlines that were all similar: the genius fired by OpenAI, who wrote a 165-page paper predicting AI trends and founded a hedge fund managing $5.5 billion...
But tags are just tags. What truly sets this fund apart is that it doesn’t buy NVIDIA, doesn’t buy OpenAI, and doesn’t buy any company building AI models. It only invests in what AI cannot live without: power generation, chip manufacturing, optical communications, data centers...
In his own words, the bottleneck for AI is not in the algorithms, but in electricity and computing power. The entire fund is betting that this statement is true.
A social media investment influencer has been called the "son of the U.S. stock market in the AI era" or the "AI version of Warren Buffett." Recently, this title has been resurfaced because his predictions have become somewhat extraordinary.
According to data released by the copy trading platform Autopilot on May 1, the simulated portfolio he follows increased by 61% over two months. Based on this, his fund size is now approaching $9 billion.
Where does the money come from? Primarily from two heavily weighted stocks. Bloom Energy, a fuel cell company providing off-grid power to AI data centers, has risen 239% year to date.
According to the position report disclosed at the end of last year, he held $875 million in stocks and options of this company, now valued at nearly $3 billion.
Also Intel. The same position report shows that in Q1 2025, he purchased 20.2 million call options on Intel, when Intel’s stock price was around $20, and Wall Street’s mainstream view was that Intel was not doing well.
Last week, Intel rose to $113, reaching a 25-year high. In less than a year, its price nearly quintupled, and this young investor’s options return multiple was even more extraordinary than the stock’s.
I understand the crowd's impulse. The U.S. investment website Motley Fool published four articles in one day dissecting his holdings, and overseas Reddit investment forums are debating whether to copy his portfolio. Everyone is trying to find the next Intel in his holdings report.
But you should know that position reports typically have a 45-day delay. By the time you see what he bought, the market move is already halfway done.
More importantly, even if you know his positions in real time, you cannot replicate the reason he consistently gets it right.
The circle is the greatest Alpha.
First, what made Leopold Aschenbrenner most impressive was his paper written in 2024, which almost predicted the current direction of AI development and investment trends.
The core argument can be summarized in one sentence: AI model training compute increases by about half an order of magnitude per year; at this rate, AGI with human-level capabilities will emerge around 2027.
But to sustain this growth rate, the key constraints are not at the algorithmic level, but in electricity, chip production capacity, and physical space. The power consumption of a single training cluster will jump from megawatts to gigawatts, approaching the output of a large nuclear power plant.
This is the core logic behind his entire fund. The pace of AI development is determined by physical bottlenecks, so you should invest in the bottlenecks themselves.
This conclusion sounds like the result of extensive research and reasoning by a smart person in their study; but in reality, I believe it was the circle he was in that shaped this judgment.
Before writing his paper, he spent a year working on OpenAI’s Superalignment team, which is dedicated to researching how to control AI smarter than humans, reporting directly to Chief Scientist Ilya Sutskever.
That year, he saw internal training programs, actual power consumption, and the specific electricity and chip requirements of next-generation models. When he wrote "gigawatt-level power consumption" in his paper, his basis may have been the internal roadmap from the lab.
In April 2024, he was fired from OpenAI after writing an internal memo to the OpenAI board warning that the company’s security measures were inadequate and that it faced risks of infiltration by foreign intelligence agencies.
This memo sparked tension between management and the board, and OpenAI subsequently fired him for "leaking information."
Two months later, the paper was published. Rather than being an independent study, it should be understood as the public version of his internal thinking at OpenAI.
AI papers have solved the problem of "which direction to look." But for investing, knowing the direction alone is far from enough.
AI requires more electricity—a judgment many analysts made as early as 2024. What truly matters is timing and position size, such as whether you dare to invest $20 million in call options when Intel’s stock is at $20.
This confidence comes not just from believing in the broad trend of AI, but from knowing precisely which companies are signing large power purchase agreements, which data centers are expanding, and the actual scale of demand.
And the fund Situational Awareness, founded by Leopold Aschenbrenner, has investors who happen to be seated in the front row of these decisions.
The fund's LPs include the two founders of Stripe, a company that processes payment volumes for the majority of Silicon Valley's tech companies and can directly sense the acceleration in infrastructure spending;
Another investor is Nat Friedman, former CEO of GitHub and current head of product at Meta AI, who is involved daily in decisions regarding compute resource procurement.
In addition to initial capital, they provided the fund with a continuously updated information channel.
In addition, the research director at his fund is also a key figure on this chain. Carl Shulman, a veteran in AI safety, previously worked at Peter Thiel’s hedge fund Clarium Capital, where he specialized in translating insights from the AI community into actionable trading strategies.
His portfolio also includes an often-overlooked crypto niche.
According to last year’s portfolio report, he established positions in CleanSpark and Bitfarms, both Bitcoin mining companies transitioning their BTC mining facilities into AI computing centers.
Crypto mining facilities naturally have large-scale power access and cooling systems, which are precisely the most scarce resources for AI data centers.
Interestingly, he is no stranger to the crypto industry. In 2022, he worked for nine months at the Future Fund, the charitable foundation founded by SBF, and left just before FTX’s collapse.
It is unknown whether this experience directly influenced his judgment of mining companies. However, it is certain that he is among the few who have deeply engaged with both the cryptocurrency industry and cutting-edge AI laboratories. This intersection itself represents a rare position of insight and potential for networking.
Another detail: his fiancée, Avital Balwit, is the chief of staff to Dario Amodei, CEO of Anthropic, the parent company of Claude and OpenAI’s most direct competitor.
He worked at OpenAI, and his fiancée works alongside the CEO of Anthropic—two companies at the forefront of the AGI race, giving him hands-on experience at one and daily exposure to the other.
Last year, Fortune magazine in the United States interviewed more than a dozen insiders who had interacted with him and concluded that he is highly skilled at “packaging ideas brewing in Silicon Valley labs into compelling narratives.”
I think this description is too polite. What he actually did was more direct: placing bets in the public market using insights gained from private circles. The AI papers he published are sanitized versions; his own investment fund holds the complete version.
A feedback loop that outsiders cannot enter
Looking back, Leopold Aschenbrenner’s fund chose a less common structure.
Most AI funding follows the venture capital route—investing in early-stage companies in the hope that one will become the next OpenAI. He did not take this path. According to Fortune, he explicitly rejected the VC model when founding his fund, arguing that the impact of AGI is too profound to be fully expressed outside the most liquid public markets.
This choice itself reveals a consensus within his circle: the greatest investment opportunity in the AI era may lie within established companies that already possess physical infrastructure.
It could be a fuel cell company with existing power access, a chip giant with wafer fabrication lines, or a Bitcoin mining enterprise with mining farms and cooling systems. These companies have been publicly traded for years and have strong liquidity, but most analysts are still valuing them using outdated frameworks that have not seriously incorporated the variable of "AI infrastructure as a necessity."
This is his arbitrage opportunity.
People in the industry already understand the pace and scale of AI infrastructure expansion, while public markets are still pricing based on outdated logic. The gap between them is the source of profit.
This informational advantage has another characteristic: it is self-reinforcing.
The better the fund's returns, the more industry insiders are willing to become LPs. The more LPs there are, the more concentrated the fund’s access to decision-makers’ insights becomes. The more concentrated the insights, the higher the accuracy of investments. This is a positive feedback loop, and for outsiders, the barrier to entering this cycle only grows higher.
Of course, this cycle also has its vulnerabilities. Highly concentrated positions combined with significant leverage mean the entire fund is extremely dependent on a single narrative. As long as the premise that "AI infrastructure continues to expand" holds true, everything runs smoothly.
But if the pace of AI development slows down, or if energy bottlenecks are bypassed by some technological breakthrough, the drawdown of concentrated positions will be much faster than their accumulation. He’s not just betting on direction, but on timing. Once the timing is off, the consensus within the circle may反而 become a collective blind spot.
Return to the original question.
Everyone is analyzing his holdings, trying to replicate his trades. But the returns of a legendary investor come with structural conditions.
The paper is public, the position reports are public, and his investment logic is clearly explained in podcasts and interviews. But even if you fully understand every one of his judgments, you cannot replicate the position from which he made those judgments.
Positions can be traced back, returns are enviable, but the source of insight cannot be shared. This is likely the most expensive form of asymmetry in this era.