AGI Fund SALP Adds Billions in Put Options Amid AI Infrastructure Bets

But SALP still added to a batch of AI infrastructure, meaning he has given up the illusion of "universal AI appreciation."

Written by Jim, MSX MaiTong

During this U.S. 13F filing season, one of the most closely watched funds is not Bridgewater or Berkshire Hathaway, but a fund with a particularly distinctive name—Situational Awareness LP.

Its founder, Leopold Aschenbrenner, is not a traditional Wall Street veteran but a former member of OpenAI’s Superalignment team. In 2024, he published a lengthy article titled “Situational Awareness: The Decade Ahead,” making a highly radical core assertion: AGI may arrive much sooner than most people imagine, and the true scarcity in the future will not just be model capability itself, but computing power, electricity, data centers, chips, storage, and the national-level resource competition driven by the AI arms race.

Two years later, it turned out he was right.

Leopold internalized a set of judgments about the next decade of AGI and mapped this framework onto the capital markets. As a result, Situational Awareness, from its inception, has never been like a conventional technology fund, but rather a direct translation of the AGI roadmap into an investment map for AI infrastructure.

This is why every move he makes in the AI investment space draws significant market attention, and the latest disclosed 13F filing suggests that this most knowledgeable AI bull appears to be quietly building large positions in put options.

Public records show that Leopold founded an investment firm focused on AGI, backed by prominent Silicon Valley figures including Patrick Collison, John Collison, Nat Friedman, and Daniel Gross.

According to market reports, Situational Awareness achieved a net return of approximately +47% in the first half of 2025 after fees, significantly outperforming the S&P 500 and the technology hedge fund index during the same period. Its distinguishing feature is that it is not simply bullish on "technology stocks," but rather highly focused on AI infrastructure, betting on where AI capital expenditures will ultimately flow.

As mentioned at the beginning, his underlying logic is that if AGI truly accelerates, the companies most likely to be revalued first may not be application-layer firms, but rather those controlling computing power, electricity, data centers, storage, optical communications, semiconductor equipment, and energy resources. Therefore, his high returns do not stem from simply buying an index, but from gaining exposure to a group of highly elastic AI infrastructure stocks—such as Bloom Energy, Sandisk, Lumentum, CoreWeave, and Core Scientific.

Here, we need to first explain what a 13F is.

Form 13F is a quarterly filing submitted by U.S. institutional investment managers to the SEC, commonly used to observe changes in the quarterly holdings of large funds in U.S. stocks, ETFs, and related options. However, it is essentially only a snapshot at the end of a quarter, revealing only what was disclosed at a specific point in time; it cannot fully reconstruct a fund’s entire trading strategy, especially regarding options—Form 13F does not show strike prices, expiration dates, whether positions are hedged in conjunction with others, or allow direct derivation of the fund’s true net exposure.

This is also the part most commonly misunderstood when interpreting this document.

The Q1 13F filing deadline was March 31. Last10K shows the document was submitted in the evening Eastern Time on May 15, but the SEC accepted it on May 18—meaning it wasn’t simply “not filed,” but rather there was a delay between submission and when the market actually saw the disclosure, which is why there was widespread discussion on social media about “waiting for Leopold’s 13F.”

More importantly, this 13F filing does not fully align with market expectations. Many had assumed that Leopold would significantly increase positions in core AI assets such as NVIDIA, Broadcom, AMD, TSMC, and ASML. However, the reality is that SALP established a large number of PUT options positions covering a range of core AI and semiconductor assets, including the SMH Semiconductor ETF, NVIDIA, Oracle, Broadcom, AMD, Micron, TSMC, ASML, and Intel.

This has caused the market to reconsider a question: Why are those who most believe in the imminent arrival of AGI now buying insurance for AI leaders?

Reducing it to merely "bearish on AI" is too simplistic; what truly deserves analysis is the macroeconomic context in which this move was made, and what it reveals about the evolving structure of AI trading.

The most striking move revealed in this 13F is that SALP has established a large number of put options positions:

• The largest is the SMH Semiconductor ETF PUT, with a disclosed value of approximately $2.043 billion;
• Second is NVDA PUT, approximately $1.568 billion;
• Then there is the ORCL PUT, approximately $1.073 billion;
• AVGO PUT, approximately $1.006 billion;
• and AMD PUT, approximately $969 million;
• In addition, it established new positions in MU PUT, TSM PUT, ASML PUT, INTC PUT, among others;

At first glance, this may appear to be a bearish bet on AI leaders, but the issue is that puts do not necessarily indicate a one-sided short position—after all, the notional value of options reported in 13F filings is based on the underlying security’s size, not the actual premium cost incurred by the fund. More importantly, 13F filings do not reveal strike prices, expiration dates, whether positions are hedged alongside others, or the true net exposure within the portfolio.

Therefore, it is not accurate to say outright that Leopold is "completely bearish on NVIDIA and semiconductors"; a more reasonable interpretation is that he is buying insurance for his long-position AI infrastructure portfolio, since many of the assets originally held by SALP—such as Bloom Energy, CoreWeave, Core Scientific, IREN, Applied Digital, and SanDisk—are inherently high-beta, highly volatile, and interest-rate-sensitive companies. While their long-term fundamentals are tied to AI infrastructure, their short-term stock prices often depend heavily on risk appetite and valuation environments.

When markets begin to de-risk due to rising oil prices, recurring inflation, higher interest rates, or geopolitical conflicts, these high-elasticity assets are often sold first. This is also related to the macroeconomic backdrop at the end of March: on one hand, tensions in the Middle East and the risk of conflict between the U.S. and Iran have pushed up oil price expectations; on the other hand, rising oil prices exacerbate inflation persistence, weakening market confidence in rate cuts.

For high-valuation growth stocks, this amounts to a "double pressure": rising oil prices push inflation higher, which dampens rate cuts; with interest rates remaining elevated, the valuations of long-duration tech assets are compressed.

In this context, Leopold's large-scale creation of PUTs becomes easier to understand—he is not denying AI, but acknowledging that even the strongest long-term AI logic cannot completely ignore macroeconomic headwinds.

In particular, for a fund like SALP, many of the holdings are high-beta assets. Holding only offensive positions can lead to significant portfolio volatility during systemic market drawdowns. By purchasing PUTs on liquid, representative AI core assets such as SMH, NVDA, AVGO, AMD, and ORCL, it can hedge against systemic downside risk in AI trading using relatively standardized instruments.

The real meaning behind this is that Leopold did not switch from being an AI bull to an AI bear, but rather shifted from aggressively going long on AI alone to continuing to bet on AI infrastructure while beginning to manage path volatility.

This is a more sophisticated approach to portfolio management.

If the new PUT addresses the "defensive issue," then adding, reducing, or clearing positions truly reveal where Leopold's offensive strategy lies.

According to the disclosure, SALP has maintained and increased positions in several AI infrastructure-related assets, including a modest increase in Sandisk common stock, as well as additions to CoreWeave common stock, IREN, Applied Digital, Riot Platforms, CleanSpark, Bitfarms, and Bitdeer. Key retained long positions currently include Bloom Energy, Sandisk, CoreWeave, IREN, Core Scientific, and Applied Digital.

This indicates that it has not abandoned AI; on the contrary, it still bets on the same long-term thesis: AI capital expenditures will continue to filter down, and the real beneficiaries will be companies that control electricity, data centers, storage, computing power capacity, and infrastructure bottlenecks.

This is very close to the main thesis of MSX Q2. In our article "AI Infrastructure Rose Throughout Q1—Who Can Still Sustain 'High Valuations' in Q2?", we emphasized that the focus of AI trading has shifted from GPUs alone to networks, storage, and power. The market now cares more about where the ongoing capital expenditures of major companies will ultimately flow—in terms of orders, revenue, and profits. These segments—equipment, networks, storage, and power—are favored not because they’re more glamorous, but because they better align with the market’s current preference for tangible execution.

From this perspective, SALP’s long positions are representative: Bloom Energy corresponds to electricity and independent energy supply; CoreWeave, Applied Digital, Core Scientific, and IREN correspond to data centers, computing power hosting, and infrastructure support; positions related to Sandisk, Micron, and TSM correspond to storage, semiconductor manufacturing, and hardware supply.

In other words, Leopold isn't avoiding AI—he's more concerned about where the AI money ultimately goes and who can turn that spending into revenue on the financial statements.

Now let’s look at the reductions and full exits—equally informative. SALP fully exited positions in INTC CALL, Lumentum, and Cipher Mining, and reduced positions in CoreWeave CALL, Bloom Energy, Core Scientific, and others. Most notably, it’s not simply withdrawing from a particular sector, but rather trimming positions that have already risen significantly, exhibit high volatility, or carry stronger leverage characteristics.

For example, CoreWeave reduced its CALL positions while still holding common shares, indicating it is not fully abandoning CoreWeave but rather shifting from a more aggressive options strategy back to a more controlled common stock position. Similarly, for Bloom Energy and Core Scientific, reducing holdings does not mean the underlying thesis is invalid—it likely reflects risk management and profit-taking at the portfolio level.

Lumentum’s sell-off is even more intriguing. In the MSX Q1 review, AI hardware and optical communications were the two strongest themes, with AXTI, AAOI, LITE, and LWLG all achieving doubling-level gains. The strength in optical communications fundamentally stems from the surge in demand for optical interconnects, optical modules, and network links in AI data centers. However, the issue is that the stronger a theme performs in Q1, the more likely it is to face crowded trades and a declining risk-reward ratio entering Q2.

So Leopold reducing positions in LITE and trimming some high-elasticity AI infrastructure holdings doesn’t necessarily mean it’s bearish on the sector—it may simply be acknowledging more realistically that the most successful trade in Q1 isn’t necessarily the most cost-effective one in Q2.

This is the most important aspect of this rebalancing. It’s not a rejection of AI, but an active structural shift—from buying anything in the AI chain to retaining only assets that can better support long-term capital expenditures, possess more infrastructure-like characteristics, and better withstand macroeconomic volatility.

What he gave up was not AI, but the linear fantasy that "all AI will rise together."

This 13F is essentially a snapshot as of March 31 and does not indicate that Leopold still holds the exact same positions by May, but it remains highly informative for current market conditions.

First, the long-term theme of AI has not ended, but the trading structure has changed: in the future, it won’t be enough to simply buy any AI-related asset—it’s about who can deliver value, who commands a premium, who is overcrowded, and who needs hedging.

Second, in an environment of high oil prices, high interest rates, and high volatility, the most effective strategy is neither purely aggressive nor fully defensive, but rather offensive with defense—allocating core positions to certainty and marginal positions to flexibility, while using hedging tools to manage portfolio drawdowns. Leopold’s actions this time essentially demonstrated this logic through actual position allocations.

Third, this also confirms a major shift in the 2026 U.S. stock market: index Beta is weakening, while structural Alpha is growing stronger. In the past, simply buying the Magnificent Seven or NVIDIA might have been enough to win; but now the market is more discerning, asking every company: Can your AI story ultimately turn into orders? Can it turn into revenue? Can it turn into profit? If not, no matter how high the valuation, it will be compressed.

This is why AI infrastructure 2.0 will become important. In the future, capital will not just focus on GPUs, but will follow the chain from compute → interconnect → storage → power → data center infrastructure to identify the true value-creating segments.

On the surface, the most eye-catching aspect of this 13F is the series of massive PUTs.

But if you look at the entire position, you’ll see that Leopold didn’t simply switch from being bullish on AI to bearish—he made a more mature upgrade: maintaining long-term bets on AI infrastructure while briefly acknowledging the volatility risks of overvalued, highly elastic assets.

This is the most important aspect of this 13F—it tells us the direction of AI, which may still be correct, but the path to that direction will certainly not be a straight line.

For true fund managers, what matters is not just correctly predicting the endpoint, but surviving the volatility along the way.

For ordinary investors, the biggest takeaway from this 13F is clear: AI trading in 2026 has moved from "buying the story" to "buying the execution"; from "buying the leaders" to "identifying the bottlenecks"; and from "one-sided offense" to "offense with defense."

This is the most interesting signal, and one that must not be ignored.