🪙 Tiger Coins | Leopold’s Final Portfolio Before Citadel: SNDK and MU Were His Biggest Bets
Leopold Aschenbrenner’s final disclosed portfolio before Citadel took over most of his public-equity positions is out. The biggest surprise is obvious: His top two holdings were not $NVIDIA(NVDA)$, but $SanDisk Corp.(SNDK)$ and $Micron Technology(MU)$.
And when you look at the rest of the portfolio, the strategy becomes pretty clear. Leopold was not simply betting on AI chips.
He was betting on the entire AI infrastructure buildout.
💾 Storage Was the Biggest Bet
His top holdings included:
Followed by $Taiwan Semiconductor Manufacturing(TSM)$, $NEBIUS(NBIS)$, $CoreWeave, Inc.(CRWV)$, $Core Scientific, Inc.(CORZ)$, $STMicroelectronics NV(STM)$, $APPLIED DIGITAL CORP(APLD)$ and $Riot Platforms(RIOT)$.
The most important part for me is still $SanDisk Corp.(SNDK)$ and $Micron Technology(MU)$ sitting at No.1 and No.2. That tells you how aggressively Leopold was treating memory and storage as part of the AI trade.
The logic is simple:
More AI servers → more compute → more data → more demand for DRAM, NAND and enterprise SSDs.
And the industry data supports that view. Memory pricing has continued to strengthen, with both DRAM and NAND benefiting from tighter supply and strong data-center demand. So Leopold was effectively betting that AI Capex would eventually spread beyond GPUs and into storage.
Memory prices are still rising in 3Q26, but the pace has slowed sharply after the surge in the first half of the year.
☁️ The New Positions Point to Neocloud
The new positions are also interesting:
🟢 $NEBIUS(NBIS)$ — Nebius, around $1.23B
🟢 $STMicroelectronics NV(STM)$ — around $584M
🟢 $Keel Infrastructure Corp(KEEL)$ — around $152M
🟢 $Vishay Intertechnology(VSH)$ — around $20M
🟢 $Cerebras Systems(CBRS)$ — around $2.2M
The one I care about most is $NEBIUS(NBIS)$.
Leopold already had exposure to $CoreWeave, Inc.(CRWV)$, $APPLIED DIGITAL CORP(APLD)$ and $Core Scientific, Inc.(CORZ)$. Adding a major $NEBIUS(NBIS)$ position shows that his AI trade was moving deeper into: Neocloud + data-center infrastructure.
$NVIDIA(NVDA)$ can sell the GPUs. But someone still has to build the data centers, provide the power and turn those GPUs into usable compute. That is where names like $NEBIUS(NBIS)$ and $CoreWeave, Inc.(CRWV)$ come in.
⚡ Why Was Bloom Energy No.3?
$Bloom Energy Corp(BE)$ being the third-largest holding might be the most underrated part of the portfolio. The AI bottleneck is no longer only: “Can we get enough GPUs?”
It is increasingly becoming: “Can we get enough power?”
AI-optimized servers are becoming the fastest-growing source of data-center power demand. That also helps explain why Leopold’s portfolio extends beyond chips and storage into power and infrastructure.
$Taiwan Semiconductor Manufacturing(TSM)$, $STMicroelectronics NV(STM)$ → chips
$SanDisk Corp.(SNDK)$, $Micron Technology(MU)$ → storage
$NEBIUS(NBIS)$, $CoreWeave, Inc.(CRWV)$ → compute
$Bloom Energy Corp(BE)$, $APPLIED DIGITAL CORP(APLD)$ → power and infrastructure
Together, these positions cover most of the AI data-center stack.
📈 A Huge Run Before the July Selloff
Situational Awareness had delivered extraordinary returns before the July rout.
By the end of June 2026, the fund was up 439% net for the year, with returns since inception reaching roughly 1,551%.
Then the AI selloff hit. Its portfolio value fell about 67% in July, although the fund still ended the month roughly 80% up for 2026.
The same concentrated and leveraged AI strategy amplified both the gains and the drawdown
📉 The Biggest Risk Was Also Obvious
The problem with this portfolio was concentration.
Most of these stocks belong to the same high-beta AI trade. When storage, Neocloud and power stocks rally together, the returns can be huge. But when investors suddenly de-risk AI exposure, they can all fall at the same time.So I would not look at this filing and blindly copy the portfolio.
The more useful signal is where one of the most aggressive AI investors believed the next money would flow.
🔍 This Portfolio Is Basically One AI Infrastructure Trade
At first glance, these holdings look very different.
But $SanDisk Corp.(SNDK)$ and $Micron Technology(MU)$ are storage, $NEBIUS(NBIS)$ and $CoreWeave, Inc.(CRWV)$ are compute, while $Bloom Energy Corp(BE)$ and $APPLIED DIGITAL CORP(APLD)$ are tied to power and data-center infrastructure.They all depend on the same thing:
continued AI data-center spending.
If that spending remains strong, more capital could continue moving beyond GPUs into storage, compute and power. If it slows, the concentration across the same AI theme becomes the main risk.
🐯 Tiger Coin Giveaway!
That's the full breakdown — one of AI's most-watched investors, betting across storage, compute, and power before it all got handed off to Citadel.
Now it's your turn. If you were building the "AI infrastructure stack" trade today — storage, compute, power, or chips — which piece would you bet on for the next 6 months, and why?
Drop your take in the comments! Best analysis gets Tiger Coins! 🪙
Good allocation isn’t just about investing — it’s also about keeping everyday life organised.
Just like a well-balanced portfolio, the right organisation can make business trips, travel and workouts a little easier. The new Tiger Toiletry Bag features a dual-layer dry & wet separation design with plenty of space to keep your essentials neatly organised, wherever you go.
Redeem the new Tiger Toiletry Bag now in Tiger Coin Mall. https://laohu8.com/J/redeemGift?goodID=100538&type=delivery
Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.
AI is increasingly becoming a data-movement problem, not just a compute problem. HBM demand remains strong, while AI servers are also driving significant demand for high-performance SSDs and NAND. Tight supply and improving pricing could provide additional operating leverage.
Micron is particularly interesting because it has exposure across HBM4, conventional server DRAM and enterprise SSDs, giving it multiple ways to benefit as AI infrastructure scales.
Power could ultimately become the biggest bottleneck, but power-generation and grid projects generally have longer lead times. Chips remain attractive, but valuations and expectations are already high across several AI leaders.
For the next six months, my ranking would be:
Memory/Storage > Chips > Power > Compute.
That’s why $SNDK and $MU stand out to me. Their upside is not simply tied to AI enthusiasm, but to a real hardware bottleneck: memory capacity and pricing. If hyperscalers continue spending aggressively on AI infrastructure, memory could remain one of the biggest beneficiaries.
The key risk is obvious: if AI CapEx slows or new supply arrives too quickly, pricing and margins could reverse. But for the next six months, I’d rather own the “picks and shovels” behind AI than chase the most crowded GPU trade.
@Capital_Insights [正经]