Leopold Aschenbrenner’s Top 10 Long Positions: The AI Infrastructure Barbell

A portfolio list circulating today claims to show Leopold Aschenbrenner’s top 10 long positions with the following weights:

Before treating this as a trade signal, remember the most important line: this is a snapshot, not a complete risk report. It does not show cost basis, hedges, short positions, put exposure, leverage, tax constraints, or whether the position weights come from equity value, options notional, or a model portfolio. For an individual investor, the useful exercise is not copying the list. The useful exercise is understanding the portfolio structure.

The Big Idea: AI’s Bottleneck May Be Physical, Not Just Digital

The portfolio is not simply a “buy every AI stock” basket. It is more specific than that.

The list is concentrated in the parts of the AI buildout that could become scarce:

• Compute access: NBIS and CRWV.
• Power availability: Bloom Energy.
• Data-center shells and hosting capacity: IREN, CORZ, and APLD.
• Memory and storage: SNDK and MU.
• Chip supply chain exposure: TSM and INTC.

That is a very different thesis from only owning mega-cap software or the most obvious GPU winners. The bet appears to be that the next constraint in AI is not only model quality. It is the ability to secure enough electricity, buildings, memory, networking, cooling, and compute contracts.

Concentration Risk: NBIS at 35% Changes Everything

The first thing that jumps out is NBIS at 35%. A 35% position is not a normal diversified holding. It is a statement of conviction.

That kind of concentration can create huge upside if the thesis is right. It can also dominate the entire portfolio if the market turns against the name.

For a retail investor, a useful rule is:

When one stock can decide your year, you are no longer just stock picking. You are managing portfolio survival.

A 35% position needs a risk plan before volatility arrives. That could mean:

• Setting a maximum position size.
• Trimming after sharp rallies.
• Using protective puts after a major run-up.
• Selling covered calls only at prices where you would truly be happy reducing exposure.
• Pairing the position with cash or lower-beta holdings instead of adding more correlated AI names.

Read: Protective Put Guide

The Hidden Correlation: These Are Different Tickers, But One Macro Trade

At first glance, the portfolio has cloud providers, memory companies, power companies, miners, and chip manufacturers. That looks diversified by industry label.

But under the hood, many of these names can move together because they depend on the same AI-capex cycle.

If hyperscalers and enterprises keep increasing AI infrastructure spending, the whole basket can benefit. If AI capital spending slows, financing gets tighter, power projects are delayed, or data-center economics disappoint, several holdings could fall at the same time.

The shared drivers include:

1. AI capex growth — the market must believe the spending wave continues.
2. Power demand — data centers need reliable electricity at scale.
3. Financing conditions — infrastructure-heavy companies often need capital.
4. GPU and server economics — hosting margins depend on hardware availability and pricing.
5. Memory pricing — SNDK and MU benefit when memory cycles tighten, but cycles can reverse.
6. Regulatory and grid constraints — power interconnection delays can slow buildouts.

So even though the list has 10 names, it may behave like one large AI infrastructure factor bet.

Why Bloom Energy Matters in This Portfolio

The reported third-largest position is BLOOM / BE, which most investors would associate with Bloom Energy. This is one of the most interesting positions because it connects directly to the power bottleneck.

AI data centers do not only need chips. They need electricity, uptime, backup power, and often faster deployment than the traditional grid can provide. If the market believes data centers will increasingly need on-site or distributed power solutions, power infrastructure companies can become a major part of the AI trade.

The risk is that power stories can be capital intensive. Investors should watch:

• Gross margins and backlog quality.
• Customer concentration.
• Project financing terms.
• Whether revenue growth turns into free cash flow.
• Whether expectations become too aggressive after a big rally.

Memory Exposure: SNDK and MU Are Not Boring Here

SNDK at 15% and MU at 6% give the portfolio a large memory-and-storage angle. That matters because AI workloads are not only GPU hungry. They are also memory hungry.

The bullish case is straightforward:

• AI servers need high-performance memory.
• Data growth increases demand for storage.
• A tighter memory cycle can expand margins quickly.
• Operating leverage can make earnings recover faster than revenue.

But memory is cyclical. When supply catches up or demand disappoints, pricing can weaken. That is why investors should avoid assuming that a strong memory cycle automatically lasts forever.

A disciplined investor can use a simple checklist:

1. Are average selling prices still rising?
2. Are inventories healthy or building?
3. Are customers double-ordering?
4. Is capex discipline holding across the industry?
5. Are earnings estimates moving up or down?

The Miner-to-Data-Center Trade: IREN, CORZ, and APLD

The combined reported weight in IREN, CORZ, and APLD is 14%. This is a big allocation to companies tied to the idea that power-rich bitcoin mining infrastructure can be repurposed or upgraded for AI and high-performance computing.

The appeal is obvious: these businesses may already have access to land, power, and operating experience with energy-intensive compute. If they can sign attractive AI hosting deals, the market may reward them with data-center-style multiples instead of bitcoin-miner-style multiples.

But this part of the portfolio is also high risk. Investors should watch for:

• Contract quality and counterparty risk.
• Capital expenditure needs.
• Dilution from equity raises.
• Debt maturity schedules.
• Execution risk when moving from mining to AI hosting.
• Bitcoin price sensitivity that may still remain in the business model.

This is not a sleepy infrastructure bucket. It can be extremely volatile.

TSM and INTC: Supply Chain Plus Optionality

The smaller positions in TSM and INTC add semiconductor supply-chain exposure. TSM is the global foundry leader, while Intel is more of a turnaround and foundry-optionality story.

The interesting part is the size difference. TSM at 5% is meaningful but not dominant. INTC at 2% looks more like optionality than a core position.

That sizing sends a useful lesson: speculative turnaround ideas can belong in a portfolio, but they do not need to be oversized. If the thesis works, the upside can matter. If it fails, the portfolio should survive.

Options Playbook for This Kind of Portfolio

A portfolio like this does not need complicated options just for the sake of being sophisticated. The goal is simple: define risk around concentrated, volatile AI infrastructure exposure.

1. Protective puts on the biggest winners

If NBIS or another high-conviction holding becomes too large, protective puts can cap downside while keeping upside open. The trade-off is cost. In high-volatility names, put premiums can be expensive, so timing and strike selection matter.

Read: Protective Put Guide

2. Covered calls after vertical rallies

Covered calls can make sense when a position has moved fast and you would be willing to sell at a higher price. They are not free money. They trade away some upside in exchange for premium.

Read: Covered Calls Guide

3. Collars for oversized positions

A collar can be useful when a stock is too large to ignore but too important to sell immediately. The investor buys a put and sells a call, creating a defined range of outcomes.

Read: Collar Strategy Guide

4. Cash-secured puts for names you want at lower prices

For volatile infrastructure names, chasing breakouts can be dangerous. A cash-secured put can define the price where you are willing to buy and pay you premium while you wait.

Read: Cash-Secured Put Guide

5. Index or sector hedges when everything is correlated

If the real risk is that the entire AI trade sells off together, hedging one stock may not be enough. In that case, investors can consider broader hedges through QQQ, SMH, or another relevant ETF. The hedge should match the risk. Do not hedge a single-stock problem with an index if the position risk is company-specific.

What I Would Watch Next

For a portfolio built around AI infrastructure, the most important signals are not only quarterly EPS beats. I would watch:

• Hyperscaler AI capex guidance.
• Data-center power availability and interconnection timelines.
• Memory pricing trends.
• GPU supply and rental economics.
• Credit markets and financing costs.
• New AI hosting contracts from miner-to-HPC companies.
• Insider selling, secondary offerings, and dilution.
• Whether the market starts rewarding cash flow instead of just capacity announcements.

My Takeaway

This reported top-10 list is one of the clearest examples of the AI infrastructure barbell: own the bottlenecks around compute, power, data centers, and memory, while accepting that the trade may be volatile and highly correlated.

The biggest lesson is not “buy these 10 tickers.” The biggest lesson is that AI investing is moving beyond obvious software and chip headlines. The market is trying to price the entire physical stack required to run AI at massive scale.

For retail investors, the right question is not whether Leopold Aschenbrenner is right or wrong. The right question is:

• Can I survive the drawdown if the AI infrastructure trade corrects?
• Do I understand what drives each holding?
• Am I confusing ticker diversification with true risk diversification?
• Do I have a plan for trimming, hedging, or adding before volatility arrives?

A concentrated portfolio can create extraordinary returns. It can also expose every weak spot in an investor’s process. If you copy the conviction without copying the risk management, you are only copying half the trade.

Not investment advice. The position weights above are treated as an unverified public snapshot as of May 31, 2026. Always verify holdings, prices, option exposure, and risk before making investment decisions.