💢💢💢

🔥🎯Fired by $OpenAI at 22, managing a $5.5 billion fund 3 years later: Leopold Aschenbrenner's AI portfolio actually reveals a bigger trend.

Many people's first reaction to this story is "genius comeback."

But if you look closely at his investment portfolio, you'll find that what's really important isn't the personal story, but the direction he's betting on.

Because this portfolio almost tells the same logic:

The bottleneck for AI isn't the model, it's the infrastructure.

In other words:

The next stage of AI competition is power + data centers + chips + network.

If we break it down, his investments are actually very clear.

First layer: Energy

$Bloom Energy(BE.US)
$EQT(EQT.US)(EQT.US)
$Solaris Energy Infrastructure(SEI.US)
$Coreweave(CRWV.US)
These companies collectively solve one problem:
AI data centers require massive amounts of power.
For example:

$Bloom Energy(BE.US) Bloom Energy provides fuel cells to power data centers.
$EQT(EQT.US)(EQT.US) is one of the largest natural gas producers in the US.
$Solaris Energy Infrastructure(SEI.US) Solaris Energy works on energy infrastructure.
$Coreweave(CRWV.US) CoreWeave provides GPU cloud computing power.

AI's computing power demand is growing exponentially.

And the essence of computing power is actually:

Power.
Many AI data centers are now even starting to directly tie up with power plants.
Second layer: Bitcoin mining companies transitioning to AI
This is the most interesting part of his portfolio.

$Core Scientific, Inc.(CORZ.US)
$IREN(IREN.US)(IREN.US)(IREN.US)
$Applied Digital(APLD.US)
$Cipher Digital(CIFR.US)
$Riot Platforms(RIOT.US)

These companies were originally:

Bitcoin miners.

But why would an AI investor buy mining companies?

Because mining companies already possess three key assets:

Land
Power contracts
Data center infrastructure

When AI computing demand explodes, these mining sites can quickly transform into:
GPU data centers.
For example:
$Core Scientific, Inc.(CORZ.US) Core Scientific
has already started providing data center hosting for AI workloads.
$IREN(IREN.US)(IREN.US)(IREN.US) Iris Energy
does both Bitcoin mining and AI cloud computing.

This is a very typical infrastructure arbitrage.

Third layer: AI hardware supply chain

$Sandisk(SNDK.US)
$Coherent Corp.(COHR.US)
$Tower Semicon(TSEM.US)
$Intel(INTC.US)
$Lumentum(LITE.US)

These companies solve another problem:

Hardware inside the data center.
For example:
$Sandisk(SNDK.US) Sandisk → AI storage
$Coherent Corp.(COHR.US) Coherent → lasers/photonics
$Lumentum(LITE.US) Lumentum → optical networking
$Tower Semicon(TSEM.US) Tower Semiconductor → analog chips
$Intel(INTC.US) Intel → US chip manufacturing

If AI data centers continue to expand, demand for these components will increase.

So if we summarize the entire investment portfolio, it's really just one sentence:

AI = Power + Data Centers + Chips.

Not:
AI = ChatGPT.

This is also why many AI investors are increasingly focusing on:
Power companies
Mining companies
Optical modules
Storage

Because the real cost of AI isn't the model, it's:

Computing infrastructure.

Leopold Aschenbrenner's portfolio is actually very concentrated.
The top three holdings are:
$Bloom Energy(BE.US) — 29%
$Core Scientific, Inc.(CORZ.US) — 14%
$IREN(IREN.US)(IREN.US)(IREN.US) — 11%
In other words:
His biggest bet is energy + AI data centers.

This is actually in sync with the actions of many tech giants today.

For example:

$Meta Platforms(META.US) is frantically building AI data centers
$Microsoft(MSFT.US) and $Alphabet(GOOGL.US) are also increasing computing capital expenditure

Some analysts have even started saying something interesting:
The biggest future bottleneck for AI isn't GPUs, it's electricity.
If AI keeps expanding, the scarcest resource might not be chips, but:
Grid capacity.
So this portfolio actually reveals a judgment:

The real winners of AI might not be model companies.

But could be:

Companies that provide energy and computing infrastructure for AI.

The question is—

If AI really becomes a "power industry," who do you think will be the biggest beneficiaries of AI in the future?

GPU companies, tech giants, or energy companies?