BeanWealth Research

Research on Stocks & Investing 📈 Investing Newsletter

[BeanWealth Research @bean_wealth]

7 charts that will teach you more about investing than school did:

1. Bill Ackman sold the bottom

[BeanWealth Research @bean_wealth]

$SMX Security Matters is focused on a simple idea:

As recycled materials become more valuable, proving what’s actually in them becomes more important.

That’s the verification layer SMX is building.

Disclaimer: https://t.co/9quOq6p6yD

Sector Peers: $IBM $PLTR $LYB $DOW

[Matt Allen @investmattallen]

$ONCY Oncolytics Biotech is focused on helping the immune system recognize and attack cancer more effectively.

Its lead therapy, pelareorep, is being studied across multiple cancers including colorectal, pancreatic, and anal cancer, with a strong focus on combining it with existing immunotherapy treatments.

The company recently brought in a new CEO who previously led Ambrx through its ~$2B acquisition by Johnson & Johnson, while upcoming clinical updates and regulatory discussions remain key catalysts.

Disclaimer: https://t.co/NMXcSQ6fJw

[BeanWealth Research @bean_wealth]

$ONCY Oncolytics Biotech continues to build around pelareorep, its lead cancer therapy candidate.

For investors, the story is increasingly centered around clinical progress and moving toward later-stage development.

Disclaimer: https://t.co/bNo1Mestmj

[Matt Allen @investmattallen]

The AI neocloud category is one of the most misunderstood sectors of the AI infrastructure trade:

1. $MARA Mara (AI Sovereignty)

American hyperscalers have a legal problem serving foreign governments. The CLOUD Act means data stored on AWS or Azure is potentially accessible to the U.S. government and foreign nations are not okay with that. MARA acquired a 64% stake in Exaion, a subsidiary of French state-owned energy giant EDF, specifically to serve governments and enterprises that need AI infrastructure their own governments actually control. That is a customer base $AMZN AWS cannot legally compete for. And once its Long Ridge acquisition closes later this year, MARA’s operational and development footprint reaches roughly 2.2 GW. This is no longer a Bitcoin miner but an AI play hiding in plain sight.

2. $NVT nVent (The Vera Rubin Buildout)

nVent is not a neocloud but you cannot build one without it. The entire Vera Rubin generation is defined by one design choice: $NVDA NVIDIA made the NVL72 racks 100% liquid cooled. That single decision turns nVent into a direct beneficiary of every Rubin rack deployed, because each one needs exactly what nVent makes: liquid cooling hardware and rack level power protection. As Rubin ramps through the second half of 2026, that demand flows straight into nVent’s order book, which hit a record $2.6 billion last quarter. It compounds quietly at 20% operating margins while the rest of the market chases the headline names.

3. $CRWV CoreWeave (The Essential Cloud)

When companies build and deploy AI models they need somewhere to run them. CoreWeave is becoming the place the best ones go. Vera Rubin is NVIDIA’s next generation chip delivering roughly five times the inference performance of what exists today and CoreWeave is among the very first to deploy it at scale. That early access is how you end up with nearly $100 billion in contracted backlog from $META Meta, Anthropic and Mistral. They are not signing decade long commitments to a GPU rental company. They are betting on the infrastructure layer the AI economy runs on.

4. $IREN IREN (Energy-to-Compute)

The thing most people miss about AI infrastructure is that GPUs are not the bottleneck. Power is. You can order chips today and wait years for the electricity to run them. IREN has secured 4.5 GW of power globally and needs only a fraction to hit near term targets. Microsoft is already a customer. NVIDIA signed a landmark partnership. And the Mirantis acquisition brought in hundreds of engineers with a decade of enterprise cloud experience across over a thousand customers. The runway here is genuinely difficult to overstate.

[Matt Allen @investmattallen]

$CRWV Coreweave just became the first AI cloud provider to fully deploy and validate $NVDA Vera Rubin

The fastest path to the newest NVIDIA infrastructure is increasingly running through CoreWeave

This is why the company continues to separate itself from the rest of the market

[BeanWealth Research @bean_wealth]

$ONCY Oncolytics is focused on advancing pelareorep across the toughest cancer markets, including pancreatic, colorectal, and anal cancer.

They rbrought in CEO Jared Kelly, who previously led Ambrx through its roughly $2B acquisition by $JNJ

Disclaimer: https://t.co/XFXT3DEOZ0

[Matt Allen @investmattallen]

Dylan Patel: “There are tons of reasons you need CPUs.”

Most investors are focused on GPUs, but CPUs are becoming a critical piece of AI infrastructure

As AI applications and inference workloads grow, demand is accelerating across the entire compute stack

[Matt Allen @investmattallen]

$AMZN AWS CEO: “Compute demand is so excessive that we have never retired old A100s.”

The AI infrastructure buildout is happening so fast that even outdated GPUs remain valuable

Most investors still underestimate how much compute capacity the AI economy will ultimately require

[Matt Allen @investmattallen]

Goldman Sachs projects token usage from AI agents will increase 24x by 2030

This is massively bullish for neoclouds like $CRWV, $NBIS, $MARA, and $IREN as persistent inference demand drives enormous compute needs

The AI infrastructure boom is still very early

[Matt Allen @investmattallen]

$GRML Greenland Mines is tied directly to one of the biggest geopolitical investment themes developing right now: critical mineral independence.

The company’s Skaergaard project in Greenland has exposure to gold, palladium, rare earths, titanium, and other strategic minerals tied to defense systems, AI infrastructure, energy, and advanced manufacturing.

As the U.S. and Europe push to reduce dependence on China for critical materials, Greenland keeps becoming a more important part of the conversation.

Disclaimer: https://t.co/QvrgWCtwNt

[Matt Allen @investmattallen]

Gavin Baker says the neocloud trade is far more durable than most investors realize

“Not all GPU hours are the same.”

He argued companies like $CRWV CoreWeave are operating at a completely different level than low-end providers

Investors need to watch this:

[Matt Allen @investmattallen]

Added 2 new positions in BeanFlagship, our AI infrastructure Autopilot fund:

CoreWeave $CRWV: Every major company in the world is racing to build AI and they all need one thing: computing power. CoreWeave owns tens of thousands of GPUs, the specialized chips that power every AI model being built today, and rents that computing power to the companies that need it. Think of it like owning the power grid during an industrial revolution. Everyone needs what you have and nobody can build fast enough to compete with you.

Demand is so far ahead of supply that prices keep rising and customers are paying upfront just to lock in capacity. That is not a struggling company. That is a company with all the leverage. They are guiding for $25 billion in revenue by 2027 and I believe they get there. This is one of the most important infrastructure buildouts in a generation and CoreWeave is sitting right in the middle of it.

nVent Electric $NVT: Everyone is focused on the GPUs. Nobody is asking how you keep them from melting. nVent makes the liquid cooling systems that sit inside every major AI data center and without them, a $10 billion investment in chips dies from overheating. They grew this business 76% organically in Q1 2026, have a $2.6 billion backlog that has tripled in a year, and fewer than 10% of data centers use liquid cooling today. The runway here is a decade long. This is infrastructure nobody talks about but everyone needs.

[Matt Allen @investmattallen]

Chamath Palihapitiya says AI companies now need three things: “land, power, shell.”

The AI race is a race for power as electricity demand explodes across data centers

The companies securing energy capacity today are positioning for the next decade of AI infrastructure demand

[BeanWealth Research @bean_wealth]

$VWAV VisionWave is building around AI-driven defense, RF sensing, and counter-drone systems.

-Closed $20M financing
-SaverOne partnership underway
-RF defense solution targeted for 2026

Pulled back from ~$6 to ~$5.25 and staying on watch.

Disclaimer: https://t.co/5ZSaAruOQd

[Matt Allen @investmattallen]

$MS Morgan Stanley wrote a sell-side research report on $MARA MARA that I think completely underestimates what the company is becoming:

-MARA is no longer just a Bitcoin miner. The company is building an AI infrastructure platform centered around owning power, AI data centers, sovereign compute, and high-performance compute infrastructure

-MARA now has a 2.2 GW pipeline across the U.S. and international markets with Long Ridge alone giving the company one of the most important assets in AI: connected power for large-scale AI data centers

-The company believes power, not chips, is becoming the biggest bottleneck in AI infrastructure and assets like Long Ridge could take up to 10 years to replicate due to permitting and interconnection constraints

STARWOOD PARTNERSHIP:

-The Starwood partnership gives MARA access to one of the most experienced hyperscaler data center development platforms in the world while allowing MARA to pursue AI infrastructure in a far more capital-efficient structure

-MARA even outlined a hypothetical 200 MW example where a powered site contribution could potentially translate into ~50% ownership and up to $50M-$100M in annualized stabilized cash flow with limited additional equity exposure beyond the land contribution itself

-The Morgan Stanley analyst acknowledged the Starwood JV could become a major growth driver, but is only assigning a 10% probability until leases are signed, which is exactly why Wall Street is often late to major value shifts like this

EXAION / SOVEREIGN AI:

-Exaion gives MARA exposure to sovereign AI, enterprise AI, and private cloud infrastructure as governments and enterprises increasingly want local control over compute, security, and data residency instead of relying entirely on traditional U.S. hyperscalers and the U.S. Cloud Act

-The company specifically highlighted expansion discussions in France, Brazil, Saudi Arabia, and other energy-rich regions as part of its international AI infrastructure strategy

I think $MARA MARA is building one of the most misunderstood AI infrastructure ecosystems in the market right now

[BeanWealth Research @bean_wealth]

$VWAV VisionWave is building around AI-driven defense, RF sensing, and counter-drone systems.

-Closed $20M financing
-SaverOne partnership underway
-RF defense solution targeted for 2026

Pulled back from ~$6 to ~$5.25 and staying on watch.

Disclaimer: https://t.co/5ZSaAruOQd

[Matt Allen @investmattallen]

Leopold Aschenbrenner: “By 2030, a single AI training cluster could consume over 20% of all electricity generated in US”

That’s not including inference demand on top of it

The opportunity in power and AI infrastructure is far bigger than most realize

Investors must watch:

[Matt Allen @investmattallen]

Most investors have no idea what $VIVO is building right now

After going deep into the company, I think the market is underestimating the long-term opportunity ahead

One of the more overlooked asymmetric stocks I’ve researched lately

Full deep dive:

https://t.co/nPMFStM2VJ

[BeanWealth Research @bean_wealth]

$MARA $VIVO

[Matt Allen @investmattallen]

The AI Investing Cold War

There is a second AI buildout happening right now

The opportunity here is massive:

https://t.co/ItrLDHrSqV

[Matt Allen @investmattallen]

The AI Investing Cold War

There is a second AI buildout happening right now

The opportunity here is massive:

https://t.co/ItrLDHrSqV