The Unknown AI Compounder
An underlooked AI company. The Friday Filter #84
Good morning investor 👋,
Welcome or welcome back to The Friday Filter—the quality investor’s market recap.
Brookfield is one of the largest positions in my portfolio. It’s a resilient long-term holding, and my only AI play that isn’t primarily tech-based. Today’s post isn’t a long read, but I wanted to share a brief overview of my Brookfield AI thesis, especially in light of the many billion-dollar AI projects it has announced in recent weeks. Truth is: Brookfield is an AI company, and one of the most underlooked1 out there.
Let’s get into it. (7 min read)
Today at a glance:
The Filter;
Brookfield; the unknown AI company
📄 In other news
📚 Resource of the week
Brookfield; the unknown AI company
The Filter
Last week, I wrote the following on Blossom Social. It was a summary on my Brookfield thesis surrounding the growth of AI and how it even indirectly supports my other holdings like Amazon and Google. It was brief, but the value is there, so I wanted to share it this week with some added thoughts:
Brookfield is possibly one of the most important players in the AI ecosystem, which a lot don’t particularly understand.
When we think of AI, we think of Google or Amazon or Microsoft, and not Brookfield. This is flawed, so let me break this down.
Simplified, AI growth has three attributes. Key attributes I should add.
Those are:
Software
Data
Chips
Storage
Energy
In other words:
Models and data to train the models
A place to host the model
A place to store the data to train the model
A chip to power the processing of the data
Energy to power it all
Google, OpenAI, Oracle, Meta, Amazon, Microsoft control the cloud and LLM side (aside from Oracle that only has cloud). These companies have the data and the storage via cloud computing (some even have the chips, but that’s overwhelmingly controlled by Nvidia as of today).
What these companies don’t have is unlimited access to data centres (the storage) and energy to power said data centres to match the demand of these newfound products.
Most data centres by big tech are leased contracts with third-party providers that own the buildings themselves. When you hear “high capex investments into cloud,” it’s not them buying out data centres, it’s them getting more leases, and more importantly, acquiring more Nvidia GPUs. (Both expenses of which affect the cash flow, hence why FCF isn’t best to use to value here.)
Demand for LLMs and AI integration with them into businesses, as well as for cloud computing, is sky high. Supply constraints, even at such large-scale companies like big tech, are hurting. They need more data centres. They need more energy. This isn’t an opinion; Google CEO and Amazon’s CEO said last earnings and in an interview how “we are facing supply constraints.”
Right now, 90% of data centre leases are booked in top markets, and clean energy access to power them is just as constrained. There’s around ±$130 billion worth of undeployed capex (total supply gaps) as I write this (most from Microsoft and Amazon).
Take a wild guess at what company is one of the largest infrastructure and renewable energy companies in the world that could offset these supply constraints.
A company that has hundreds of billions of dollars of capital to deploy and invest.
A company that partners with other third parties to acquire and develop large scale physical assets that it can then lease out for passive cash flow (distributions).
A company that could deliver the energy and data centres that big tech needs.
Ding ding ding. Say hello to Brookfield.
It depends on what Brookfield you own. Maybe you only own the renewables business (BEP) or the asset management (BAM) business or infrastructure (BIP), but this post I’m referring to BN (the mothership; exposure to all Brookfield businesses).
(I talked about why I own this instead of the other ones like Brookfield Asset Management in my Brookfield stock analysis on my newsletter (“Managing Resilience.”)
Overall, since 2022 (the launch of ChatGPT), Brookfield has invested around $72 billion globally into data centres and data centre related energy projects. That includes €20 billion into France (15 for data centres, 5 for energy and other), €10 billion into Sweden (data centre), $10 billion in partnership with Microsoft for energy, and tens of billions more.
Brookfield has also been acquiring data centres from other companies alongside investing in building the projects. As of now (not sure how accurate my total is here), Brookfield has invested and pledged $38 billion into data centre projects ($31 billion into Europe, ~$7 billion in North America and Australia), and about $35 billion into energy projects worldwide.
Big tech controls the distribution and growth of scaled AI. AI is much more storage intensive and energy consuming (over 10x even 20x more than the traditional internet). AI is the future, and Brookfield is a crucial part in this future.
Brookfield at the moment has about $100–$110 billion in deployable capital ready to invest. One of the benefits of the business is capital integration. So across the BN ecosystem—whether it be with asset management or insurance, renewables or infrastructure—it doesn’t matter what business of Brookfield it is, Brookfield can use the income it earns from those businesses alongside its deployable capital to grow its projects and total AUM. In the context of AI, this matters, since with current supply constraints, the ability to cross-invest is an amazing advantage.
There are only a handful of companies in the world with (1) leading AI infrastructure and renewable energy, (2) hundreds of billions of dollars in deployable capital and passive income from its operational businesses to fund new AI-aiding projects, and (3) an ability to integrate all of the above.
For example, Brookfield can use its lease distributions from its data centres (infrastructure business) to fund new energy projects to power said data centres… and vice versa. This applies to every business of Brookfield. Capital earned from insurance, fees earned on the asset management business; it’s all cross-invested. This is one of the great benefits of the BN business. And this, among the reasons above, is what creates the foundation for an inevitable win for an investment into Brookfield regarding AI. Ergo, Brookfield = winning AI stock.
For more on Brookfield, its operational structure, management quality, valuation, and everything you would ever need to know to form a foundational thesis on the company, you can read my full Brookfield analysis here:
a. 📄 In other news
WhatsApp officially rolls out ads
I was quite surprised to hear this announcement, and while I don’t (and have never) used WhatsApp, it’s a very bullish sign for Meta to turn a profit beyond network growth on its largest acquisition ever. In fact, I wrote about Meta’s complicated history with acquisitions and how it always seems to work out profitably. Point proven here, maybe?
OpenAI wins $200 million U.S. DoD contract
OpenAI will “develop prototype frontier AI capabilities to address critical national security challenges in both warfighting and enterprise domains.” Maybe this is your sign to not send personal data into your preferred LLM, but otherwise good news for an unprofitable subscription AI company.
Nippon Steel finally buys U.S. Steel
The ~$15 billion deal announced back in December 2023 (around the time I started this newsletter) is now officially settled, following an agreement to the Trump administration’s “national security agreement.” The deal grants President Trump the authority to name a board member, as well as a “non-economic golden share” that gives the U.S. government veto power over a long list of corporate decisions.
AI will replace many jobs at Amazon
In a memo this week (valuable read), Amazon CEO Andrew Jassy said: “As we roll out more generative AI and agents, it should change the way our work is done. We will need fewer people doing some of the jobs that are being done today, and more people doing other types of jobs…”
Overall, great news for shareholders. But the everlasting concern over AI operational efficiencies causing labour shortage? The economic consequences of that only God knows.
b. 📚 Resource of the week
I found this video very interesting.
I know for a fact it wasn’t just me who used GDP per capita as a weak measure of per-person growth in wealth or what have you every so often, so this was an insightful watch that offered a decent alternative explanation. One of the best videos I’ve watched lately as a whole actually.
—-
Thanks for reading today. Hope to see you soon (only figuratively; I’m not spying on you, I swear). Cheerios, friend.
All of my links here.
From the archives:
The technically correct word would be “overlooked,” but I find it to be a running joke when I write how “underlooked” always comes first. So I’m just going to leave it.