In the Software/AI thread, @PaulWBryant asked us to focus on companies with high growth prospects. I think it’s useful to pretend we’re writing a Netflix documentary, and so the familiar trope of “follow the money” is worthwhile.
To whit:
• Google just announced they will spend $175B to $185B on AI Capex this year. This comes on top of the $91B they spent last year, which was 74% higher than 2023.
- 60% of that to servers
- 40% to data centers & networking equipment
• Amazon recently announced they will spend $225B in AI Capex this year.
• Microsoft is at $145B per year in AI Capex. That’s like 45% of its revenue - for a software company that’s astounding!
• Meta will spend between $115B and $135B in AI Capex this year.
• Oracle is in the tens of Billions of dollars, relatively small potatoes, lol.
• Tesla didn’t break the numbers out, but a significant portion of the $20B in Capex it’s going to spend this year is for AI build out.
I go back to the early 2000’s and think about Amazon. Saul and I suspect most of the board, sold out of Amazon since its bottom line wasn’t showing much of a profit. Turned out what was really going on is that the company was investing in infrastructure, not just to run the eCommerce business, but to develop and scale up its newly created Cloud Computing business - AWS. Since Mr. Market (and us) for the most part didn’t understand the future that Amazon saw, the company poured almost all of its profits into developing this infrastucture without calling it out specifically (that I recall, but could be wrong on this point). And so it looked like Amazon was barely profitable, with some people assuming the costs of running the eCommerce was high and/or Amazon was keeping prices too low.
Today it’s different - the “AI trade” is a stock market thing, and so companies aren’t afraid to call out the spending they’re doing on building AI infrastructure out. Heck, Amazon today, despite its high profits and FCF is tapping the debt markets to get more money for the build-out.
So, now, where is this $700B (or so) going?
• Nvidia is obviously one beneficiary. Although it’s worth pointing out that the new growth is more in inference than training, so keeping track of inference chip competition is needed.
• I think the networking hardware companies are likely beneficiaries: ALAB, ANET, CRDO.
• Compute needs memory, and for AI that’s currently HBM (High Bandwidth Memory), which is Micron and Samsung (and SK Hynix, but I don’t know how to invest in that). Samsung has lots of businesses, so I think it’s hard extrapolate its memory business as having a big impact on its stock price.
• Data center build out companies. Perhaps Vertiv, but I’m sure there are others.
• AI Infrastructure companies: NBIS, CRWV, IREN, etc.
• Chip making companies. I thought about TSMC, but it’s production limited and bring new fabs online is a slow, expensive process, and there’s the geopolitical risk. Samsung is a player here, too. Companies like Amazon and Google are making their own ASIC-based AI chips, but I don’t know how to invest based on that - do you?
Now, with all of this hardware, obviously software is next. Most of these companies aren’t public yet. Claims are that OpenAI has $20B in ARR already. Anthropic isn’t public either, but Claude and its variants are all over the news, and the company reportedly has 300,000 business customers. Google’s got Gemini and Veo and other products, but they don’t break those out separately.
Part of my problem is that if, as @wpr101 wisely stated, we’re looking for companies that can 3x-4x in 1-2 years, then these big companies, no matter how innovative and smart, are probably not the targets for our investing dollars (expect maybe for a non-Saul portfolio portion one might want to have for other reasons).
With the big dollars being spent, what are the companies of a size that can 3X-4X in 1-2 years we should be looking at? Outside of the networking hardware companies I mentioned above, I don’t know. And, what software or application companies will find AI markets, even niche markets, that will propel their growth? Some of this may be too early to tell, but it’d be good to get some on our radars.
