Premium subscription to my newsletter gets you: - Over 50 equity research reports per year - Membership to our Discord community - Access to our outperforming portfolio - Live support - Q&A Join with the link below: capitalist-letters.com/

This year both $MSFT and $META jumped almost 50% from their lows. Last year it was $GOOG. These are trillion dollar companies. And you still believe markets are efficient?

The “futuristic” bubble on X is insane. People think agents will disrupt platforms and payments all at once within a few years. Meanwhile, Europe would still be struggling to adopt card payments if it weren’t for Covid. Many small businesses in Germany are still “cash only.”

Compute demand is just insane. Neocloud revenues will grow 8x from $50 billion this year to almost $400 billion in 2031, according to Synergy Research Group. $CRWV and $NBIS will together account for $150 billion of that. Compute is the new gold.

JP Morgan on $CRWV: “pricing & margins will overwhelm the increase in debt.” This is exactly what I’ve been trying to explain for months. As you see in the model below, if the demand holds up and $CRWV executes on capacity targets, interest expense as a percentage of revenue will drop from 22% levels now to 7% in 2030. Note that the model disregards any price increases and assumes just around $10 billion revenue per GW. In reality, we’ll likely see this rate go higher as short term rates are currently 3-4x the long-term rates. $CRWV is not a leap of faith trade anymore, it’s a perfectly logical investment.

What types of aggregators are threatened by AI agents? Let me put it clearly: If your network aggregates commoditised supply, like $AMZN, you have a problem. You won’t likely lose sales, but you will likely lose some ad revenues. If your network aggregates differentiated supply, like $BKNG, agents won’t make a dent in your business. Commoditised supply is the best environment for agents, pick the best price/performance. Easy. When supply is differentiated, discovery, taste, preferences gain prominence, and agents won’t be of much use there.

Bill Ackman: “If a company is increasing growth capex while earning above average returns, you should applaud.” The market is finally applauding $META after seeing all the great products they cooked instead of spending all that money to do lazy-ass buybacks.

I am a big $META bull and I love Muse, but stop with that “Muse disrupts everything all at once” bs. At this point, people are simply not thinking realistically, they just enjoy writing futuristic stories. There is no way Muse is making a dent in $BKNG. Remember, being far away of your time is practically the same as being wrong.

This is going to drive some people crazy but I’ll say it: $CRWV is a better bet right now than $NBIS. I own both but I shifted a large part of my $NBIS position to $CRWV after a 9x return. Here is why: I previously made the case that $NBIS’s lower leverage was just an illusion and it would leverage up with scale. We are seeing this as $NBIS quarterly average interest rate on its debt has been increasing, while $CRWV’s has been coming down since it’s getting increasingly derisked as the revenue scales. Exactly because of that scale, its capex per total revenue is substantially lower than both $IREN and $NBIS. Meanwhile, it’s trading at 2x 2027 revenue while $NBIS is at 6x. As I said, I own both, but my $NBIS average is around $30. If I were considering an entry now, I would definitely go with $CRWV.

Nice to see Capitalist-Letters on the list. Thanks @StockJabber

Dropping inference costs is incredibly bullish for AI infrastructure. Token consumption explodes as inference costs come down, Jevons Paradox. And compute supply can’t accelerate due to power constraints. Result? Compute prices are rallying. $CRWV $NBIS $MSFT $AMZN $ORCL

Cost of AI performance isn’t same as cost of intelligence. While both are in downtrend, cost of frontier intelligence comes down way more slowly. Combine this with 24x expected growth in token consumption by 2030 and you see why cloud businesses will still print. $NBIS $CRWV

Copper demand is exploding. Bloomberg expects copper demand to grow 40% by 2050. Energy transition, AI buildout and robots will be the main drivers of demand. No new supply coming over the next decade. Copper shortage is here.

Is copper really this obvious? Futures already hit all time high. Meanwhile, S&P Global says supply will decline by 7% between 2025 and 2040 while the demand will grow by 50% in the same period. Copper super-cycle is coming.

Neo-clouds will literally print money. $NBIS’s on-demand price for H100s was $2.95/hour at the beginning of the year, and it’ll be $4.50/hour from October 1. A 50% increase in just 9 months. This will likely be reflected in long-term pricing as well. We could see long-term prices go from $10 billion per GW to maybe $20 billion. It makes sense, as short-term, large-scale pricing is around $40 billion per GW. Either short-term prices will come down, or long-term will go up. The latter is more likely given the supply constraints.

Michael Burry doubles down on $NBIS short.. You know what happened last time he did this? $NBIS posted blockbuster results, shot up 20% a day, ripping faces. Why repeat?

We should think about taking some profits on $META as Cathie Wood has entered the stock. If you don’t know how strong a top signal she is, here is a reference for you: She said last year that $BTC was going to $1.5 million. It’s down 20% since then.

Iran war will escalate after the midterms. It’s obvious. Trump promises a deal after the midterms to prevent a slump in the polls. But he also says: “I’m not running and I’ll be here for 2 more years.” He’s like, “I don’t care. I’ll can blow you up.”

Literally...

AI will completely change advertising, and $META will be the biggest winner. Two reasons: McKinsey predicts 80% of advertisers are planning to purchase AI ad formats over the next 12 months. If AI agents will do transactions, there is nowhere better to publish ads. This means the winner in the personal AI assistant market will have a huge advertising opportunity. Nobody is better positioned than $META to win this market, as it already owns consumer distribution with over 3 billion daily active users. Second, those ad budgets flowing to AI interfaces won’t emerge out of thin air. They’ll be dollars reallocated from other channels, mainly search. Thus, $META won’t just keep its own ad allocation in social, but will also take the lion’s share of ad dollars reallocated from other sources to AI interfaces.