I started a Substack called The Investment Case.
I originally built these investment cases for myself.
I wanted a faster way to understand a company before spending days on filings, earnings transcripts, expert calls, sell-side research, investor conference transcripts, and presentations.
Not the final answer. A serious first pass.
What does the company do?
What is the thesis?
Where do informed investors disagree?
What would break the case?
They have saved me a lot of time, so I’m publishing them in case they help others too.
They are produced by AI workflows I built, using filings, transcripts, expert calls, sell-side research, investor conferences, and presentations as the source base.
I posted the investment case on Vertiv $VRT, a business between the grid and the chip that scales roughly 1:1 with gigawatts deployed.
It adds ~$1 of revenue with ~2c of fixed capital, collects cash before paying suppliers, runs 20%+ ROIC, and converts 100%+ of adjusted net income to FCF.
I published the Mastercard investment case today.
The network is not really the question. The core moat is wide, tested, and still intact.
The debate is growth: services, cross-border, stablecoins, sovereign A2A rails, and whether Mastercard can keep monetizing new payment flows as the card core matures.
Thank you to everyone who read and shared feedback on the first few Investment Case posts. They got more attention than I expected, and the comments were genuinely helpful.
The most useful criticism was that some pieces read a little too much like public explainers. Easier to read, yes, but softer than the underlying work. That is probably my fault. I was trying to make them accessible and sanded off some of the density that made the memos useful to me in the first place.
So I went back and tightened the first five posts. Going forward, I’ll keep them closer to the source memos: serious, investor-focused, and willing to stay dense where the argument needs it.
I’ll publish Mastercard later today.
Meta’s Business Agent launch matters because it is not in the base case.
The base case does not assume a new AI revenue line. It only assumes AI improves the core ad machine. So all AI-native revenue is upside.
5-year forecast sets a low bar: capex efficiency falls from Meta’s historical ~0.9x to ~0.3x, and ad revenue still grows at a +16.4% CAGR.
If Business Agent, paid AI tools, or other AI-native revenue starts to matter, that upside can get meaningful fast.
At 0.6x capex efficiency, the revenue path moves above 25%.
I’d start with Meta.
The question is not whether Meta’s AI works today. Q1 2026 says it does.
The question is whether one capex stack can keep lifting both content ranking and ad pricing fast enough to pay for the depreciation it creates.
Starting points, not recommendations.
The business. The debate. The breakpoints.
I started a Substack called The Investment Case.
I originally built these investment cases for myself.
I wanted a faster way to understand a company before spending days on filings, earnings transcripts, expert calls, sell-side research, investor conference transcripts, and presentations.
Not the final answer. A serious first pass.
What does the company do?
What is the thesis?
Where do informed investors disagree?
What would break the case?
They have saved me a lot of time, so I’m publishing them in case they help others too.
They are produced by AI workflows I built, using filings, transcripts, expert calls, sell-side research, investor conferences, and presentations as the source base.
The first five are live and free:
$META - one capex stack, two AI channels, one unresolved question
$BKNG - an OTA built on supply, tested by distribution
$LLY - the efficacy ceiling on a dated clock
$CPRT - a wide moat compounding through a unit decline
$CELH - a narrow distribution moat, gated by one partner
https://t.co/RfHrJ2pZgk
7/ The cleanest read for me:
CELH is going through an air pocket from the SKU reset.
But the underlying Celsius brand may be healthier than headline scanner data suggests.
That creates an interesting setup if the reset starts showing through in the numbers.
1/ I think the market may be over-reading weak $CELH scanner data.
The headline numbers look soft, but JPM makes a useful point: scanner data is being distorted by SKU rationalization.
The brand may look weaker than it actually is.
6/ The Costco/Kirkland concern also looks less severe so far than the stock reaction implied.
JPM estimated Kirkland private-label energy had reached ~6.7-7.7% share of Costco’s energy drink category in early data.
Still early, but not obvious category disruption yet.
@bjmtweets Thanks for sharing. You got most of it right, but the mechanics are slightly off. They are not discounting to clear, instead pulling back from shelf, which is why there is a temporary decline in TDP.