Green Dot Cap

L/S Analyst. Mostly tech. Not investment advice

[Green Dot Cap @GreenDotCap1tal]

My interpretation wasn’t that this was negative for the buildout because the labs will stop spending. This risk is that 1/2 political parties is going to run on slowing or stopping data center builds, and Dario and Sam just gave them a silver bullet talking point to sway public opinion in their favour

[Green Dot Cap @GreenDotCap1tal]

This chart continues to amaze me. What a disaster

[Green Dot Cap @GreenDotCap1tal]

Not yet seeing anything that makes me think a big rebound is coming. Data centers will be great and limit downside, but need a better rate environment and more labor availability for a broad improvement

That said, the market is very bifurcated though. Small/local projects suck, while large GCs are doing quite well because of DC demand. PCOR is more levered towards these large GCs and should outperform overall non-resi

[Green Dot Cap @GreenDotCap1tal]

Is Procore about to get a huge construction bid?

Stock going back to $60 in 2-3 months

[Green Dot Cap @GreenDotCap1tal]

Not worried about this Anthropic number in the slightest. Their summer update was always going to imply a slowdown - their ARR is usage based and we know the summer is seasonally "weak" for the labs

They probably put out this number, knowing it would be a disappointment, so they can then release a much higher figure in the fall ahead of their IPO

[Green Dot Cap @GreenDotCap1tal]

Exactly, I still like it. I think upside is $90+ and downside is ~$45 over the next 6-12 months, so good R/R still.

A lot to like with potential AI upside (or at the very least less disruption risk than others), margins going MUCH higher (no reason they can't get to 40%+), and an improved management team. I do ultimately think non-resi construction growth inflecting is what makes this a great or not though

Biggest risks are AI disruption (which I haven't heard any new competitors gaining traction), non-resi construction volume staying weak, and the fact that they don't have GAAP valuation support

[Green Dot Cap @GreenDotCap1tal]

@P_Remarks I’ve learned that first lesson time and again with CRWD. Even when numbers were bad, it would go down 7% and then be back higher 2 days later

Agree broadly on software. Selling my MDB, SNOW, and TWLO. Keeping DT, FIG, and PCOR. Might chuck some into ORCL and INTU

[Green Dot Cap @GreenDotCap1tal]

@DanielSLoeb1 AI slop making it into the investor letter now?

[Green Dot Cap @GreenDotCap1tal]

@TBU12345678 Can you DM me? Working with them right now

[Green Dot Cap @GreenDotCap1tal]

@sama That’s nothing. I just have chatgpt talk to my kids now. It’s completely freed up my schedule

[Green Dot Cap @GreenDotCap1tal]

@JaredKubin Most people I talk to are $350-$400bn. I know spec sales was saying $300-325bn

Wild numbers when you step back

[Green Dot Cap @GreenDotCap1tal]

Quick thoughts on the GOOG quarter:

- Search was in-line with expectations. 2Y stack flattish, but deccelling for the next few quarters
- Youtube finally beat
- Cloud was great, but whatever? Everyone knows cloud is accelerating and going much higher into 2027. This doesn’t really change numbers
- EBIT margin light on the step up in G&A
- Cloud incr margins staying at mid 50% (very important as cloud becomes a bigger mix of $ growth)
- Capex raised, as expected. 2027 capex probably going to $350-$400Bn

The stock is well owned, numbers don’t really move this Q, and the debates don’t change in either direction - so nothing to attract the incremental buyer for now and would expect the stock to be choppy / rangebound in the NT

That being said, the stock is ~18x on 2027 numbers. Tough to see too much downside. I also think the visibility we’re getting into future growth should help re-rate the multiple over time (vs the majority of revs coming from ads)

[Green Dot Cap @GreenDotCap1tal]

@P_Remarks Nooo Okta has NEVER disappointed on a reaccel before

[Green Dot Cap @GreenDotCap1tal]

I wonder if this was at all inspired by the AI 2040 project?

Regardless, one of the world's leading experts just said that "a system that exhibits all the cognitive capabilities the brain has, is probably only a few short years away" and that recursively self-improving systems will be here soon.

We are approaching a point that will ultimately lead to utopia or annihilation, and I don't think most of us (including government officials) understand how fast it is approaching

If it took 50 years to reach AGI, ASI could happen in the week after

[Green Dot Cap @GreenDotCap1tal]

Personal view is the benchmarks don’t really matter anymore, but I’ll at least agree that direction of travel is positive for them.

Genuinely asking, how are you assuming they will monetize their model in the future? Given how big they already are, I just struggle to believe that it will provide enough uplift to their ad engine to justify $200Bn of spending. And it isn’t clear to me what new business they can enter to monetize it

Maybe I’m overthinking it and the correct answer is just that if they can’t actually monetize it, they’ll just sell the computer to Ant and OAI at a 40% margin. Then the shitty ROIC they’re seeing right now is temporary one way or another

[Green Dot Cap @GreenDotCap1tal]

@obsidiancap1 @evrgn11112231 Nothing released makes me change my view. They’re still far behind Ant, OAI, and Google. They’ve caught up to open-source though, I’ll give them that

[Green Dot Cap @GreenDotCap1tal]

@ContrarianCurse EXACTLY. This is why it’s been so important for them to develop a frontier model or create some new AI driven business - the revenue/growth required to get a good ROIC on this level of spending is more than they can get just from the ad market

[Green Dot Cap @GreenDotCap1tal]

@PythiaR *Guy who is in charge of AI saying AI doing well*

[Green Dot Cap @GreenDotCap1tal]

@TMTLongShort curious if you have any view or data here

How does model revenue/growth trend over time (ie Opus 4.8)? Specifically, I’m trying to think about what happens after ~6 months once open source has closed much of the gap in capabilities - obviously not the entire gap

I would also love to understand how this has compared across different generations of models (ie Opus 4.6 vs 4.8)