David Orr Profile picture
Jul 31 5 tweets 1 min read Read on X
In the outcome that $NVDA:

A: Stays dominant in AI hardware, it's going to beat $TSM a fair bit.

B: Stops being dominant in AI hardware, it's going to drop a lot. Plus, $TSM will benefit from even larger volumes of hardware spend.

$TSM seems the much better risk:reward.
The tail risk with $TSM is China invades Taiwan. But after studying modern submarine tech, I'm convinced that's very unlikely. And over time, $TSM is going to bring more of its capacity to the USA reducing that risk.
How can people see $TSM and think we're in an AI bubble?

It's 20x earnings, 30% return on capital, return on capital is trending up and the growth runway is obvious.

People always say how cyclical semiconductors are - even when they haven't been cyclical for many years.
This is not a cyclical business. Just because people say 2+2=5 many times, doesn't mean they're right. Image
$TSM can easily triple in the next 5 years.

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with David Orr

David Orr Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @orrdavid

Jul 31
Subtle part of the Kelly Criterion:

Sharpe goes up as fraction of Kelly goes down, especially from .5 -> 1 Kelly.

You should want a margin of safety on never going above .5. Which is an argument to aim for most like .25 Kelly. Image
It's impossible to put a hard number on what the true Kelly size is. Edges on bets are not fully known. So is downside risk. Correlations between bets isn't even that clear. And all of this constantly changes.
Say that you know that a factor like AI has a pretty big edge. What is that in rough terms compared to a rigged coin? Pretend it's 70/30. Full Kelly is a 40% bet, half is 20%. To get your margin of safety, 10-15% is correct.
Read 7 tweets
Jul 28
I'm seeing many posts calling an ai peak, or pointing out near record high positioning in semiconductor stocks.

I'm in the camp that this is more like 1995. The ai show hasn't even gotten started.
AI is not a fad, and the great white-collar displacement hasn't even really begun yet. And that is going to happen.

I think it cannot come soon enough. We can once again make real progress in the physical world.
"Software is eating the world" has been the ongoing theme for my whole life.

I couldn't tell you how software got much better in the last 15 years. It peaked around windows 7. That's when everything just worked well already, with only very minor improvements from there. I even had internet that worked great, the more expensive physical side of this 15 years ago - in Thailand!

The last 15 years of software have been hardcore rent seeking for the benefit of nobody except a tiny class of people. Many of humanity's best people worked at those companies because that's where the monopoly man rents came from, and where the largest pay packages were.

In many cases, software has gotten a lot worse. For example, search is worse today than it was in 2018, it's not even close.

I think this great disruption cannot happen soon enough. And I think many more companies - even operating systems like Windows - are in a precarious spot today.
Read 4 tweets
Jul 7
A few guys I know already do pretty well beating the market.

If they'd just take on a fair bit of leverage, they'd be absolutely killing the market instead.
This is the key way to use leverage well. If you're only 40% gross long - both in terms of beta adjusted and the raw gross - then you're so unlikely to drawdown even as much as the market generally during a meltdown. Never mind having a crazy blowup. The only way that's really even possible is concentration, which also does not mix with leverage.

"Stress-test at 100 % correlation: Size gross leverage so the portfolio survives if every sleeve gaps down together. (I aim for <–30 % projected drawdown at 100% correlation).
Some firms run with 700% leverage perpetually. Some run with 1,200% for long periods. They make it work just fine. They do it this way, getting more and more diversified not just assets but asset classes. The other tools the author mentions - like non-margin debt - are other layers of safety.
Read 5 tweets
May 22
Eight people asked me for my thoughts on the market setup in the last couple of days, and normally that's rare/intermittent.

I'm not sure what that's a tell of by itself. I'm a short seller first and foremost, so I guess people are finding it hard to be bullish the market? Which seems like a bullish signal at least short term.

My thoughts on the market: Maybe we get a recession maybe not, hard to say, nobody knows. And this trade war stuff essentially worked out well in the end, which people still seem in denial about. But far more important than either of those is AI. Focus on AI. I think this is 1997.
The main thing I can be sure of: AI is going to kill most knowledge jobs - lawyers, programmers, doctors, etc. It's going to kill drivers. It might even kill plumbers and electricians - at least the simpler jobs - as people can use the AI to easily do their own work? I used AI for interior design and that seemed to work great - I bet it refined my setup more than all but the high-end interior designers.

AI will be way cheaper and way better than humans.

People still seem in denial about AI or at least are underappreciating it. I still read comments once in a while that AI isn't really AI at all, that it's just a glorified search engine, etc. Which means AI has a long way to go up.

AI is going to lower everyone's costs/barriers to starting a business. But it's also going to put lots of people out of work. It should be very deflationary. But that's tricky too given the world's fiscal situation.
I am not pessimistic about this transition. Right now our brightest people are doing jobs that anyone would agree are a total waste - corporate tax lawyers, for maybe the best example. But also the people expanding $MSFT's monopoly moat. Or the analysts in our business. Or even a few more years out, my job.

It'd be a lot better if we did something real for the world. Perhaps all these smart people working in the physical world will create far more progress, rather than this extreme waste. A nice high rise apartment doesn't need to be so scarce, for example. Neither does nice furniture - isn't it strange that in the richest country in the world most furniture in the stores is junk? Maybe that'll finally improve.
Read 6 tweets
May 17
First chart are start dates of speculative rallies. These have always been tough periods for me.

Second chart are the results of my hedge fund partition. Around -3.5 Sharpe including the .7% I lost today. Image
Image
From my Q3 2023 hedge fund letter,

Autocorrelation

Militia runs with two key factor mismatches. We’re long low to medium volatile companies, and short very volatile ones. We’re long profitable businesses and short cash incinerators. This causes the autocorrelation.

There are stretches where volatile, cash burning companies trade up far more than the market as a basket. Sometimes these moves are explosive which isn’t worrying. If anything, I might get more bearish after a rapid blow off top in junky stocks. That’s often the point of peak madness, a fine time to try pressing short bets. The problem is when bad companies are persistently strong, like from May through July 2023.

There are many possible causes of sustained flows into junky stocks. A couple examples,

* A new, big technology is coming out. Charlatan management of the worst companies will reliably lean into the hype. Like AI this year.

* Too much government stimulus, causing easy money to flow into speculative investments. Like during Covid.

* Junky stocks sold too hard, too fast and are more likely to rebound for a while
Also from my Q3 2023 hedge fund letter,

Flocks of Sheep

Say that a flock of sheep starts coming from a bend in the road, but you can’t see how big it is yet. More and more sheep start coming at once and the flock keeps getting thicker. This trend could suddenly stop but I’d gamble on it continuing. You can make an educated guess that the herd is likely larger than it immediately appears given the early trend information. This is the time to gradually increase caution given all the leverage I use.Image
Read 5 tweets
Apr 30
Should have less focus on the dollar and more focus on winners and losers relative to the dollar.

Stronger MXN seems obvious
Weaker EUR seems quite likely
Stronger JPN less clear than before, but probably yes
No clue on INR, will avoid since I can't play their stocks anyway.
US stocks are screaming that bears are wrong. Flat since the tariff shock and major world reorientation began.

US stocks got a boost from the weak dollar, adjusting for that it's still basically down 5-10%, but still that's no crisis.

What happens if/when countries agree to reorient against China? Zooming out, most would say that is good news for stocks long term. And like I pointed out in a post yesterday, a 1-2 year pain period that the market knows is temporary doesn't really matter at all. Once other countries agree, there's less uncertainty than covid even.
The elephant in the room, what people stopped talking about for a month, why I think the market keeps going up:

We're on the cusp of true AGI and the fastest wave of human progress ever. Everything else seems like noise, including even a minor recession.
Read 5 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us!

:(