The current AI industry stack:

1. Mostly monopoly or oligopoly compute companies, from ASML to NVDA to MU etc.

who mostly sell to,

2. Three serious hyperscalers.

who mostly sell to,

3. Two serious model companies.

who sell to,

4. All sorts of companies, large and tiny. What seems clear:

1. Is most vulnerable to a boom/bust cycle, since revenue can collapse 80%+. But stands to win the most if compute demand stays strong. If demand doubles from 2029 to 2032, they are still very cheap. If demand maintains 2029 level long term, they are reasonably priced or a bit cheap. If demand collapses, they will collapse.

2. Has good upside if compute demand stays high, low downside given long term contracts. Heads they are worth double+, but don't benefit more than that from an ever increasing demand scenario. Tails they tread water for a few years as they work through over investment.

3. Are an extremely binary bet which comes down to, "Is it AI hardware improvement driving most AI gains, or will rearchers/other going to drive it?" If the former, these are going to be terrible commodity businesses. If the latter, they will become the most valuable companies in the world and it's not close. I think the bullish odds are very low, but if someone disagrees these are a *screaming* long.

4. Seem screwed generally.

$AMZN $MSFT and $GOOG are just expanding their existing cloud business/capabilities. It's a long term oligopoly that I doubt anyone can compete with. I was worried about competition 7 years ago, but the sheer scale of these operations today makes that seem near impossible.

$META is randomly joining because Zuck has no other real way to expand the biz (hence the old metaverse catastrophe).

The rest are a big question mark. And China never seems like a real threat, because a business would have to be insane to rely on Chinese systems which have lots of extra risk. Three companies buy huge volumes from semiconductor companies. These three simultaneously lock into a few year contracts to sell that compute. Given those contracts, the compute business just isn't that risky even in a bust scenario. Maybe they lose a year or two of earnings.

The (software) companies signing up for contracts to buy this compute are taking a bigger risk, and they are essentially forced into that decision. That's why they have been exceptionally weak today. Their old business went from wider moat capital light to weak moat capital intensive... and lots of them are just going to be commoditized.

And of course AI hardware companies make a ton of money until the bust happens. But that bust will be spectacular... and seems quite likely short to medium term.

I'm glad $GOOG isn't trying to be the most advanced frontier model company, which eems like a fool's bet. A cash incinerator. $GOOG clearly has the resources to do it, but just don't seem to be.

It's better to be a company:

* With a scale advantage running hardware, which there are really just three of.

* Already have wide moat businesses that benefit from AI. For example, this new youtube mode that automatically shrinks videos based on what matters and doesn't. A better user experience and packs in more ads.

* TPUs, which were specifically designed to run AI efficiently years ago so they're not just playing catch up. It's very possible that $GOOG will have the best AI simply because these have the most power in the end. If I'm right about both of these (and maybe not):

1. AI improvement will depend mostly on hardware advances.

and

2. There will be a boom/bust cycle in hardware, causing hardware to be vastly oversupplied.

Hyperscalers should be the winner. They'll take a brief one time hit as the latest wave of hardware they ordered is oversupplied. But from there, they can stop orders (causing AI hardware companies to collapse) and demand will catch up within a few years so it won't even be that big a loss. And meanwhile, their physical world scale advantage will continue long term... it seems so hard for anyone to catch up in that game in a real way.

This thread discusses my long investments sorted by bucket. I'll often skip tiny positions out of laziness.

AI hardware comapnies:

$TSM $GOOG (partial) $AMZN (partial) $4966 $5393 $6920 $8053 (partial) $6490

AI hardware companies has been an easy theme to ride for the last 3 years.

Headline valuations aren't *totally* crazy. For example $TSM trades for 40x. These companies are growing so fast and biz quality is so high that those valuations could really be fundamentally justified, which is the main reason I remain long.

Another reason to stay long: if a dotcom comparison is right - and that seems kind of obvious at this point - we should expect far crazier valuations at the peak of the bubble. Based purely on multiple expansion, we could see 100% upside. I'm not big on investing this way. I almost never think this way. But this one seems pretty damn straightfoward and there's a time and a place for it. Japanese conglomerates:

$9435 $9042 $8053 (mostly) $8058 $8078 (partial) $8001 $8472 (partial) $2768

They are just good capital allocators, good businesses trading for good valuations. Hikari I've been pounding the table on and is the outlier large position of the group since it seems both best and cheapest.

Nobody actually knows what's going on with Iran, what the real deal was, who is even in control of Iran today, etc.

All I know is the price of oil this whole time hasn't lined up with the narrative that Iran successfully applied max pressure via the SoH on oil prices. Maybe the USA capitulated to Iran, sure. I'll even assign that a solid 35% odds.

Maybe we already, quietly got a very favorable outcome after we killed 50% of their senior leadership.

Maybe Iran capitulated to us, and the deal is face saving nonsense.

Nobody knows.

The amount of time an AI can spend on a task: 6 weeks of programming 24/7, probably with more capability than a human.

Knowledge jobs are totally cooked.

I'm very bullish on AI over 20 years.

But now I'm confident that AI is most likely going to be a boom/bust cycle in the shorter term. Probably within a few years. The issue:

1. Most AI end user spend is just on coding. **Coding is where the money is actually coming from to pay for today's extreme capex**.

2. Compute costs will keep dropping a lot because hardware keeps improving rapidly. Even if the volume of capex doesn't keep going up (and it is today), the cost of software development will keep going down as the hardware gets better. Software developement isn't that hard a problem and it's easy to how **AI is going to drive development costs very low long term**.

3. **Back in 2020, it made sense for software companies to continuously improve their products because the cost of developing software was going to be roughly flat long term**. Thus, money invested into development was a moat because it would cost a competitor way too much to ever catch up. The bet was that each niche piece of software would be winner takes all and it would never make sense for anyone to catch up.

4. But now because of 2, 3 is no longer the case. If you assume software will cost 99% less to develop in 10 years (I do), software's moat sucks. You keep investing into something, where the future cost of that thing is going to be lower.

At some point the market will realize all of this. And then software companies will stop making this poor investment. And then AI capex will overshoot almost for sure vs the short term demand. And then AI companies are going to drop hard / we have a bear market.

Longer term, I expect companies to spend big on compute on things besides software. But this dominating variable is key today.

3 years ago I wrote that $NVDA would become the most valuable company in the world on the back of AI. Today, I'm nearly as confidently calling a short term boom/bust cycle, unless companies quickly find something else to spend huge on AI. This post is *not* meant to be a short term call by the way. I have no clue short term. There's even a good chance the market will come up with a short term narrative for how AI is in fact good for software companies due to cost savings, even though it's really terrible.

Anyone have counterpoints to this / other data sets?

If this is accurate, which I have no clue, and it craters, then $QQQ is going to go into an extremely fast bear market.

https://twitter.com/zerohedge/status/2066976979312607248

This is the key variable for the entire stock market. And I just don't see that many people talking about it.

So please, and comments are welcome.

SOXX/NVDA Choppy sideways until 2015, right when value investors starting performing terribly. The pair makes its really wild move starting October 2022, when markets bottomed. This, AI, is really what saved the market from higher rates.

I'm trying to understand YTD's factor action:

1. Through the end of February, there was a rush to safe investments because the Iran was was obvious/likely to happen. This is why my result was so good leading up to the event.

2. Next, the right tail risk from the Iran war didn't happen. The risk was that Iran would blow up a lot of oil infrastructure and/or cause serious damage to the US military.

3. I gave back Feb's gain at the same time. It was a very sharp move down where I lost 15%.

This is my partition only. It's basically the same move as gold over the window - save haven assets going up a lot and then dropping a lot *because the war went well, not poorly*. I feel pretty confident about this answer now.

Took me quite a while to work through what this was. And I bet some guy betting on only factors knew week by week while it was happening

Safe haven assets, and even some oil plays, peaked at the end of February because:

People knew the war was coming and the move got front ran big time. Gold went up the first of March, only a single day it went up, and then immediately it started going down.

One interpretation is the market realized the war was over as soon as Iran wasn't able to inflict any real damage. That's my interpretation because Iran was supposed to to retaliate cause serious damage if we attacked them. Specific examples:

We didn't know if their Chinese supplier "carrier killer" missiles would work. They didn't.

We didn't know if their ballistic missiles would reduce Israel to rubble. That didn't happen, either.

We didn't know if Iran would be able to shoot down our F-35s with modern Chinese tech. And instead, when those systems failed completely, China disappeared a bunch of their top military researchers and commanders.

We didn't know how much of their top leadership we'd kill. And we got like half the top guys.

The only place we "lost" was the near certain bad outcome, that Iran would be able to apply pressure to the SoH. But even there, the oil price is saying that thesis wasn't nearly as bad as everyone thought.

I'm in a particularly bad drawdown in my hedge fund partition, -24% before fees or -18% after fees (I have to report both numbers for compliance reasons). I'm personally around flat YTD now.

This year I've been leaning far more into diversifying factors, like global low volatility stocks. This because the hedge fund is multimanager now and the other PMs don't have that same exposure/risk. Without doing that, I might be in more like a 15-18% drawdown instead. That's still bad, but basically in line with my past bad drawdowns. Fortunately, this model is working and the fund level is still performing well YTD (+18% net of fees).

When I zoom out and look at the market environment, being flat YTD is I guess not bad. This has been a particularly bad environment for what I do.

I did a couple degrossing / risk management moves on my way down already. Each time I do this it locks in a 1%. However, if I hadn't done this voluntarily, I would be in a much worse drawdown right now and have been forced to degross anyway, which would have increased greatly the long term loss being locked in.

I think my job through this is to survive. To keep risk managing. To probably degross even a bit more than normal given the persistance of this move. And then just keep planning and be ready to attack when the time comes. Another thought: It is incredibly lower stress to run a 200% gross levered investment vehicle instead of a 400%+ one. 200% feels like a damn a cake walk risk wise to me now.

Jan 2021 was way worse than what we just had.

In Jan 2021, the most shorted stocks went up ~55% as a basket in 3 weeks. 18%/week moves. And very important: this happened while the general market was flat.

In their recent jump, the most shorted stocks went up ~35% as a basket in 8 weeks. 4.5%/week moves vs the above 18%/week. And this happened while the general market was up 19%. Really, that's just in line with the beta - it's just what you'd expect the short term moves to be.

A small counterargument is that breadth was weak in the last couple of months, with $RSP only up 10%. If someone has no exposure to AI stocks, and they were short lots of junk, they probably made nothing on longs and lost big on shorts. They had plenty of time to risk manage, vs Jan 2021 where the gap ups were extreme. Thinking on this, I guess that's the factor that just blew out, right?

That was the bet.

Long 150% $RSP
Short 10% $SOXX (Negating $RSP's exposure to AI)
Short 30% most short

Those guys lost bad.

It's really been since just April 13th that my results have been so bad. Before that, YTD alpha was very strong.

What changed on April 13th specifically? For people who specialize in market regimes: What factors went wild from April 13th until April 22nd?

I've been reviewing what's been going wrong for me since March 1st in the hedge fund. I'm personally in a 20% drawdown (-15% after fees), which is particularly bad for me.

Usually during tough periods it's my short book causing problems. This time it's not. My shorts are only losing me 11% since then, while I have been 165% gross short and the Russell 2000 (best proxy for my short book) is up 9%. Gross adjusted, my shorts should be down 15%. Plus, my shorts run higher beta so really they should be down 20%. So alpha on shorts has actually been strong. The market did go up real fast, so of course I lost a lot, but I still consider this a big victory.

It's the long book where I am getting *killed*. This is so different from past junk meltups.

Stocks like $9022, JR Central who's main asset is the famous Shinkansen that links Tokyo with Nagoya, Kyoto and Osaka. This is one of my big losers. This trades for 7x earnings with solid earnings growth. More of my big losers:

$3288, Open House, a rapidly growing home builder, who's signs I notice constantly when going around the city. This trades for 9x earnings.

The more bottlenecks there are to AI, the better the AI business becomes, right?

If the world is beholden to $NVDA, $NVDA makes all the money.

With 6 different bottlenecks, AI supply is capped and prices can just keep going up. Committed hyperscaler spend becomes great. And aren't these bottlenecks going to last at least a few years?

I'm still skeptical on memory specifically. But some of these ... who knows.

Militia Capital my hedge fund lost 10/12 of the last trading days. Pretty brutal stuff! A straight line down of death. Alpha is sometimes correlated, even when the bets aren't fundamentally related whatsoever.

In this case, the explanation is incredibly obvious, which is rare. This time is happening because money is flowing out of other stocks and into AI stocks. The last time it was obvious what was happening was during covid, when the infinity money printer caused money to flow into the most speculative and trendy trash possible.

It's really nice to see why this happens. Most of the time you don't even really get to know why. It just seems like market spirits.

When things like this happen, long/short funds are getting slaughtered on their long beta, they have to degross/derisk as a group. Which pushes up bad companies that are more shorted on average, and pushes down non-AI companies.

These are the of moments that make long/short investing so hard. And who the hell knows when it will end.

I think I've solved the game theory of the Hormuz situation.

The IRGC doesn't care about their people. But they care a lot about their oil revenue.

"End your blockade or every day we will blow up 5% more of your oil infrastructure." That's how to end this very fast. I'm pretty sure it's that simple. And the people running this strategy probably already know. But I still think it's a good, loud and clear wake up call for the world that they let themselves get into this situation. So letting the world see cause and effect is important.

People investing with Lynch won a lot less than his fund's long term performance because they attempted to time the market.

It's genuinely sad that people do this. I think it's an argument to only even have the market open once a month for an hour. A big upside of a hedge fund structure is that people do only get monthly redemption windows, so only one chance a month to make a mistake. And they're expected to give advanced notice. And because of the human/relationship element, they are a lot less likely to do it.

To young people considering living in Japan or Asia generally, this area of Osaka is a very good option: $300/month in an outstanding (truly) location. Tiny room, but good enough for a bed and desk.

suumo.jp/chintai/bc_100…

People who want to break into the investing industry need to realize just how bleak things are for the legacy active management industry. Their marketing-first (marketing-only) playbook does not work anymore, and this will keep getting worse. When one of these legacy marketing professionals tells you how to manage your own career, do not listen to them.

You have to be different.

And, you actually have to add value / have an edge today.