NairoLB

Rational Optimist arguing over stocks, marco events and anything that makes money | MoM | GiG

[NairoLB @NairoLB242]

Going to be active on X again.

Substack stays my main platform, that's where the full research lives. This account is for shorter comments, closer to Substack notes.

To get you up to speed, everything on there so far:

https://t.co/u01qkHUodX

[NairoLB @NairoLB242]

Who Actually Keeps the AI Savings?

I came across an old Charlie Munger quote about textile capex while reading @gregoryblotnick, and it got me thinking about AI.

The quote is basically about one of the easiest mistakes to make when looking at technological investment: confusing how much value a technology creates with how much of that value the company actually gets to keep.

Munger was talking about textile companies buying better machinery.

Every few years, a salesman would come along with a new machine. It was faster, more efficient and cheaper to operate. The maths looked obvious. Spend the money today, save enough on costs over the next few years, and the machine effectively pays for itself.

And importantly, the machine really did work.

The cost savings were real.

But what Munger talked about is the second order consequences of these cost savings.

The machines worked. The shareholders still didn't win

I think he means it like this: If one textile company could buy the new machine, so could everyone else.

Everybody's costs came down. The industry became more efficient. And then competition did what competition does. Prices fell.

Instead of the textile company permanently keeping the cost saving as higher margins, much of the benefit was passed through to the customer.

Munger's point was thus that companies would repeatedly show him projections explaining how a new machine would pay for itself within a few years. What those projections never showed was the second step: how much of the saving would actually stay with the company once competitors made the same investment?

How does this translate to today?

Obviously, when I read these dynamics around capex, I started thinking about how this could repeat itself with today's AI capex.

A large part of the discussion around non-software/AI-native companies' enterprise AI still starts from the first-order impact.

AI automates a process. The company needs fewer employees. SG&A comes down. Margins go up. EPS goes up.

I think that is probably right initially.

But Munger forces you to take the second step.

If one mining company can use AI to automate parts of its finance department, procurement process or back office, why can't another mining company do exactly the same thing?

If both companies eventually save 20% of the cost of doing something, the industry cost curve has simply moved lower. In which case, in a Bezos-ian fashion, "your margin is my opportunity" becomes a thread.

So how should I go about modelling AI margin savings then?

Say a company believes AI can remove $100m of annual costs.

It is easy to put that $100m into next year's EBIT.

But should it still be sitting in margins five or ten years from now?

Because if competitors can make the same investment, some portion of the benefit should eventually be competed away.

And this matters even if you value the company on next year's earnings. A stock trading on 20x next year's EPS is not being valued purely on the cash it generates next year; that multiple reflects what investors believe about the durability, growth and risk of those earnings well beyond that period.

If that $1 of incremental saving is a temporary advantage which disappears as competitors catch up, the market should also apply a lower multiple to it. 
A permanent $1 of incremental earnings deserves a very different valuation from a $1 productivity windfall that disappears within three years.

So how should I go about analysing AI savings?

It's:

AI investment → productivity or revenue benefit → competitor response → value passed to customers → value retained by shareholders.

Rather than:

AI investment → cost saving → permanent margin expansion.

This approach changes what I want to focus on.

I would be more sceptical of businesses where the entire AI thesis is based on taking costs out of relatively standard functions that every competitor has access to.

Those benefits may be real for the next few years but I would be careful about putting them into terminal margins and paying a high multiple for them.

The more interesting companies are probably those where AI can drive the top line and where something around the AI makes that revenue difficult to replicate.

So what was Munger's Textile capex, is today's companies' AI investments. The difference will be if instead of just saving costs in delivering their products, the "intangible AI capex" can actually help out top-line innovation and differentiation instead.

NLB

[NairoLB @NairoLB242]

Some interesting thoughts around how accounting can help out the overall AI trade/narrative but also other industries which are consumers of AI...worth a read imo https://t.co/haathB15B7

[NairoLB @NairoLB242]

Semi caps are down today, and more broadly it looks like AI is seeing some risk-off.

What I find quite interesting, though, and something that is making me almost bullish again on the AI trade, is a report that surfaced three days ago.

After the spectacular blow-up of the Situational Awareness Fund, the big story was Citadel buying the public portfolio in an off-market transaction. I was fairly certain they would get rid of those positions quickly. You buy great positions at a discount, and some of these names subsequently rallied 20-30%, and in some cases 50%+ from the late-July lows. Seems logical to lock in the profit.

But that means these stocks rallied this hard despite billions of dollars of selling pressure.

That selling pressure is now gone. And in an environment where AI is still the talk of the town, retail investors are leveraging up again, and rate hikes are becoming less likely, I don't see why these names wouldn't just resume their rally.

The obvious question is whether everyone who wanted to position has already done so, meaning the buying pressure we are waiting for simply doesn't materialise.

Still, if the market could bounce this much despite such strong selling pressure, it is worth thinking about what happens when that seller is no longer there.

Maybe today's weakness is actually a pretty good buying opportunity.

P.S. Chips, which SALP was short, saw continued pressure despite what was likely billions of buying pressure. If that buying pressure is now gone and we get some weakness in GM, especially if Nvidia's planned price hikes are too late or not well received, we might just see the same dynamic play out here, just in reverse.

NLB

[NairoLB @NairoLB242]

I have to say, though, some of the names lifting on the back of this just don't make sense other than through basket and mechanics. $CRM, $NOW, $ADBE . These are $100 billion market cap companies. They will not be PEs' first target...Am I smelling a short opp here? JK, as you know, I'm long Software but just saying...

[NairoLB @NairoLB242]

If you wondered why software names lifted today, very abruptly and with no obvious news attached, this is why.

Silver Lake is reportedly in talks to acquire Workday. A c.$40bn market cap. WDAY up 15% on that shouldn't be a surprise. The question is why the rest of software lifted with it.

Quite simply, shorts got scared.

The big trade this year was short software on AI fears. Some of it unwound over recent weeks as momentum switched, but short interest in these names is still high, and plenty of people will have reshorted into the rally. What today introduced is a tail risk becoming more likely.

You’re short a company, it gets taken out at a 30% premium, and you've lost 30% in a single print. That's gap risk. This isn’t a just crack in the thesis. This is an event that is going to f you no matter how bearish you are.

The reason the reaction was this violent is that not many saw it coming. Sitting right next to short software as the other big theme this year was private credit, precisely because of its software exposure. The fear was that the PE deals of recent years (buy the software company for the cash flows, load it with debt, don't invest in the business to pay off the debt, sell it on) would end with those companies dying as they get replaced by AI. Private credit provided a lot of that debt. That was the thing expected to blow up.

So how unlikely does it seem that PE would come back for more software while the existing book was assumed to be in trouble? But as it usually goes. The things nobody expects are exactly the ones that happen.

Put everyone in the same trade, then hand them a 30% gap risk, and they don't wait around to see if it's real. They cover. That's the squeeze you saw today.

That’s positioning and short squeezes for you.

NLB

[NairoLB @NairoLB242]

@gregoryblotnick If anyone from London reading this and wants to get together, dm me!

[NairoLB @NairoLB242]

@gregoryblotnick Love the contrarian take. Do you think this also applies to Europe/UK?

[NairoLB @NairoLB242]

Seems like the market is finally gaining some clarity that AI will transform the economy, after hyperscaler earnings.

Should write more on this, so stay tuned.

NLB

[NairoLB @NairoLB242]

Interesting that the semicaps and other AI names have been selling into the close these last two days.

↓

[NairoLB @NairoLB242]

The market has moved on and is now pricing the FCF coming out of the hyperscalers a few years from now, after the big investments.

Beta is strong too.

↓

[NairoLB @NairoLB242]

speechless

[Michael Saylor @saylor]

Our Q2 Earnings Call, remixed. 🎶

[NairoLB @NairoLB242]

Look, I have no clue if he is right or wrong, but I must admit I did have to giggle at the simplicity of it: We are worse than others on benchmarks? Well, that means benchmarks are wrong.
Gigachad

[Jawwwn @jawwwn_]

.@ssankar says Palantir was able to make Nvidia's Nemotron Ultra model "better than frontier":

"I literally almost felt gaslit when, within 24 hours of getting Nemotron up with no post-training, this is vanilla Nemotron Ultra, it did better than frontier."

"If you just looked at the numbers, you would say, 'It's nowhere near the Frontier. That shouldn't even be possible.'"

"But of course, the benchmarks are wrong. I mean, the benchmarks are right for what the benchmark's measuring, but that's not my business. Those are not the tasks my customers had that they were trying to solve."

[Gregory Blotnick @gregoryblotnick]

"the way u take losses in life = the way u take losses in markets" (PART 1)

Why? Because: "THE WAY YOU DO ONE THING, IS THE WAY YOU DO EVERYTHING."

Thoughts:
(0:00) "MARKET WIZARDS" INTRO / emotional self discipline as all-encompassing
(2:00) dealing with loss - love, friendship, business
(3:00) broken heart - losing dignity and self-respect
(4:00) cutting a loss clean and keep moving
(5:00) obviating stop-losses
(6:00) taking L's like a professional
(7:30) where you fall on the risk spectrum
(9:00) knowledge of self: "never make the same mistake twice"

[Invest Like the Best @InvestLikeBest]

Claude is Walter Cronkite for the stock market.

@mjmauboussin talks about how a breakdown in diversity leads to bubbles and crashes. Gavin thinks Claude might be causing that right now.

”Everyone I know in the public equity investment business, whether retail or institutional, everything immediately, every piece of news gets fed into Claude—Claude, Claude code, sometimes a Claude agent.

Claude is probabilistic. There's probably not that much variation in the way it's interpreting this news.

People talk about the fragmentation of media and how it used to be like Walter Cronkite was the only voice of truth, and now we don't have that anymore.

It's like Claude is Walter Cronkite for the stock market, and everybody just believes whatever it says.”

[NairoLB @NairoLB242]

Google shipped 3 new models late July but still not the one everyone wanted.
No Gemini 3.5 Pro. Just cheaper, faster Flash models built to run under Search.

Not a bid for the coding crown. But for its own product range. What I called below

Full piece 👇
https://t.co/6tUr0rWbvB

[NairoLB @NairoLB242]

And one that isn't about markets: the habits behind sustained excellence, and which of them actually held up when I lived through them:
https://t.co/7YfRGnSUiW

[NairoLB @NairoLB242]

Going to be active on X again.

Substack stays my main platform, that's where the full research lives. This account is for shorter comments, closer to Substack notes.

To get you up to speed, everything on there so far:

https://t.co/u01qkHUodX

[NairoLB @NairoLB242]

Outside the series, three on what the tape is doing:
Software ripping while semis sold off:
https://t.co/57UKjfANHd
Earnings season and AI Returns:
https://t.co/GfmoJCGlu1
Google down 5% on a Gemini delay, which I don't think it deserved:
https://t.co/AVJPc0M72k