To The Moon

Investing is about risk and reward.

And I think the current state of the AI buildout has more risk built in than many investors want to think.

Those risks came to light over the weekend.

In essence, Dario’s argument is that AI development should be “paced” in three ways:

1. Embedded evaluators to verify the safety of models.

   1. This is similar to how big banks are regulated today.

2. AI labs coordinate to create safety frameworks and limit “unchecked AI progress”.

3. Coordination globally.

I have a lot of thoughts, and I don’t think there are 100% certain answers to the safety, risk, or future development of AI.

But I do think the argument and the fact that all of the major lab CEOs seemed to agree bring forward some major risks for investors.

And it seems to be pulling the rug out from under some of the biggest trades in the industry. 🤔

I’m going to try to go through the big ones, but I’m sure I’ll miss some.

Safety First

• Argument

  • AI safety is a concern, and frontier development should be slowed.

• Investment Risk

  • Training accounts for ~2/3 of AI workloads, and a reduction in training could lead to less demand for chips, data centers, energy, etc. If growth slows, valuations compress.

  • If safety is that big a deal, everything is at risk. Banks. SaaS companies. Healthcare. The market could crash if a major attack happens.

    • This could be a COVID-level event…

  • I have to ask: Why was this brought up now, and why did all of the CEOs seem to agree that AI safety is a huge risk? What did they see?

Pulling Up the Ladder

• Argument

  • Frontier labs just want to regulate AI in order to pull the ladder up from upstart competitors and open-source models.

• Investment Risk

  • Regulatory capture would lead to fewer competitors and potentially less innovation and growth.

  • If the AI labs win a dominant position, they’ll squeeze suppliers like $NVIDIA(NVDA)$ ( ▼ 3.36% ), AMD, etc.

The IPO Pop

• Argument

  • Frontier labs are trying to IPO so they want to reduce model development costs in order to become more profitable.

• Investment Risk

  • See above. Any reduction in model training would be a net negative for infrastructure demand.

The China Problem

• Argument

  • If the U.S. labs slow down, China will catch up/pass U.S. labs.

• Investment Risk

  • This is an acknowledgment that moats in AI are weak and can only be measured in months.

  • If the entire thesis is predicated on running faster and faster to stay ahead, eventually capital will run out, and the bubble pops.

The uncomfortable truth is that huge swaths of the market are buoyed by the AI trade.

Not only are revenue and (sometimes) profits up, valuations have exploded as well.

With higher valuations come higher expectations, and if those expectations aren’t met, valuations can revert back to lower levels very quickly.

$Intel(INTC)$ ( ▼ 5.59% ) has done little to prove it deserves its higher valuation.

$Advanced Micro Devices(AMD)$ ( ▼ 4.4% ) has performed well, but its price-to-sales multiple is up 869% in the past decade, which would have beaten the market!!

And what if demand doesn’t grow as fast as supply in a commodity market like memory, where there’s new factories coming online from every player by the end of the decade? Could $Micron Technology(MU)$ ( ▼ 5.25% ) see margins and multiples compress?

I haven’t chased the AI trade because I can’t tell what’s sustainable and what isn’t.

Anything unsustainable runs the risk of falling and potentially even going to zero. And if there’s a coordinated effort to slow AI development and spending — maybe even with the help of the government — that could lead to lower growth and lower expectations.

Today’s move is a reminder of how quickly the market’s sentiment can change, and it was an essay that caused the damage.

What happens if it’s something worse?

When markets keep you watching, knowing when to switch off matters too.

A strong U.S. jobs report has put rates back in focus, with this week’s CPI data set to be another key market mover — keeping overnight trading firmly on investors’ radar. But investing is a long game, and you don’t have to watch every move. The new Night Trading Dollar Eye Mask is soft, lightweight and comfortable, made for quick breaks between market moves, power naps or long-haul travel — so you can switch off, recharge and come back ready for what’s next.

Redeem the new Night Trading Dollar Eye Mask in Tiger Coin Mall. Keep an eye on the market — and give your eyes a break.

https://laohu8.com/J/redeemGift?goodID=100557&type=delivery