Lee says AI differs from past bubbles while innovation continues

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

1. 5:40

   Do you think that it exists? If not, why not? And if so, what are the dynamics going into it? I've witnessed several bubbles, you know, because I've been covering markets since the early 90s. And the two that people would describe as bona fide bubbles are the internet bubble and the housing bubble. I mean, those are genuine bubbles. and it's hard to compare the current cycle to either at this moment because you know the internet bubble was apparent because the the type of innovation that was taking place was simply cost savings the internet in into 1999 was not a revenue story it was a brick and mortar to e-tailer model. It was trying to basically say, this is a new way to do it, but you could save money. And the housing bubble, of course, brought leverage and capital into the hands of unqualified

2. 6:43

   holders. So it was basically a bubble where money was, debt was given to people, and then a lot of people borrowed a lot of money. This AI story so far is one of pretty remarkable innovation taking place, like otherworldly innovation. And so I think as long as that pace of innovation is happening, and then we're seeing it as productivity for businesses, because revenue per workers going up, there was no productivity metric in the internet we could point to at all at any time into 99. I think AI becomes a bubble when we start making up use cases that don't actually have end markets yet. As far as circular financing goes, and I know people are making claims that that's the reason it's a bubble. I think people do need to be students of history.

3. 7:37

   There are many cases where circular financing is not called circular financing. It's really ecosystem development. Some examples are, of course, the energy industry then would be notorious for circular financing, but that is the nature of resources. Another example is the early days of radio. RCA made radios. Then they had to actually create radio stations across the country. then they had to invest in programming across the country and imagine if someone said radio was a bubble because rca is investing in radio stations and in in radio programming they were creating an industry and that was not it would have been a huge mistake because rca was probably the biggest stock of its time until the 1929 crash which was not a radio bubble it was just basically a stock market bubble So I think, you know, it was more than 100x move. Cisco was 100x move. NVIDIA has been about a 10x, you know, from the start of ChatGPT. So to me, it's, of course, has all the makings eventually to create a bubble, but I think it's very early.

4. 8:54

   I mean, I would just say during the internet period, and I covered it from 93 to 99, well, I covered it all the way through the crash and recovery. No one even called it a bubble until after 2000. In terms of the circular financing, I think one of the concerns, even if we would call it ecosystem development, is that it is implying that demand might be stronger than it would have been if the demand were organic.

5. 9:24

   Like, if you didn't have these circular relationships, then you wouldn't see as large of a business. And therefore, you know, the argument is that this is sort of artificial growth that we're seeing. I think the same thing would be said about token maxing as an example. You know, businesses tell their employees, you have to use as many tokens as possible. And the reason that we want to do that is because we want to be part of this revolution and we feel very strongly about this.

6. 9:54

   And then suddenly they say, actually, this is too expensive. Let's pull back on the token spend. And as a result, it does seem at least the growth of the token use is beginning to slow. Whether that's directly a result of token maxing, I'm not sure. But it does seem in those cases that there are these dynamics that point to an ecosystem where the signals that we're getting aren't maybe the purest signals of organic growth that are completely dependable. Do you disagree with that or do you not worry about that?

7. 10:27

   I mean, I just want to point out as a practical matter, exponential growth industries don't have the same rate of change sequentially. Cisco, for instance, in 1990. 1997 had two quarters of 20% sequential growth, and it went down to five. And in 1997, Cisco stock fell 50%, and the data networking analysts said the growth story is over. And they said there's competition coming from internet protocol switching.

8. 10:57

   So Cisco fell from $9 to $5. A year later, it was at $18. So growth, an exponential S-curve industry doesn't grow at the same pace. I think token usage still has the Jevons paradox at work, you know, and it doesn't mean every day and every month and every quarter it grows exponentially. I covered mobile. There might have been six growth cycles in there and growth recessions because, hey, you roll out GSM, then growth slows.

9. 11:27

   Or if you roll out a new Nokia phone, there's an upgrade cycle and it slows. But everyone thought that was the end of the penetration cycle for mobile. And of course, it continued for 30 years. I think that token maxing and people thinking rate of change for tokens is going to explain the peak of the cycle, they haven't covered prior growth cycles. And again, I'm just emphasizing I've covered many in my lifetime. I'd say the real key is is ai no longer providing remarkable innovation and if it doesn't that's really the end of the cycle when you look at anthropics growth as an example does that make you feel more bullish