00:00Lisa, let's start with one before we get to the other. Chips, meaningfully overbought.
00:05What gives you that idea right now, Lisa? Oh, my goodness. I mean, all anyone needs to do
00:11is look at some of these charts that have gone completely hyperbolic. You know, whether you're
00:17talking about, you know, the Philly Sox over the last eight to 10 weeks, we're literally at certain
00:24points were up literally 100 percent on the year. We've obviously had some churn and some pullback
00:31recently. But, you know, whether you're looking at the participation in the Philly Sox, you're looking
00:39at semiconductor ETFs, you're looking at levered ETFs. It's all part of the same trade. And so much
00:48of it is related not just to swollen order books in across the industry, but quite frankly, folks
00:57extrapolating the sustainability of pricing and the pricing power, obviously, that companies like
01:06Micron have been able to embed in their earnings and their huge earnings surprises. Lisa, so many
01:12people are waiting before they get out to see a reduction in CapEx. They're waiting to go to the
01:17hyperscalers once they see a reduction in CapEx. You noted that early this week. Why not wait until
01:24then? Because I think that we're in the early innings of the beginning of that deceleration in
01:30the CapEx process now. Like everything, what we tell people is you can't market time these things to
01:37the day. You have to kind of sense changes in the second derivative. And we're sensing it. I mean,
01:43just look at some of the news flow that you're seeing from companies like Meta. It kind of signals
01:50to you that there are conversations about the rate and speed and return on investment on some of this
01:57CapEx and how can they pull forward their monetization strategies. Lisa, yesterday,
02:04we talked a lot about Meta now creating their own chip, Meta trying to be much more like a hyperscaler.
02:10Do you think that some of the reduction in CapEx that you're feeling out, not only is coming from
02:15the likes of Meta saying, maybe we'll sell some of our compute capacity, that's a better proposition
02:20than using it internally, but also because of pushback that you're seeing in financing markets,
02:25particularly in fixed income. Yeah, I'm not sure that the way these guys are behaving has to do with
02:34access to capital, just quite bluntly. I mean, I think that there's ample capital still available
02:41broadly to this trade and to most of the companies in it. I think this is much more
02:49about engineering and economics. Whenever you have a scenario where supply chains get bottlenecked and
02:57folks are extracting excess rents as some of the memory chip guys are, the engineers get to work. And
03:05that's what we're seeing. And we're seeing the AI data center tech stack, if you will, being re-engineered
03:14to include lower cost proprietary chips that many of the hyperscalers are now designing themselves and
03:23having manufactured white label. They are using a combination of open source as well as Frontier Lab
03:32models, all in the effort to compete on price and essentially to consume less energy per token.
03:43Do you see us entering a phase where AI is less inflationary as a result, where there is more
03:50discretion when it comes to how much hyperscalers in particular are willing to pay for chips for other
03:56particular inputs that actually could reduce how much it's really adding to the US inflationary backdrop?
04:03Absolutely. I mean, I think the front end of this story is NVIDIA, to be honest. You roll back the
04:11clock a
04:12year. Almost every hyperscaler felt that they had no choice but to buy NVIDIA GPUs exclusively and continue
04:23to queue up for the latest and greatest addition. I think now those GPUs are actually net losing total market
04:34share of data center AI designs and some of these proprietary ASICs are taking share. And so I think
04:43what we've seen with NVIDIA and the stock has kind of stalled out over the past six to nine months
04:50as a
04:50result, we're going to start to see with some of the other semis. So, Lisa, we're talking about 20%
04:56of the
04:57index. That's semis on the S&P 500 right now. Crazy. For some people, 100% of the sentiment, the
05:02enthusiasm for this market. Where does it leave the index outlook, Lisa, if you believe that that big
05:07chunk of this market is overbought? Yeah. So, what we have to remember is that the market is also
05:13concentrated on the other end, right? It's concentrated in some of the hyperscalers that have sold off.
05:19And so what's happened is that, you know, as we've had this churn, right, we've had this huge move and
05:27the change in market capitalization of the semiconductor industry from about 3% of the index to 18, as you
05:33point out. And the hyperscalers have actually lost share of the index and we've quote-unquote
05:42deconcentrated. And that's kept the overall index in a zip code. I think if this intra-tech trade
05:51continues to churn, right, we can absorb, you know, a 10%, 20% decline in semiconductors as we already have
05:59in the last couple of weeks and not have a lot of damage to the headline index. That's one of
06:05the good
06:05things, if you will, of having the hyperscalers be as big as they are as a group.
Comments