What's Next For Markets

David O'Connor discusses NVIDIA leadership transition, AI CapEx concerns, and semiconductor outlook

29 min Listen to the original
NVIDIASemiconductorsAI Capital ExpenditureValuationMarket SentimentAgentic AI

Summary

David O'Connor argues that NVIDIA's leadership transition will not disrupt its multi-generational AI roadmap, which remains sold out through 2028. He highlights a disconnect between NVIDIA's 10x P/E multiple and its 50% CAGR earnings growth, identifying a $150 stock price floor as a long-term buying opportunity. O'Connor expresses concern over AI capital expenditure projections reaching $2.5 trillion by 2030, questioning whether current revenue streams can justify such spend. He recommends Broadcom, AMD, and ARM while advising avoidance of memory and consumer semiconductor stocks due to supply normalization pressures.

Key takeaways

  1. 1
    These guys, these hyperscalers are in the space race basically to get to AG. AGI.

    Unidentified speaker (speaker_2) · 14:11

  2. 2
    hyperscalers, they only get probably half of the supply among them. So the big kind of four guys, maybe fifty, fifty-five percent of the supply. And then the rest of the supply kind of is allocated to things like neo cloud, guys with new business models, who again have different sets of customers. Also it's going to sovereigns, also going to enterprise as well.

    Unidentified speaker (speaker_2) · 15:10

Episode sections

  1. 1:23 – 3:20

    NVIDIA Leadership Transition and Roadmap Certainty

    Speaker 2 argues that NVIDIA's CEO departure does not change the company's multi-generational AI roadmap, which is already sold out through 2028. The market is expected to quickly appoint an internal successor, likely the head of the data center business, to maintain execution.

  2. 3:20 – 9:02

    Market Sentiment and Election Impact

    Investor conviction in semiconducters has been weak since late summer due to macro uncertainty and election anxieties. Speaker 2 notes that while anti-data center sentiment exists, it has not yet impacted the fundamental demand-supply imbalance for AI chips.

  3. 9:02 – 10:10

    NVIDIA Valuation and Stock Floor

    Speaker 2 points out that NVIDIA is trading at a 20-year trough P/E multiple of 10x despite 50% CAGR earnings growth. He identifies a stock price floor of $150 as a buying opportunity for long-term investors.

  4. 10:10 – 17:55

    AI Capital Expenditure and ROI Concerns

    Speaker 2 highlights that AI CapEx is projected to reach $1.3-1.4 trillion in 2027 and normalize to $2.5 trillion by 2030. He questions whether current revenue streams from companies like Anthropic and OpenAI can justify this spend, identifying monetization as the key uncertainty.

  5. 17:55 – 24:45

    Bull and Bear Cases for Semiconductors

    The bear case parallels the 2000 internet bubble, where infrastructure build-out outpaced monetization. The bull case relies on accelerated enterprise adoption of AI and potential earlier-than-expected robotics integration driving further semiconductor demand.

  6. 24:45 – 29:25

    Recommended and Avoided Sectors

    Speaker 2 recommends Broadcom, AMD, and ARM for their positioning in inference and agentic AI. He advises avoiding memory and consumer semiconductor stocks due to supply normalization pressures and demand destruction from high memory prices.

Assets discussed

AssetKindStanceContext
NVIDIANVDAequitymixedThis week, we're happy to be joined by David O'Connor, senior research analyst covering semiconductors and capital equipment here at Piper Sandler to discuss NVIDIA semiconductors.

“Stance” is how the speakers talked about the asset, not a recommendation.

Claims worth verifying

NVIDIA's roadmap is sold out for 2027 and 2028.

Unidentified speaker (speaker_2) ·2:03

How to verify: Check NVIDIA's quarterly earnings call transcripts or investor presentations for forward-looking capacity guidance and supply commitments. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

NVIDIA is trading at a 20-year trough multiple of 10 times.

Unidentified speaker (speaker_2) ·9:21

How to verify: Calculate NVIDIA's current P/E ratio using its stock price and trailing twelve-month earnings, then compare to historical P/E data from financial databases like Bloomberg or FactSet. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

NVIDIA earnings are growing at a 50% CAGR for the next three years.

Unidentified speaker (speaker_2) ·9:21

How to verify: Review NVIDIA's forward guidance provided in recent earnings reports or analyst consensus estimates for FY2027-2026. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

AI CapEx will be up 70% in 2027 and 50% in 2028.

Unidentified speaker (speaker_2) ·5:17

How to verify: Analyze public CapEx disclosures from major hyperscalers (Microsoft, Meta, Google, Amazon) in their 10-K/10-Q filings and compare year-over-year changes. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

Anthropic and OpenAI have a combined run rate of $100 billion.

Unidentified speaker (speaker_2) ·11:49

How to verify: Estimate revenue based on public reports of user growth, subscription prices, and enterprise contracts, or wait for private company disclosures if available. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Consumer PC demand is down 10-15% this year.

Unidentified speaker (speaker_2) ·27:44

How to verify: Check shipment data from market research firms like IDC, Gartner, or Canalys for global PC shipment volumes. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Over 30% of the tech sector had negative earnings in the late 90s/early 2000s, compared to 3% today.

Unidentified speaker (speaker_3) ·20:38

How to verify: Calculate the percentage of S&P 500 Technology sector companies with negative net income for the specified periods using SEC EDGAR filings. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

NVIDIA allocates 50-55% of its supply to the top four hyperscalers.

Unidentified speaker (speaker_2) ·15:10

How to verify: Review NVIDIA's customer concentration disclosures in SEC Form 10-K or 10-Q filings. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

Smartphone shipments have grown from 100 million to 1.2 billion units.

Unidentified speaker (speaker_2) ·23:26

How to verify: Compare historical smartphone shipment data from IDC or Gartner for the early smartphone era versus current annual figures. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Consumer semiconductor demand will remain under pressure through 2027 due to high memory prices.

Unidentified speaker (speaker_2) ·28:04

How to verify: Monitor memory price trends from industry trackers like TrendForce and correlate with consumer electronics shipment forecasts. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Do your own research

  • Verify NVIDIA's supply allocation percentages by reviewing recent earnings call transcripts or investor presentations.
  • Check current P/E ratios for NVIDIA, Broadcom, AMD, and ARM against historical 20-year averages using financial data providers like Bloomberg or Yahoo Finance.
  • Assess AI CapEx forecasts from major hyperscalers (Microsoft, Amazon, Google, Meta) in their latest 10-K or 10-Q filings.
  • Research Meta's 'Muse' application and its current user adoption metrics to evaluate agentic AI ROI potential.
  • Review semiconductor inventory levels and supply chain reports from Taiwan-based manufacturers like TSMC to validate demand-supply imbalance claims.
  • Analyze historical parallels between the 2000 internet bubble and current AI infrastructure build-out by comparing CapEx-to-revenue ratios.
  • Monitor consumer semiconductor and memory sector pricing trends to assess demand destruction risks from high prices.
  • Identify potential conflicts of interest for speaker David O'Connor by checking his fund's holdings in recommended stocks like Broadcom, AMD, or ARM.
Full transcriptexpand
0:00
Michael Kantrowitz: What's next for markets?
0:05
Unidentified speaker (UNKNOWN): I'm
0:05
Unidentified speaker (speaker_1): Michael Kantrowitz, Chief Investment Strategist at Piper Sandler, where I lead the portfolio strategy team ranked number one on Wall Street in the All-America Institutional Investor Research Survey.
0:16
Michael Kantrowitz: This week, we're happy to be joined by David O'Connor, senior research analyst covering semiconductors and capital equipment here at Piper Sandler to discuss NVIDIA semiconductors.
0:27
Michael Kantrowitz: and some of the tough questions that have been coming up around the AI spending cycle. So David, welcome to the show and a pretty exciting time the semi space recently.
0:40
Unidentified speaker (speaker_2): Yeah, there's never a dull moment in the semis, that's for sure. Particularly given the all the drama the last the last couple of days, who the guest, who the guest we'd be facing, NVIDIA be facing, no C CEO this time last week. So yeah, so a lot of a lot of uncertainty there, a lot of concern from investors, but at the same time it's yeah, it's it's it's probably
1:00
Unidentified speaker (speaker_2): an opportunity as we highlighted this morning as well coming out of this.
1:03
Michael Kantrowitz: So we're sitting here November 12th. There's been so much activity and interest in the volatility around the tech space in general, which these days is led by semis. But as you mentioned, some news broke today. Maybe before we get jump into the big picture, kind of addressing the current elephant in the room, does the
1:23
Michael Kantrowitz: news around the NVIDIA CEO no longer being there matter to you? Is that a big deal?
1:31
Unidentified speaker (speaker_2): Yeah, the all has changed, but nothing has changed in one respect. It's it was such a unique company and his position there, because he was the founder in since the very for the last thirty years, very unique structure. So he was it was a one man band internally. He's involved in kind of everything. But at the same time,
1:51
Unidentified speaker (speaker_2): a lot of the big decisions in AI have already been made. When for someone like NVIDIA who makes money selling GPU chips, but all is based off a roadmap which is multi multi-generational. generational.
2:03
Unidentified speaker (speaker_2): And if we look down the tracks, yeah, the all the big decisions already have been made for 2027, 2028. In fact, yeah, 2027 is sold out, same in 2028. One new CO coming in isn't going to change that. He's just just more focused on execution, but yes, steering the ship in the same direction.
2:23
Unidentified speaker (speaker_2): So it doesn't change massively. Obviously there's uh a an execution, there's a succession risk which carries a bit of execution risk. So that's what uh investors are grappling with. But I think the market and NVIDIA will move on quickly. We saw the resignation this morning. And our guess is the that they will
2:43
Unidentified speaker (speaker_2): appoint someone internally who would be a safe pair of hands here to execute on that roadmap who knows the customers and yeah, we think it's the head of the data center business, is the guy who gets the nod there ultimately, but still early days in the process, right?
2:57
Michael Kantrowitz: Interesting. Okay. We addressed the elephant in the room. Now to talk a little bit about the broader space and not just about NVIDIA. Tech has been under pressure. And that has mainly been led by some uncertainty within AI and semiconductors specifically. What is your view on the
3:17
Michael Kantrowitz: recent market action and the outlook going forward?
3:20
Unidentified speaker (speaker_2): Yeah, no, you fit the nail on the head there. Yeah.
3:23
Unidentified speaker (UNKNOWN): The big
3:23
Unidentified speaker (speaker_3): question. That's the big
3:24
Unidentified speaker (speaker_2): question. I thought this was meant to be easy questions.
3:27
Unidentified speaker (UNKNOWN): The
3:27
Michael Kantrowitz: softballs come
3:28
Unidentified speaker (speaker_2): later. They come in the post. Yeah, I would say since late summer, there hasn't been a great amount of conviction among investors. They've been they've all been long semis, right? And they've been long kind of networking trade, long the memory trade. And we had a bit of a sell-off in the summer, but coming
3:48
Unidentified speaker (speaker_2): out of the summer we had a bit of a rally, but still there was no broad conviction out there. Investors are worried on the volatility. They were worried on the macro and the thinking was at the time once we get to the elections and get that kind of digested, whatever comes out of that, then that sets it up for a kind of a rally
4:08
Unidentified speaker (speaker_2): into year end. That was the kind of ideal picture. We also had the kind of the anthropic IPO and stuff and the news around that. So there was a lot of moving pieces, but investors more focused on top down versus earlier in the year, I would say they were focused more on the bottom bottom-up up side of things. And I think now we're post the election with the
4:28
Unidentified speaker (speaker_2): anthropic IOPO coming next week. That'll be helpful. But yeah, we're still, I would say, lacking a kind of catalyst of what's going to drive the AI trade into kind of year-end. It's tricky. It's tricky, right? No easy answer to that question.
4:44
Michael Kantrowitz: Does the election results and the anti-data center sentiment that bubbled? up in the recent election results affect the earnings picture for these semiconductor stocks?
4:57
Unidentified speaker (speaker_2): Yeah, effectively it doesn't. It's when we've been through earnings and the message from earnings, which is the effectively a kind of a snapshot of the fundamentals. Fundamentally, AI demand is remains very strong. Most AI chip suppliers, they're all sold out, as I mentioned earlier. for 2027 and even
5:17
Unidentified speaker (speaker_2): 2028, in fact. So demand, the picture still is demand is ahead of supply. We're supply constrained and compute. And if we look to next year, 2027, AI Capix next year is going to be up another 70%. Next year, and looking to 28, it's and some of the supply chain checks out of Taiwan. That's going to be up another probably
5:36
Unidentified speaker (speaker_2): 50%. fifty percent. So you know the that uh picture is unchanged, regardless of the elections, regardless of the sentiment. Those two things are add to the noise, but they aren't really going to change the picture over the twenty twenty seven, twenty twenty eight. Now, maybe let's see when the newly elected candidate, newly elected politicians take their seat,
5:56
Unidentified speaker (speaker_2): maybe they the policies will change on data centers, etc., and that kind of that could affect maybe. 2029, 2030. Because it takes three years to build these data centers. And also we're talking about we're talking about large hyperscalers who have global footprints who can pivot to adding capacity
6:16
Unidentified speaker (speaker_2): from one state to the next, from one country to the next. So yeah, it's one one the the kind of rising anti data center sentiment. We don't really see that kind of
6:27
Unidentified speaker (speaker_2): impacting the uh the AI trades, but it's unhelpful, but no fundamental impact uh would say at this point. So
6:34
Unidentified speaker (speaker_3): is it fair to say that you are continue to be constructive on the space? We've seen a 20% pullback in semis over the last just quarter and date already. What are the opportunities that you see and where would you be focusing if you were an investor?
6:49
Unidentified speaker (speaker_2): Yeah, absolutely would still be constructive. A lot of I'm guided a lot on the fundamentals here. And as the picture I painted earlier is a very strong outlook for 2027 and 2028. And despite all kind of what we talked about, the elections, the data center, anti data center sentiment, none of that is
7:09
Unidentified speaker (speaker_2): Is going to change. So I think given the pullback we had, obviously investors are concerned. We talked about the volatility, but it does present the pullback does present opportunity here when we look at the top line growth, the earnings growth, when we look at the multiple. So yeah, so we'd be we'd still be quite constructive here, looking out over the next
7:29
Unidentified speaker (speaker_2): 12 months. Probably can't exactly put Emily a finger on what that catalyst is going to be to drive the next leg up in the trade, but certainly.
7:39
Unidentified speaker (speaker_2): on a kind of l mid to longer term basis, we'd be quite constructive here on on the pullback and some of these valuations.
7:46
Unidentified speaker (speaker_3): Now, next week obviously is going to be a big earnings week report with NVIDIA coming out. What is it? Is there anything specifically that you are looking for from their from their release to give a better sense of where the tides are likely to be going for the space for the next couple of months? months.
8:03
Unidentified speaker (speaker_2): Yeah. The NVIDIA, in fact, is already guided for next year, revenues up 70%, as has kind of Broadcom as well, you could say as well, up around that 70% mark next year. So a lot of the bigger guys have already given a steer for next year. And then that kind of matches the data points we get out of Taiwan. So that probably kind of paints the picture of a very
8:23
Unidentified speaker (speaker_2): strong kind of year in 2027. Maybe if it's with all the certainty, probably there's there's a lot known, is my point for next year. So I don't expect from NVIDIA next week for us to learn too much more about kind of 2027. Their numbers are going to be strong. As we talked about earlier, the fundamentals
8:43
Unidentified speaker (speaker_2): are remained very strong. That is unchanged. changed. For NVIDIA, the key thing is the messaging around the the new the the new CO, CEO, the CO CEO transition. and a kind of to reassure the industry that it's all's changed but nothing's changed is a key thing for NVIDIA. So not really looking to learn. I don't expect us to learn too much
9:02
Unidentified speaker (speaker_2): incrementally more that we don't already know. But certainly we need to hear a very reassuring message here from the next week. Maybe just one thing just on NVIDIA. We talked about for investors a concern has been around like a catalyst for the stock the the D-rating as well. And when we look at the valuation of NVIDIA, NVIDIA is trading at a 20-year trough
9:21
Unidentified speaker (speaker_2): multiple here of 10 times. When we look at the earnings growth of the company's growing earnings at a 50% cagger for the next three years, we see a kind of disconnect there and we see a a floor in the stock of Of $150. It briefly touched there this morning, but certainly for longer longer-term term investors, we see that. as a buying opportunity here.
9:41
Unidentified speaker (UNKNOWN): The
9:41
Michael Kantrowitz: semiconductor space is a lot bigger than just data centers and AI, even though that's what we associate it with. How does that and the uncertainty around the future profitability of AI affect the long-term earnings growth? Is this a n no matter where that kind of the chips fall on the ROI
10:01
Michael Kantrowitz: for AI? is the semi-space still going to be an area that puts up these incredible growth numbers going forward?
10:10
Unidentified speaker (speaker_2): Yeah. The the the semispace really for the last three years has essentially all the growth, so 100% has been driven by AI. And that is likely going to continue for the next three years. But the spending whether you like it or not, the semis semi space is totally dependent on the kind of growth of AI and the success of AI. And for
10:30
Unidentified speaker (speaker_2): a big question, Mark, at the moment, for a lot of investors on AI is to your point on the ROI here. So the return on all this investment that we're doing. And that's less clear, I would say, at the moment. And that's where we we need better answers, I would say, going forward. When you look at the CapEx, for instance, this year $800 billion, we've done a double to the
10:50
Unidentified speaker (speaker_2): eight hundred $800 billion. level the last three years. And next year, as we talked about earlier, it's up another 70%. So we're going ahead like a $1.4 billion number, $1.3, something like that next year. And then, as I say, the 50% growth is the metric the year after, based on supply chain checks. And then you normalize by 2030,
11:10
Unidentified speaker (speaker_2): you easily get to $2.5 two and a half trillion dollars on a kind of a run. a rate for capex. Now they're big numbers, but it also means you need to justify that through the ROIs. Where is that in where is the revenue growth coming from from that? One way to look at it is as well, typically in the industry for every dollar you invest in capex,
11:30
Unidentified speaker (speaker_2): you need to generate a dollar in revenue somewhere in the in in the ecosystem. So at $2.5 trillion by 2030, that means you need to create an AI. industry generating a similar level of kind of revenues. When we look at Anthropic and OpenAI today, they're probably run rate going forward, maybe end of the year closer to 100 billion. That's
11:49
Unidentified speaker (speaker_2): 200 billion combined. We're a long way off hitting what makes sense from an ROI perspective. So we need better answers there. And that's where kind of I think going forward into next year, that could be the catalyst for the next. leg of growth for a lot of the AI trade. If that's a Gentic, agentic AI is new. It's it's a
12:09
Unidentified speaker (speaker_2): lot of the hyperscalers and AI labs are launching their own version of a Gentic AI with meta we have Muse, certainly gathering traction. What's useful, I'm not sure if you guys have used it, but very useful. That is consumer basically has to be a big part of that to answer that kind of ROI question. But there's no clear answers to that at the
12:29
Unidentified speaker (speaker_2): moment. But yeah, that's as we go into next year, that's where we're looking for better answers.
12:34
Michael Kantrowitz: Can the consumer support these levels of capex?
12:38
Unidentified speaker (speaker_2): We not just consumer. We need it's not consumer spending a couple of hundred million, consumer spending 30 bucks a month. That's probably not going to get us there. We need consumer, though, to be a contributor. We need entity. enterprise as well to be a contributor. So far, the big success has been things like Claude and Claude Code and so forth. That's been helpful,
12:58
Unidentified speaker (speaker_2): but that's not going to get us there, right? It's an early adopter, but it's we need more than that. So yeah, the consumer is important to justify the level of the capix. It always is in these types of super cycles, but we need more. Probably enterprise is where the biggest kind of unknown is at the moment. Lots of companies, I go to
13:18
Unidentified speaker (speaker_2): conferences, lots of companies I talk to, they're all working on AI. Everyone is doing something out there in the real world. Here at Piper, we're working as well on stuff on AI. So very useful. But again, we're still in the early innings of that. We need to see a lot broader employment in the real world or deployment of a
13:38
Unidentified speaker (speaker_2): of AI. to drive kind of those incremental revenues. And that's that's yeah, it's important. All these end markets contribute to justify that ROI. Because otherwise, it doesn't make sense from a business perspective.
13:51
Michael Kantrowitz: Is the demand you're seeing and the the semiconductors being swallowed up by all the hyperscalers and AI plays? How does that affect kind of the future build-out within the space? Because semiconductors are a broader use case than just data centers and compute for AI.
14:11
Unidentified speaker (speaker_2): Yeah. Probably any hyperscaler would take a hundred percent of the supply of GPU chips if they could. But not that they need it, but just to lock everyone else out of the market. These guys, these hyperscalers are in the space race basically to get to AG. AGI. And the two key ingredients you need is you
14:31
Unidentified speaker (speaker_2): need power and you need GPU chips in data centers, basically. And that's the kind of the that's the kind of the raw ingredients here to be successful in AIs. They would take double, if they could, of the kind of semiconductor supply. Now, from a semiconductor supplier perspective, what's not interesting, just supply one guy. So
14:51
Unidentified speaker (speaker_2): you want to enable many actors out there and you want to enable competition as well amongst them. So guys like NVIDIA who control the supply chain thus far have done an excellent job in allocating supply. So everyone gets their fair share, you could say. So all the supply, although it could go to
15:10
Unidentified speaker (speaker_2): hyperscalers, they only get probably half of the supply among them. So the big kind of four guys, maybe fifty, fifty-five percent of the supply. And then the rest of the supply kind of is allocated to things like neo cloud, guys with new business models, who again have different sets of customers. Also it's going to sovereigns, also
15:30
Unidentified speaker (speaker_2): going to enterprise as well. Although enterprise Price is a much smaller portion, they are getting their kind of a fair share so far. So yeah, no. it's not just the hyperscalers driving the bus here, because but the demand is significantly stronger than the ahead of supply this year and next year as well.
15:48
Unidentified speaker (speaker_3): As we think about NVIDIA, and obviously there's gonna be a new uh a new NVIDIA with without Jensen at the helm, but just wondering, we were talking about this as a team, whether as a company, NVIDIA has become too big to fail versus too big to bail as the way.
16:03
Unidentified speaker (speaker_3): the sort of its role in the ecosystem of the economy. And just I know it's a it's a big question, but just wondering if you have thoughts about sort of like a downside detection element in terms of investing in NVIDIA as far as there being so many supports for it. Do you have any thoughts about limiting the downside, limited downside risk there, just given how important it is? Yeah.
16:23
Unidentified speaker (speaker_2): Make no bones about it. NVIDIA is driving the roadmap for. AI. It's effectively created the whole AI industry or was an enabler, I guess is the right expression, an enabler for the whole AI industry based on its GPU chips. And it's been driving the roadmap for the last six,
16:43
Unidentified speaker (speaker_2): seven years for AI. And it's built up the uh its supply chain to be able to support now, as as we talked about, a kind of eight hundred billion dollar infrastructure build out. So yes, I would say it's too big to fail, but I think the risk of its fail failing is very limited
17:03
Unidentified speaker (speaker_2): here. The uh we have a had a we were having a CO transition now, but at the same time, like we talked about earlier, the all the big decisions for the next three years. have already been made. The die is cast on all the products, all the architectures. Next year sold out, 28
17:23
Unidentified speaker (speaker_2): is effectively sold out. So yeah, we don't see a risk of failure here. Maybe there's a higher risk premium attached to NVIDIA. But I think speaking of of failure Fader is just way too way too pessimistic, I think.
17:36
Michael Kantrowitz: Interesting. So I won't be so mean as to ask, are we in an AI bubble and put you on the spot there? But I will ask what in your eyes, speaking with investors and your knowledge of the space, what is the bull case for semis and what would be the bear case for semis?
17:55
Unidentified speaker (speaker_2): Are we in an AI bubble? Oh
17:57
Michael Kantrowitz: if you want to answer that one.
17:59
Unidentified speaker (speaker_2): We're
17:59
Unidentified speaker (UNKNOWN): in
17:59
Unidentified speaker (speaker_2): uh the happy to answer all questions here, Stephen. The we're in a super cycle. Now if you wanna some people call it a bubble, some will call it uh a super cycle. I prefer to call it a super cycle because a bubble has quite nasty connotations attached to it. But look, the essentially we're in a super cycle where we have a new technology and we're building out the
18:19
Unidentified speaker (speaker_2): infrastructure for it. Now this is the exact same as we saw through the smartphone super supercycle cycle or smartphone bubble, and as we saw through the internet. super cycle and the internet bubble way back in 2000, when at that point I was working in the industry as a chip designer. So, yeah, a very first hand experience of a lot of these super cycles or
18:39
Unidentified speaker (speaker_2): bubbles as you call them. And typically they they're quite similar, they have similar characteristics in that there's a kind of a build build-out out phase, and then there's a kind of adoption phase, and then there's a monetization phase. Right now, within AI, we're in the infrastructure build-out phase, and the as I say, we've gone double
18:59
Unidentified speaker (speaker_2): double the last couple of years. Now, as we move forwards, the growth, the rate of growth will slow. It's only natural, right? Particularly you look at the kind of law of big numbers, that's going to slow. But we really need to get the monetization phase. And there's, as we talked about earlier. there's no great answers yet on the monetization stage. But there
19:19
Unidentified speaker (speaker_2): has to be good answers to keep to get to extend the kind of cycle here. I do believe we're in a multi-year kind of AI is very transformative to lots and lots of businesses out there. And those that embrace it best or quickest will be the ones who sir who survive. But we do need to see that monetization because also
19:38
Unidentified speaker (speaker_2): otherwise we'll end up in a very similar situation as we had in the internet build out where we had a couple of years of infrastructure build out and there was no monetization at the end of that. And then when the kind of there was no cash flow being generated, the bills weren't being paid, that's when it all grown to a halt basically. So there is similarities in and all these kind of
19:58
Unidentified speaker (speaker_2): super cycles, all these bubbles. Where we are right now, we're probably in the as I say early to middle cycle of the infrastructure part of this build out. So it's it's it's it's still early days, I would say. But yeah, better answers on monetization is what we're looking for going forward to give investors that kind of conviction that it's it's not yet all over.
20:18
Unidentified speaker (speaker_3): In terms of what you were just talking about, in terms of comparing today to the early two 2000s. thousands, there's we get that question all the time. C investors that are concerned about any type of repeat of that. And the backdrop to your point is so dramatically different today. One of the charts we were just pulling up season night this morning, looking at the percentage of the tech sector that has negative earnings,
20:38
Unidentified speaker (speaker_3): which back in the late 90s, early 2000s was over 30% of the sector, where that's about 3% today. So just the stocks that are driving performance of tech over the last several years have all been making money. They've been profitable for the most part. It's just a totally much more established, stable business model that than what we had back then. And so while there
20:58
Unidentified speaker (speaker_3): may be stocks that are not as not quite as profitable, the general trends of the industry have been is a much more solid footing today than it was back then. And so that's something that people like to talk about all the time with the comparison. And when I was asking earlier about being NVIDIA being too big. to fail or too big to bail, and not that I thought there was going to be any imminent
21:18
Unidentified speaker (speaker_3): threat to that for NVIDIA, but just the concept of where do we go from here and can these stocks continue at the pace that they've been going. I
21:26
Unidentified speaker (speaker_2): think the question was it the the bull case for semis and the bear case for semis for here. I think the the bear case and to your point Emily, the bear case here is we draw a parallel with the internet build out in the late nine uh the late nineteen nineties. Sorry, in in two thousand, where we did an infrastructure build out and there was no monetization.
21:46
Unidentified speaker (speaker_2): I think that's the bear case here. And our monetization takes a lot longer to come through and we have to slow the build out and before maybe monetization doesn't happen in the next one or two years, maybe it takes really another three, four or five years before that comes through. I think that's the uh I think that's the bear case here. And that's why it's so important to get better answers to ROI
22:05
Unidentified speaker (speaker_2): here, to monetization. So to bat away that kind of bear case or to reassure ourselves that we're not in the same situation as we were back in 2000. The bull case here is I would say the AI build out continues and probably accelerates again and driven by faster adoption
22:25
Unidentified speaker (speaker_2): in the enterprise. of AI, that all those companies out there who have been working on AI models internally in their companies start to deploy AI models and we get significant spike or demand for tokens basically, and that would require a lot more semiconductors, Conductors,
22:45
Unidentified speaker (speaker_2): a lot more data centers, a lot more chips, and just semis in general. I think and then you have probably maybe something like robotics coming a bit sooner as well. Robotics is a big application for AI down the line. Now it's not meant to come until kind of volumes 2029 timeframe, but if that kind of comes a bit sooner,
23:06
Unidentified speaker (speaker_2): that would be a new application as well and drive a lot of demand for semiconductors and for kind of tokens essentially. So yeah, so the again it's stripping it back. It would be a kind of reacceleration of AI from here, which has happened before. If we go back to the smartphone cycle, there were stages in the smartphone cycle, but that where
23:26
Unidentified speaker (speaker_2): we thought when we were shipping like a hundred million smartphones, that was the that was a going that was it. There was a ceiling to number smartphones, and we went on and today we're shipping like 1.2 billion smartphones. So there was been on previous cycles. there has been times during the cycle where there's you get a collision between macro concerns and also
23:46
Unidentified speaker (speaker_2): maybe a kind of a not a pause in the kind of fundamentals, but maybe a kind of a yeah, maybe a pause before the next kind of technology wave comes. We have here be been here before. So I think going from here, yeah, the bull case would be a kind of a reacceleration driven by broader adoption of a AI across across the enterprise.
24:05
Michael Kantrowitz: Dave, a quick question. We've hammered you on the semi-space, and we have talked about NVIDIA because they're very much in the news and they are the biggest. Though we on the strategy team always highlight that sectors aren't always representative of all of the stocks underneath them, and every
24:25
Michael Kantrowitz: industry tends to be a diverse group. So closing out, can you give us some companies that you follow in track that you think are good executors and semiconductor stocks to be watching going forward? And if you have any ones to maybe avoid?
24:45
Unidentified speaker (speaker_2): Yeah, sure. The from us taking it down to the stock level, the way I think about stocks and certainly. semiconductors, I like to keep it simple. Either the earnings drives the stock or it's on the multiple kind of expansion drives the stocks or contraction. Works both ways. On the earning side of things, as we talked about over the last
25:05
Unidentified speaker (speaker_2): 30 minutes, there's a lot of no knowns as we go into 27 and 28. So we're probably not going to get as much surprise on the earnings earning side in going forward from here over the next 12 to 24 months. So to drive these stocks, it really has to come from the multiple side of things. And typically in semiconductors, there's probably three things drive the multiple.
25:25
Unidentified speaker (speaker_2): It's new technology, new application, or market share gains. So whatever stock you're looking at, it has to drive the multiple and drive the stock. It has to fit the thesis, it has to fit into kind of one of those three stocks. segments. And when I look at kind of the the coverage right now, I would say the uh the guys who I would see is
25:45
Unidentified speaker (speaker_2): very well positioned as you call good stocks going forward into 2027, Broadcom is one I'd be pushing here. Again, inference is just taking off for a lot of these for a lot of these the hyperscalers. It's extremely well positioned on the inference side of the of the equation also i'll be looking at
26:05
Unidentified speaker (speaker_2): amd agentic ai we spoke about it earlier agenticai is going to be a massive theme again in 2027 among our coverage amd is very well positioned there i also would arm i think has some catalysts again on the kind of soft bank gpu as a catalyst there it also ticks
26:24
Unidentified speaker (speaker_2): the agentic A ai box So very well positioned. We spoke about NVIDIA earlier is trading a trough multiple. It's the kind of the elephant in the room, so to speak, but the fastest runner in AI. And there's value being the fastest runner. And at a trough multiple and going earnings 50% Kagar over the next three years. It's difficult to look
26:44
Unidentified speaker (speaker_2): beyond that as well. Again, dependent on your investment horizon. Maybe no short-term catalyst, but longer term very all these cost companies are very well positioned. I would say where to avoid to the other part of your question, as we look to 2027. Certainly a lot of investors have been long the memory trade
27:04
Unidentified speaker (speaker_2): in the first half of 2026 and late 25. I think that becomes a lot more tricky going forward because, as again, memory is more of a reflection of supply and demand. The demand's significantly ahead of supply at the moment, prices popped from here, they kind of normalize. And as we go into 28, we
27:24
Unidentified speaker (speaker_2): probably have a lot more supply coming online. So they they're not going to benefit for as much from pricing going forward as they have over the last six to nine months. So I think that's a sub-segment sub segment of semis. I would tend to avoid. I would say. And then also another sub segment is obviously consumer, right? Because of these
27:44
Unidentified speaker (speaker_2): high memory prices, it that's really resulting in demand destruction across consumer. PCs this year down kind of ten, fifteen percent, smartphones similarly. If we look to next year twenty 2027 twenty seven, and and a lot of that is demand destruction at the low end because of memory prices. They're and just too high and they've eaten into cash flows
28:04
Unidentified speaker (speaker_2): and margins. If we look to 2027, again, consumer is going to be under pressure until we get that normalization of memory pricings, which at the earliest now that's going to come in 2028. At the earliest, so consumer really is going to be under pressure through 27, and maybe only in 2028 it'll get some
28:24
Unidentified speaker (speaker_2): reprieval. So probably consumer. and memory are two sub segments within semiconductors in general that I would tend to search of.
28:33
Michael Kantrowitz: Thank you so much for your time. This was great. Thanks for jumping on, and we would love to have you on again soon.
28:38
Unidentified speaker (speaker_3): We hope you enjoyed today's episode of What's Next for Markets. You can subscribe to the show on Apple, Spotify, YouTube, or wherever you get your podcasts.
28:47
Unidentified speaker (UNKNOWN): Be
28:47
Unidentified speaker (speaker_3): sure to follow Michael on X at Michael Cantro. If you like, the show, please let us know by leaving a positive review. The preceding commentary is not financial advice. You should always seek appropriate financial advice from a professional financial advisor in your state. You should not consider this as a substitute for these services. Piper Sandler does and seeks to do business with companies covered in
29:07
Unidentified speaker (speaker_3): its research commentary or coverage. As a result, investors should be aware that the firm may have a conflict of interest that could affect the objectivity of this commentary. Investors should consider this commentary as only a single factor in making their investment decisions. For full disclosure information, visit www.pipersandler.com/slash research disclosures.
This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.

More from What's Next For Markets

All insights