Brown Advisory CIO Perspectives

Brown Advisory CIO Perspectives: AI Infrastructure Politics, Anthropic IPO, and Market Volatility

53 min Listen to the original Episode page
AI InfrastructureAnthropic IPONVIDIACredit MarketsQuant VolatilityJapanese ActivismBiotech M&AMarket Structure

Summary

Sid Ahl and Jordan Wruble discuss how US election results, particularly in Texas, are introducing political uncertainty into AI data center build-outs, potentially slowing 2029+ timelines. They analyze the shift in AI investment dynamics, noting that hyperscalers like Amazon and Microsoft are the primary drivers of capital expenditure, while questioning the sustainability of Anthropic's usage-based business model ahead of its IPO. The speakers also address NVIDIA's leadership change and its implications for industry financing. In credit markets, they describe a bifurcation between high-quality and lower-quality credits, driven by higher rates and increased supply from AI infrastructure needs. Finally, they highlight extreme factor-level volatility in equities due to quant deleveraging and identify Japanese activism and biotech M&A as key investment opportunities.

Key takeaways

  1. 1
    Texas is slated to bring on approximately 30% of 2028 planned data center capacity, making state-level political shifts critical for infrastructure timelines.

    Sid Ahl · 0:59

  2. 2
    Bloomberg estimated that up to 40% of total US data center build-out is in states where newly elected governors advocated for restrictions or pauses, introducing a new political uncertainty dimension.

    Unidentified speaker (speaker_1) · 2:32

Episode sections

  1. 0:38 – 7:23

    Market Volatility and Political Impact on AI Infrastructure

    Sid Ahl and Jordan Wruble discuss how US election results, specifically governor races in Texas and other states, have introduced uncertainty into AI data center build-outs. They note that incoming governors in key states like Texas have expressed concerns about local impacts, potentially slowing development.

  2. 7:23 – 14:41

    AI Investment Dynamics and Hyperscaler Dominance

    The speakers analyze the shift from a 'beta' phase of broad AI investment to an 'alpha' phase requiring selectivity. They argue that hyperscalers (Amazon, Alphabet, Microsoft) are the primary drivers of capital expenditure, not just frontier labs, due to cloud migration and enterprise AI adoption.

  3. 14:41 – 25:25

    NVIDIA Leadership Change and Ecosystem Impact

    Discussion centers on the recent leadership change at NVIDIA and its implications for the AI ecosystem. The speakers highlight NVIDIA's role as a financier for the industry and question whether the new leadership will maintain the aggressive capital support for neo-clouds and frontier labs.

  4. 25:25 – 34:03

    Anthropic IPO and Business Model Analysis

    The hosts discuss the upcoming Anthropic IPO, noting its strong October revenue run-rate but highlighting the complexity of valuing a usage-based business model. They debate the sustainability of Anthropic's lead against open-source models and the potential for cybersecurity and coworking to become major revenue drivers.

  5. 34:03 – 36:46

    Anthropic IPO Economics and Compute Costs

    Sid Ahl notes Anthropic's projected $200 billion cash warpost and $80 billion revenue base, arguing this provides a training advantage. Jordan Wruble counters that compute costs may be lower than feared, citing an estimate of $11-13 billion per gigawatt per year, compared to SpaceX's higher short-term rates.

  6. 36:46 – 39:59

    Credit Market Bifurcation and Supply Dynamics

    The speakers discuss a divergence in credit markets where high-quality credits remain tight while lower-quality credits struggle due to higher rates. They note that AI infrastructure build-out is shifting funding needs from operating cash flow to credit markets, increasing supply and creating patience among lenders.

  7. 39:59 – 41:15

    Credit Hedge Fund Deleveraging and Forced Selling

    Wruble describes a cycle of forced deleveraging in credit hedge funds, where capital withdrawals from multi-manager funds trigger liquidations. This creates a feedback loop of pressure on credit prices, similar to dynamics seen in equity markets.

  8. 41:15 – 42:38

    Equity Market Structure and Momentum Volatility

    Sid Ahl highlights that while index volatility is subdued, factor-level volatility is extreme, citing Morgan Stanley's high momentum tech index moving >5% on 65 trading days in 2026, exceeding the combined total of 1999-2001. This is attributed to multi-manager hedge funds and quant strategies.

  9. 42:38 – 45:31

    Quant Strategies and Short-Term Orientation

    Wruble explains that quant strategies isolate idiosyncratic risk and apply leverage, leading to high portfolio turnover (10-12 times per year). This creates a short-term orientation where stop-losses and drawdown limits trigger synchronized selling, causing dislocations.

  10. 45:31 – 49:07

    Portfolio Construction for Volatility Management

    The speakers argue that long-term managers must use diversification and patient capital to withstand short-term dislocations caused by quant deleveraging. Brown Advisory is shifting clients toward passive or passive-linked strategies to manage tracking error and free up capital for active managers.

  11. 49:07 – 50:07

    Japanese Activism and Corporate Reform

    Wruble identifies Japanese activism as a key opportunity, noting a cultural shift where management teams are now welcoming investor input rather than resisting it. This reflexivity is seen as a driver for improved economics and higher valuations.

  12. 50:07 – 53:07

    Biotech M&A and Patent Cliffs

    Biotech is highlighted as an area benefiting from deleveraging and continued M&A activity. Large pharma companies face patent cliffs at the end of the decade and have cash surges from GLP-1 drugs, creating a need to acquire new therapeutics, potentially accelerated by post-election regulatory clarity.

Assets discussed

AssetKindStanceContext
NVIDIANVDAequityneutralMaybe we could talk about another recent news story, which is the change of leadership at NVIDIA, which has people up in arms as well.
Meta PlatformsMETAequityneutralZuckerberg at Meta, Sundar. at Google, you have big, very convinced and deep pocketed folks backing this.
AlphabetGOOGLequityneutralZuckerberg at Meta, Sundar. at Google, you have big, very convinced and deep pocketed folks backing this.

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

Claims worth verifying

Texas is slated to bring on something like 30% of 2028 planned data center capacity.

Sid Ahl ·1:19

How to verify: Check Texas state government infrastructure reports or major data center developer press releases regarding 2028 capacity projections. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Bloomberg estimated as much as 40% of the total US data center build out is in states where newly elected governors advocated for restrictions or pauses in development.

Unidentified speaker (speaker_1) ·2:32

How to verify: Search Bloomberg News archives for articles citing this specific 40% statistic regarding US data center build-out and state-level political restrictions. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

High bandwidth memory supply will not catch up to demand even in a slower build-out for the next few years.

Unidentified speaker (speaker_1) ·7:09

How to verify: Review semiconductor industry supply/demand forecasts from firms like Gartner, IDC, or major memory chip manufacturers (e.g., SK Hynix, Micron) for the 2026-2026 period. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Semiconductor stocks have sold off 30 to almost 40% from their very peak.

Sid Ahl ·7:43

How to verify: Calculate the percentage decline from the all-time high to the current price of major semiconductor ETFs (e.g., SMH, SOX) or index components using financial data providers like Bloomberg or Yahoo Finance. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

NVIDIA is generating obscene amounts of cash flow and has a 70% revenue increase next year.

Unidentified speaker (speaker_1) ·16:45

How to verify: Check NVIDIA's (NVDA) latest quarterly earnings report (10-Q or 10-K) filed with the SEC EDGAR for revenue guidance and cash flow statements. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Primary source

NVIDIA traded at less than 10 times forward earnings.

Sid Ahl ·17:31

How to verify: Calculate NVIDIA's (NVDA) forward P/E ratio using current stock price and consensus analyst forward earnings estimates from sources like Bloomberg or FactSet. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Primary source

Anthropic announced a $17 billion project in Wyoming.

Sid Ahl ·6:31

How to verify: Search for press releases from Google (Alphabet) or Wyoming state government regarding a $17 billion data center investment in Wyoming. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Anthropic's October run-rate revenues were $8.1 billion.

Unidentified speaker (speaker_1) ·27:52

How to verify: Check Anthropic's S-1 filing or investor presentations for revenue data covering the four weeks ended in October. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Primary source

Anthropic's gross margins on inference are 80 plus percent.

Unidentified speaker (speaker_1) ·30:50

How to verify: Review Anthropic's S-1 filing or financial disclosures for gross margin percentages specifically attributed to inference services. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Primary source

Over 10% of Anthropic's revenues were from Chinese companies paying for model distillation.

Sid Ahl ·13:01

How to verify: Check Anthropic's S-1 filing or investor presentations for breakdown of revenue by geography or customer type, specifically looking for mentions of Chinese clients or distillation services. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Primary source

Anthropic will have approximately $200 billion of cash after its IPO.

Sid Ahl ·34:23

How to verify: Check Anthropic's S-1 filing with the SEC EDGAR database for the IPO prospectus and cash position disclosures. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Primary source

Anthropic has $80 billion of revenue tied to its models.

Sid Ahl ·34:03

How to verify: Check Anthropic's S-1 filing with the SEC EDGAR database for revenue disclosures. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Primary source

SpaceX charges $40 billion for short-term compute deals.

Sid Ahl ·35:46

How to verify: Verify via press releases or contracts related to SpaceX's Colossus data center compute leasing agreements. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Anthropic's compute costs are estimated at $11-13 billion per gigawatt per year.

Unidentified speaker (speaker_1) ·35:04

How to verify: This is a manager estimate; verify by checking investor presentations or reports from the specific manager cited (unnamed) or industry reports on AI compute economics. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Morgan Stanley's high momentum technology index moved by more than 5% on 65 trading days in 2026.

Sid Ahl ·41:35

How to verify: Calculate using historical daily return data for the Morgan Stanley High Momentum Technology Index from a financial data provider like Bloomberg or FactSet. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

The 65 trading days of >5% moves in the Morgan Stanley index exceed the combined total of 1999, 2000, and 2001.

Sid Ahl ·41:35

How to verify: Compare the 2026 count against historical data for the same index for the years 2026-2026 using a financial data provider. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Hedge funds are the most underweight tech and AI since COVID.

Sid Ahl ·48:07

How to verify: Check weekly hedge fund positioning reports from prime brokers like Goldman Sachs, JPMorgan, or BAML. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

2026 has been a rare year in which earnings expectations have risen throughout the year.

Unidentified speaker (speaker_1) ·52:14

How to verify: Compare consensus earnings estimates (e.g., S&P 500 EPS) at the start of 2026 versus current estimates using FactSet or Bloomberg. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Large pharma companies are facing a patent cliff at the end of the decade.

Unidentified speaker (speaker_1) ·50:27

How to verify: Review patent expiration schedules for major pharmaceutical companies (e.g., Pfizer, Eli Lilly, Novo Nordisk) from company investor relations pages or FDA databases. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Biotech deleveraging has created opportunities, with M&A driven by GLP-1 cash flows and patent cliffs.

Unidentified speaker (speaker_1) ·50:47

How to verify: Analyze M&A transaction data in the biotech sector and correlate with pharmaceutical revenue reports for GLP-1 drugs and patent expiration dates. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Primary source

Do your own research

  • Verify Texas data center capacity projections and political statements from incoming governors via state government websites or Bloomberg Energy reports.
  • Check NVIDIA's forward earnings guidance and current valuation multiples against recent SEC filings and analyst consensus.
  • Review Anthropic's IPO prospectus for detailed revenue run-rates, compute cost assumptions, and cancelation clauses in power agreements.
  • Analyze credit market spreads between triple-B and double-B rated bonds to assess the impact of interest rates on borrower cash flow.
  • Monitor Morgan Stanley's high momentum technology index performance data to validate claims regarding factor-level volatility in 2026.
  • Research Japanese corporate governance trends and activist investor activity reports to evaluate the cultural shift in management responsiveness.
  • Examine large pharmaceutical companies' patent expiration schedules and GLP-1 drug sales data to assess biotech M&A potential.
  • Compare Anthropic's projected compute costs against actual industry benchmarks for gigawatt-scale power commitments.
Full transcriptexpand
0:04
Unidentified speaker (UNKNOWN): Hello
0:05
Sid Ahl: and welcome to CIO Perspectives. I'm Sid all, the co-CIO of Private Client Endowment Foundations here at Brown Advisory. And today I'm joined by my colleague Jordan Rubel, the head of our investment solutions group, responsible for external manager research. Jordan's done a bit of everything in his career from investment invest in banking to private equity and running money at hedge funds and bank
0:24
Sid Ahl: proprietary trading desks. And we've got a lot to talk about today. So I'm excited to talk about the broad swath of external managers that we invest with, what they're saying about the current investment environment, and what Jordan's views are. So Jordan, welcome back to the pod.
0:38
Unidentified speaker (UNKNOWN): Thanks,
0:39
Unidentified speaker (speaker_1): Sid.
0:39
Sid Ahl: As mentioned, lots to talk about today. Maybe we could start a little bit with the volatility in markets that followed the elections in the U.S. So I know you and I talk, we all talk a lot about how hard it is to predict the results of an election and how the market's going to react to them. In some ways, this is maybe another example of that. Markets are higher
0:59
Sid Ahl: after the election, even though they sold off initially. But there were some pretty important developments in the blue sweep. So most importantly, maybe the governor races in a few states which flipped Democratic. Texas is a prime example, the incoming governor. is decidedly less data center friendly. And Texas
1:19
Sid Ahl: is slated to bring on something like 30% of 2028 planned data center capacity. So, what are you hearing from managers with regard to how this might impact the build-out and perhaps the AI infrastructure stocks that are so dependent on it?
1:33
Unidentified speaker (speaker_1): I think the election introduced yet another dimension of uncertainty. And I already think just as we think about the environment we're in, a wide range of outcomes from what the technology can accomplish to how the markets will react to it. The technology, on one end, is going to cure all diseases, and on the other hand, we're talking about the
1:52
Unidentified speaker (speaker_1): probability of doom. I think you can make a credible argument for all kinds of outcomes. From the investment perspective, we've got unbelievable earnings potential, but the durability of the drivers of that earnings are uncertain, not just because of the technology, but the financing piece, potential for competition. And yes, now we've got politics. So the
2:12
Unidentified speaker (speaker_1): election did raise concerns that the hyperscaler investment in compute would get delayed or severely limited as this sort of NIMBYism of building data centers in one state became a big campaign issue. And it was an issue that. really was bipartisan. In Texas, we saw this from the incumbent governor and the incoming
2:32
Unidentified speaker (speaker_1): electee. And as you've noted, a huge percentage of the build-out is going to take place in Texas. And I think Bloomberg estimated as much as 40% of the total US data center build out is in states where newly elected governors advocated for restrictions or pauses in development. So that's a new uncertain dimension. And we already had this
2:52
Unidentified speaker (speaker_1): weird tension where on the one hand, we were afraid we would overbuild, and on the other hand, we wouldn't put enough compute into operations to satisfy earnings expectations for a lot of the infrastructure stocks, including the chip stocks. So that's part of it. I think there's a couple of healthy aspects to this. For one thing, I think in the race to
3:12
Unidentified speaker (speaker_1): deploy capital to secure compute, need you know, we've seen some very expensive prices paid. And it's certainly conceivable that some of the deals that were struck weren't the best ones that might be available. I think slowing the pacing down allows those who want to maybe slow walk some of the deals they've done or walk back some of the deals they've done to do
3:32
Unidentified speaker (speaker_1): so. And that could be To healthy. And I think eventually the compute and the infrastructure trade will depend a lot on. the technology and the return on investment, and less necessarily about the timing, right? I think people are very, very focused on the timing of what happens in what year. If this sort of smooths that growth curve out a little
3:52
Unidentified speaker (speaker_1): bit, it might be in fact very healthy. The other thing is I have trouble seeing that we won't continue to invest, particularly some states will be ready to negotiate. And we saw some announcements. that suggested that maybe in a state like Wyoming, you'd have a more friendly environment. So I think ultimately this allows states and municipalities to
4:12
Unidentified speaker (speaker_1): strike a good deal for people who live there. And given the gigantic cost of deploying compute, a little extra to smooth things over in a local community to give some incentives for that to make people in that community better. off should hopefully help get us to the needed level of compute in the ground.
4:31
Sid Ahl: Yeah, I think your point about slower might be better is something I've definitely thought a lot about. I think the pace that we were setting was spooking investors to a certain degree. It was certainly spooking debt markets. The pace of supply of new debt coming online was pushing rates higher. We started having a lot more talk. about is this
4:51
Sid Ahl: a bubble? There was a lot more talk about an earnings cliff in the future. And if we could smooth that out, that might be a positive. Yeah. I mean, there's kind of no way to sugarcoat the outcome as being potentially bad for data center development in 2028 and beyond. I do think a lot about next year, 2027, right?
5:11
Sid Ahl: We as investors were always thinking, what's the probability that we've we have a recession? What's the probability that earnings go down in the next 12 to 18 months? I do think it's important to remember that 2027, those projects are overwhelmingly approved or under construction. So it's seems more to be like a 2028 issue. And and the big question is you
5:31
Sid Ahl: bring up is are those 2028 projects that may get delayed to 2029 or might they just not happen at all? I know in Texas specifically, the new governor wants legislation to govern any new projects. That's going to take a bit of time, and you have no idea how that's going to turn out. But I agree with you that, you know, my sense
5:51
Sid Ahl: is that a lot of these states don't really want to stop the data center development. They just want it done right. They want it done more slowly. They don't want electricity bills to go up. They want water recycled. And one thing I think a lot about with regard to Texas is that not every data center developer there is one of the big hyperscalers. There's a lot of wildcatters, so to speak, a
6:11
Sid Ahl: lot of, you know, unproven neo clouds that are not playing by these same rules. And I think we forget that sometimes. This isn't a rebuke necessarily of how Microsoft or Amazon or Google are doing things. It's the others. And then, as you noted, right, Google announced. just last week it was the $17 billion project in Wyoming. The president of Meta
6:31
Sid Ahl: came out right after the election and said that one of these governors who'd been pretty anti-data center called the moment the results were in and said, I want to meet, I don't want these jobs going elsewhere. So you know, we'll see how this all shakes out. But I think as you point out, one thing that hasn't really changed is the demand. We'll get into that a little bit later.
6:50
Unidentified speaker (speaker_1): One of the concerns. we had, and we talked to managers about right, there's certain markets like high bandwidth memory, where some of the dynamic in it today is there's a shortage. If you slow down growth, does that tip that into surplus? And at least their view was no, there's not an expectation of the supply of high
7:09
Unidentified speaker (speaker_1): bandwidth memory catching up even in a slower build-out, build out, at least for the next few years. And so you don't have the risk of that turning into a glut. And I think those are some of the questions we have for managers about how the slower build-out might affect economics.
7:23
Unidentified speaker (UNKNOWN): Well,
7:23
Sid Ahl: that maybe leads into the next topic, which is obviously there's been a lot of consternation with regard to the quote unquote AI trade. Semiconductor stocks have sold off 30, almost 40% from the very peak. And as you note, memory stocks or other pockets, neo. clouds, have sold off even more. And it's driven by this growing
7:43
Sid Ahl: concern about the ability to finance and even just build the data centers, but also, as you noted, the revenues that may be generated to pay for them. I'm curious what you're hearing from investors on these key debates. You kind of alluded to a handful of managers talking about probably.
8:03
Sid Ahl: not arriving at a glut, but we've seen the valuations because the earnings haven't moved, the earnings expectations haven't moved, the valuations have just been decimated for a lot of these companies. What are managers doing about positioning in this area?
8:16
Unidentified speaker (speaker_1): We're talking about the range of outcomes. It's been such an unusual and volatile time in which, because outcomes are so wide, a lot of things change just on narrative. There's not enough data to support. a very tight range, whether it's valuations or earnings expectations. And so you've seen these huge swings. I think a year ago you and I spoke and super bullish on
8:36
Unidentified speaker (speaker_1): AI build-out. And it was because of that point, we're continually underestimating the need for capital investment. And so every time a bank or somebody made an announcement, it was we're going to spend more, we're going to spend more. And that was bullish for the supply chain. As we got into Q2, I think we reached a little bit
8:56
Unidentified speaker (speaker_1): more toward fuller expectations, right? A little higher set of expectations, and probably just in terms of CapEx, reached the upper bound of what could be expected from the ability to finance it to actually execute and shovels in the ground and so forth, which meant you probably. he had a little more
9:15
Unidentified speaker (speaker_1): downside risk from the market perspective. So I think the way we've been talking about it broadly is the beta phase is probably over or ended around then. And it became more of an alpha phase. It became more of a not all of these things can be true. We can't be both overbuilding and undersupplied. And so being very selective
9:35
Unidentified speaker (speaker_1): is pretty interesting. And the beginning of the year.
9:39
Unidentified speaker (speaker_1): with the disruption of Claude Code and all software businesses, Motes being in question, you actually saw the hyperscalers be victims of a little bit of the fear around AI. And it was a little bit of this well, are they going to overbuild? But also are their core businesses at risk. And as we get now toward the end of the year, the
9:59
Unidentified speaker (speaker_1): hyperscalers seem really well positioned. And whether it's because future build out will get slowed down or not, it's having compute is seen clearly as very, very valuable. The overall AI capex cycle and build out, we often heard people talk about will the OpenAI and anthropic revenues primarily be enough
10:19
Unidentified speaker (speaker_1): to support a return on that capital investment. And the kind of numbers that are being floated for ending 2028 run. rate were really, really high. Upwards of a trillion dollars combined for leading lab run rate revenues ending 2028, some numbers below that. But like we've come to believe from talking to managers is probably not the only way
10:39
Unidentified speaker (speaker_1): to think about this. And maybe partly it's because the capital being outlaid is actually more from hyperscalers like Amazon, Alphabet, Microsoft. And what's their return is going to be not just lab rep. revenues. So part of the drive for compute is we've had for a while now this migration to the cloud. But as you have these AI tools, you need
10:59
Unidentified speaker (speaker_1): more and more workflows in the cloud. And so you're seeing that migration accelerate. And they're just more drivers of cloud revenues than pure frontier token delivery. And so we think the right focus is probably yes, leading labs have to perform well. but also the hyperscalers are really the ones paying for most of this capital. And
11:19
Unidentified speaker (speaker_1): they are seeing the data. They are seeing the returns. They're seeing the incremental spend. They're also hearing from their customers. And so I think some of the fear is probably less if you think about it from the hyperscaler lens and whether or not they're getting the proper return. And then we do think that there's like still a really enormous opportunity for anthropic and open.
11:39
Unidentified speaker (speaker_1): AI and that intelligence is going to be a really big market.
11:42
Sid Ahl: Yeah. I think it's interesting. People, as you say, they're grasping at straws at times to define the narrative of what's most important in AI. And I think there was kind of a shift to saying, well, it's all about the anthropic revenue number or the OpenAI revenue number. And you kind of miss the fact that 80 to 90% percent of the hyperscaler spending is
12:02
Sid Ahl: not related to. the frontier labs. And even if the frontier AI companies are gonna make up a growing piece of some of the investment in this CapEx, the hyperscaler is just still the dominant force. And as you say, companies are continually moving more data to the cloud and they want increasingly to use AI on that
12:21
Sid Ahl: data. And whether that's a frontier AI or an open weight AI or a combination of the two. that's the clear trend. And that is what is mainly supporting the infrastructure build out. I also think people have probably become a little too focused on another issue, was the slowdown for anthropic
12:41
Sid Ahl: in the third quarter. And as you and I've talked about, we heard a lot of feedback from people on some of the drivers of that. Some of it was cutting back on. the Chinese distillation, right? Over 10% of revenues at one point were Chinese companies that were paying Enthropic to use the model and reverse engineer it to make their own
13:01
Sid Ahl: models. That's been mostly cut off now. And that was a one-time hit, but probably a pretty big long-term positive for the business. There's also, we forget, it's only a few months ago, you know, the government pausing the release of some of their models. in the beginning of the quarter. That was another delay. They launched a new model towards the end of the quarter and they
13:21
Sid Ahl: cut prices. And usually there's a bit of a J curve to that, and they're coming out of it now and had a great October. So I just think it's been really hard for investors to understand what are the key drivers. And as you say, it's become a lot more narrative driven, I think, as a result of that, which has led to some. pretty crazy volatility, which I'm excited to talk to you
13:41
Sid Ahl: about. I think one thing also, the last thing I think about is just again, it keeps coming back to demand. The unit of demand, one of the units of demand we talk about a lot is tokens, right? These AI tokens. That growth is still going up 20 to 30x per year, which is a big number. And I think
14:01
Sid Ahl: in some ways the biggest risk to AI infrastructure and the build-out is actually more technological. It's kind of a race to see how fast are the chips and the models getting better so they can produce more tokens per unit of compute versus that 20 to 30x per year.
14:21
Sid Ahl: And so I think that's one of the things that I'm monitoring the most. It still appears. that compute is not growing nearly fast enough to meet that 20 to 30x growth, but we shall see. Maybe we could talk about another recent news story, which is the change of leadership at NVIDIA, which has people up in arms
14:41
Sid Ahl: as well. How do you think this might impact the AI ecosystem?
14:45
Unidentified speaker (UNKNOWN): Well,
14:46
Unidentified speaker (speaker_1): it's worth pointing out that this news is a day old, a few hours old at this point. And so I think we're going to continue to learn a lot. And I don't know when this gets released, but we haven't yet had the anthropic IPO just a level set when people are listening. But that's expected very soon. So the NVIDIA news again, just like another thing that this investment environment
15:05
Unidentified speaker (speaker_1): didn't need was just again more uncertainty. NVIDIA obviously has leading accelerator technology and is an incredibly important enabler of everything happening in AI. But it also is playing this secondary role as sort of the banker to the build out that I think is concerning.
15:25
Unidentified speaker (speaker_1): People have talked about circular finance in the sense that whether it's equity stakes in companies or credit enhancement or different roles that they're playing is they're tying capital to customer purchases. which kind of is smart, right? If you're them, you're locking in customers on your technology and you're enabling them to
15:45
Unidentified speaker (speaker_1): buy more of it, which the market wasn't willing to provide. NVIDIA is generating obscene amounts of cash flow, just enormous growth. And so it's been this big facilitator of some of the acceleration and some of the excitement around all whole space with Jensen out. I think we don't. know. I think for one thing, we're still learning a lot about the
16:05
Unidentified speaker (speaker_1): details of what was happening at NVIDIA. I think there'll probably be a lot more that comes out over the coming days. And so it's a little hard to see. But will whoever formally takes that role on a full time basis have the same strategy of doing those kind of deals? Will the counterparties want to do the same kind of business as
16:25
Unidentified speaker (speaker_1): they were doing with Jensen? I think those are really hard to say right now. And so I do think that there's a healthy fear premium right now around that. That said, there's certainly a lot of near-term visibility on demand for NVIDIA chips that they got to do a 70% revenue increase next year, which was
16:45
Unidentified speaker (speaker_1): capacity constrained and could have been higher. That puts them on an incredibly cheap near-term. earnings multiple. And tying back to your question before of will there be a cliff of CapEx? And it's hard to say whether the future earnings certainly may not grow as much, perhaps will decrease. And that seems to be what the market has been pricing in is a real
17:05
Unidentified speaker (speaker_1): healthy decline in earnings because otherwise it's just hard to explain why that earnings multiple would be so low. Yeah,
17:11
Sid Ahl: it's funny. Nvidia's become a value stock now, right? I mean, briefly a couple days. of ago, I traded at less than 10 times forward earnings, which is a shocking multiple for a company that has just told us it's going to grow 70% next year. And in some ways, I guess you could say a more conservative management strategy might be more appealing to
17:31
Sid Ahl: value investors looking at NVIDIA as a stock to own. But when I think about the ecosystem, I think it's hard not to say that this is probably a negative for areas like the neo clouds, which Jensen was so early and important in backing the core weaves and nebiuses of the world and
17:50
Sid Ahl: many others, right? Cutting equity checks, backing them explicitly and implicitly. And you've certainly seen those stocks punished in recent days as a result of that. It was a smart move for him, right? Diversifies his customer base. It promotes new entrants, makes a more competitive ecosystem. And also they were captive
18:10
Sid Ahl: to the NVIDIA ecosystem because of that support. Probably are going to get less support even of the Frontier Labs, although they may not need it as much. But NVIDIA has announced or announced a few weeks ago they were going to cut a $10 billion check for the Anthropic. IPO. We don't know if that's definitely going to happen,
18:30
Sid Ahl: but you might imagine there will be fewer of those $10 billion checks cut in the future for OpenAI or anthropic if they were to need capital again in the future. And I guess, you know, for me, like the one question I'm wrestling with is has this maybe happened at a time where he was so important in setting up this ecosystem, but certainly
18:50
Sid Ahl: anthropic. it looks like it's ready to kind of exist on its own. It's about to go public. It's about to have a lot of cash on its balance sheet. OpenAI, not quite there yet, but these neo clouds maybe have approached a scale where they can go it on their own. And Jensen a few weeks ago was saying when everyone was up in arms about Core Weave and some of their debt
19:10
Sid Ahl: issues, that Core Weave kind of has to figure it out on their own. So maybe we've reached. a point where it's more self-sustaining. And I get a lot of questions about just in general, did you need that magic of Jensen to kind of power this industry forward? We have a lot of other pretty big AI pilled and very well capitalized people,
19:30
Sid Ahl: including Dario at Anthropic and Sam at OpenAI, and obviously the leader, you
19:36
Unidentified speaker (UNKNOWN): know,
19:36
Sid Ahl: Zuckerberg at Meta, Sundar. at Google, you have big, very convinced and deep pocketed folks backing this. So it may be that the whole industry is at a point where they're a little bit less dependent on Jensen than they were maybe three or four years ago.
19:55
Unidentified speaker (UNKNOWN): It
19:55
Unidentified speaker (speaker_1): all kind of fits together, right? If we're slowing down just a little bit, there's a little less need maybe to raise as much capital a little. less need for them to be the banker in the middle of it all. To you might have to digest some uncertainty and maybe some mistimed allocations, but I think you can still land in a very healthy place. And it is
20:15
Unidentified speaker (speaker_1): interesting in the context of anthropic, right? Nvidia strategically is trying to support open source. They don't want to have one or two big customers. They want to have a lot of inference being done. And so their efforts to kind of be the provider of more palatable provider of open weight models has been one of their strategic initiatives. That's another thing that
20:35
Unidentified speaker (speaker_1): perhaps slows down. And maybe that benefits anthropic and OpenAI a little bit.
20:40
Sid Ahl: That's a really good point. And something people have written and talked about a lot in the last few days is, you know, he was probably the biggest champion for open weight models and making sure we didn't ban Chinese open weight models and white. what have you. And who knows where that might go from a regulatory standpoint in the coming years. So something to keep an eye on. How does this overall just
20:59
Sid Ahl: impact your longer-term views about how to position portfolios to benefit from this growth of AI capabilities and usage?
21:07
Unidentified speaker (speaker_1): Well, the volatility makes everything pretty tough, but it's funny because I feel like the next couple of years are pretty uncertain, but I feel much more confident. about 10 years from now. Now it doesn't take away what I said earlier about the range of outcomes, but I do think that we're headed to a world in which intelligence is ubiquitous and it's going to be pretty different. And a
21:27
Unidentified speaker (speaker_1): little bit I draw from the 90s and the internet when we knew this cool technology was doing interesting things, but you still hadn't had so much of the development that was yet to come. And whether that was broadband, and persistence, you know, eventually mobility, there were things that were hard to even envision at that point that the internet
21:46
Unidentified speaker (speaker_1): enabled.
21:47
Unidentified speaker (UNKNOWN): I
21:47
Unidentified speaker (speaker_1): think AI will have some of the same things. And already whether the history has been we've got chat, and that's exciting, and then reasoning models and now agents. And maybe the next leg there is persistent agents. And I think there's some parallel to the broadband persistent. on with like a always running intelligence workforce.
22:07
Unidentified speaker (speaker_1): So I feel like the tenure outlook is maybe a little bit easier. And again, I don't know who the winners will be necessarily, but you can see why people are excited about Anthropic. Having a lead on intelligence matters not just to get premium pricing, but actually like they can deliver, I think, strategic advantages beyond just the frontier. They can
22:27
Unidentified speaker (speaker_1): be the most cost effective. I think they already are the most cost effective at serving intelligence. And so if they chose to compete with open weight by cutting prices, they'd still be able to do that and because they're the cheapest to serve. And in some ways, like their advantages might be almost Walmart-like, where they have the most compute and they have the ability to be the cheapest
22:47
Unidentified speaker (speaker_1): provider. They don't necessarily have to price that way. Certainly, I don't think they will. at the higher end of intelligence, but they might have that ability at lower ends to compete with or more than compete with where open weight is today. So we're excited about that coming to market. I think the infrastructure players, they've reset expectations very
23:06
Unidentified speaker (speaker_1): substantially here from where they were at the end of Q2. And I think there are a number of opportunities that we hear from managers where it's not the hyperscalers, it's not just Frontier Revenues or with hard disk drive companies, it's not just the data center build-out, but like the creation of data and the amount of stuff
23:26
Unidentified speaker (speaker_1): that needs to get stored. So there are going to be, I think, you know, winners in this, regardless of all the uncertainty. And I think that that's the approach that we've been taking, which is find great managers who are experts in this area. let them do bottom-up diligence on who the most likely to prevail are, both in terms of their understanding of the
23:46
Unidentified speaker (speaker_1): technology and the roadmap, but also like where are expectations for the stock. And as we mentioned with NVIDIA, and you could bring up some of the memory companies, like expectations are pretty modest right now. So I think that's creating a very interesting opportunity. At the same time, we have a lot of exposure. to managers who do great work on the short side, and there'll be companies that
24:06
Unidentified speaker (speaker_1): aren't going to be as successful that get a lot of hype and premium for their potential that the reality doesn't meet. And there's still going to be a lot of disruptive businesses. It's almost easier to be confident on the short side in some ways, because whether or not the economics of AI are durable, the disruption to incumbents seems very likely to remain a risk.
24:25
Sid Ahl: Yeah, I think one has to have a lot of humility in this environment in terms of where the value may accrue. Because as you said, the range of outcomes is so wide. How much is going to go to the model companies? How much to the hyperscalers? How much to the infrastructure companies? How about just the companies and the consumers that are using the technology? How much of this might be surplus
24:45
Sid Ahl: to them? And those questions will be decided over time. I want some exposure to all. of those areas. I've been focused most on making sure we have coverage and potential exposure to those areas. So we have a lot of exposure to the hyperscalers and the companies that may be benefiting from using AI. I think the pieces of the story
25:05
Sid Ahl: we've been talking about the last couple of years, we've had less exposure to until last year was to the AI infrastructure companies and to the model company. companies. So I think that's where we had to be more proactive in getting that exposure in portfolios. And I still feel pretty strongly we want that exposure or that potential exposure, that
25:25
Sid Ahl: expertise in portfolios today, despite the big drawdown in the infrastructure stocks. Because I mean, as you note, it's almost been like a market on fast forward. We've seen this really, really violent resetting of expectations, which has hasn't actually shown up in earnings estimates yet, but it's very clearly shown up in multiples. And there's a lot of carnage
25:45
Sid Ahl: out there that I want some folks sifting through at the moment. And I'm excited to have them sifting through it. So you touched on anthropic and the upcoming IPO. Again, recording note, we're recording this on Friday, the 13th, pre-anthropic pre anthropic planned IPO. But what are you hearing about the interest in the IPO for
26:04
Sid Ahl: managers across the portfolios? And what are the key debate points you're hearing? I know you touched on a couple of them already, but what are you hearing?
26:12
Unidentified speaker (speaker_1): Yeah, you know, as with all things, it's mixed. We have some funds that have been invested in Anthropic privately, and they remain extremely bullish. And I think that transparency into its development has really helped them. One thing that's become clear, and I think this is true historically investing in technology is these things don't develop linearly. And so there's a
26:32
Unidentified speaker (speaker_1): patience with those who've been close to the company with some of the vicissitudes of what's happened in revenue growth, for example. And you mentioned Q3 deceleration for anthropic. Q2 was extraordinary and probably over-accelerated. And this is how technology diffusion works. So we have some
26:52
Unidentified speaker (speaker_1): managers that are, I think, really broad generalists showing real interest in the IPO. And I think we're pleased to see that. But I think it'll be a hard company for a lot of people to underwrite initially. I mean, there's a lot of uncertainty around the business and even how you value it. So my guess is there's going to be a lot of
27:12
Unidentified speaker (speaker_1): learning post-IPO post IPO to understand the business. and get a better sense of what it can do. Right now, they're doing a roadshow, and I think people are getting some access, but I think it'll take time, right? You talked on the Q2 piece, and I do think that was a big part of it, both the cutting off of distillation and the price cut that they did for new models, as well as that tough comp. And
27:32
Unidentified speaker (speaker_1): I think that kind of spooked Spook the market. We just got October run rate. They updated the S1 and saw really, really strong acceleration from October coming out of that price cut and from the new models. I think it was 81 billion run rate revenues using the four weeks ended in October. And
27:52
Unidentified speaker (speaker_1): that will help. That I think is a great backdrop for the IPO. There's a bunch of debates around it. So one is we've had coding as is like the perfect first use case. It's being used by developers who are comfortable and understand how to deploy technology. It's got great data sets to train on. That was adopted faster than
28:12
Unidentified speaker (speaker_1): probably anybody could have imagined. I think as you look out to the next vectors, the knowledge work opportunity is enormous. I mean, just to put it in context, I think Anthropic said something around an average customer or their largest. enterprise customers, they're spending less than 0.1% of opx on
28:31
Unidentified speaker (speaker_1): Anthropic and the most closer to 1%. But there's a huge potential opportunity, but rolling that out will be harder than it was for coding, right? There's people like me trying to make sense of how to use these things best. And there's a decent amount of trial and error. I think that we're just in very, very early days. And what that means is there will be some. volatility
28:51
Unidentified speaker (speaker_1): around revenue. And that will probably create interesting opportunities, but that's part of it. There's a big debate around whether Anthropic can maintain its lead. Our view is probably yes, that there are real returns to being in the frontier as they're using their own technology to develop the next model. And, you know, this pursuit of RSI of just like
29:11
Unidentified speaker (speaker_1): using the technology, you know, recursive self-improvement to like get further. and further ahead. And again, their investment in compute and their commitment to spending a ton and using a ton of their compute on training, you know, should keep that lead ahead. But that's a debate. Open source is a debate. We've seen a huge shift in tokens toward open source. What we haven't
29:31
Unidentified speaker (speaker_1): seen is revenue shift toward open source. And so I think that my guess is where this goes is company like Anthropic. and we'll see about OpenAI. As they get enough compute, they're more competitive in the quality. I guess you talk about like cost per intelligence rather than per token, but like in the cost per intelligence for
29:50
Unidentified speaker (speaker_1): lower intelligence tasks that they will be able to compete there. Part of it, I think one of the interesting things just developed was open source looked very good on benchmark testing in part because these tests were open book. and some of the new testing doesn't disclose what they're testing for. And in that scenario, the open source looks considerably worse
30:10
Unidentified speaker (speaker_1): relative to leading models. So I think open source is certainly a debate, but we still see a very strong case for the Frontier Labs to win that and across the Pareto Frontier. So then you come down to business model, and I think this. is where it's tricky. We always talk about ARR, and we're used
30:30
Unidentified speaker (speaker_1): to that being recurring. These are still run rate revenue ARRs. So what that means is you get into the holiday season here in November with Thanksgiving and the holidays in December, and you're going to see revenues slow down. And I think their guidance was like they met half their quarter guidance from just October. And I think that's quirky, right? And what that means is
30:50
Unidentified speaker (speaker_1): like the visibility. on some of this is still to be determined. Now, will they eventually sign software like ARR deals? I would imagine so. But then how do you value a business where their gross margins on inference are extraordinary, 80 plus percent, but so much of their spend is in R D and I'm developing the next model. I
31:09
Unidentified speaker (speaker_1): think people can make the argument that, well, if you slow down R RD D like any business, you'd have massive cash flows. I think that's true. But does that mean something about your competitive positioning? It's certainly not something we anticipate anthropic doing anytime soon. And so the revenue multiple needs to at some point tie to a
31:30
Unidentified speaker (speaker_1): cash flow multiple. And when you see that, unclear. But as we talk about sort of a range of outcomes, there's certainly a right tail here. where this is the most valuable company in the world by market cap, and then some. And so I think it's a pretty exciting opportunity. And I do think that it'll take some time for some managers to get comfortable with it. But if the
31:50
Unidentified speaker (speaker_1): development happens as our managers who are closest to the company expect, we'll see a lot of interest in owning it.
31:55
Sid Ahl: Yeah, it's funny. Even some of those managers that we know who've made big bets, very large bets. And it was one of the things I think that caught both of our eyes, just how much conviction people had in this company the last year and a half. Even those investors would say this is one of the hardest companies to model that
32:15
Sid Ahl: they've ever invested in. Because as you point out, right, this isn't really recurring revenue, it's a usage-based model. And we are still exploring what all these use cases are.
32:24
Unidentified speaker (UNKNOWN): I
32:25
Sid Ahl: do think it's interesting. that Anthropic has been telling investors on this roadshow and in their investor day that the adoption curves, the revenue ramps for their cybersecurity and for their cowork are faster than for coding, even if they're still very early.
32:45
Sid Ahl: That is at least encouraging and suggests there could be two more big markets for them. But your point about diffusion, how fast organizations are going to be willing to adopt those two areas versus coding, I think is such a key question. And we just don't know the answer to it. But as you say, there's so many other
33:04
Sid Ahl: potential motes for this company. It's the talent that they have. I think people underestimate how long Anthropic was starved for capital. Everyone was. funding OpenAI and they weren't giving money to Anthropic. And that made them focus on being very efficient in how they design their models to use a very
33:24
Sid Ahl: limited amount of computing power to produce the best results. And that, to your point, many argue has made them the lowest cost provider of compute. That is a potential moat. The recursive self-improvement self improvement is a potential moat, right? If you have the best mode. models that aren't released to the public and you're increasingly using them to train the next model,
33:44
Sid Ahl: that could give you a compounding lead. And pulling up the drawbridge on the Chinese models by cutting the distillation perhaps will allow you to see more of that over time. And then probably it shouldn't be lost on people that it's a big advantage to be the company that is monetizing. AI
34:03
Sid Ahl: the best with regard to models right now. Having $80 billion of revenue that is tied to your model allows you to invest the most of anyone relative to that revenue back into training the next model. And that I think could be a significant advantage for them over time.
34:23
Sid Ahl: Also, after this IPO, they're going to be sitting on, probably. 200 billion dollars of cash. And yes, they have a lot of big commitments over the next few years, but if their revenue kind of ramps the way people think, then that could be an advantage for them as well. So really interesting, very excited. Uh, by the time people are listening to this, the IPO likely will have already
34:43
Sid Ahl: happened. But
34:44
Unidentified speaker (UNKNOWN): yeah,
34:45
Unidentified speaker (speaker_1): I think that just as you bring up the cash and their commitments, I think one of our managers made a really important point. which was we're seeing the sort of cost for compute go up. And some of this is just the enormous revenue generation that like SpaceX's Colossus has gotten for short term deals. And it's sort of
35:04
Unidentified speaker (speaker_1): thrown off a little bit where the hurdles are, I think, in people's minds. And this manager's estimate was that anthropics costs will be more in the eleven 11 to thirteen. 13 billion per gigawatt per year, which is a much lower number than I think what some have talked about. And so what they need to achieve in terms of
35:24
Unidentified speaker (speaker_1): revenues to continue to be able to make those build-out commitments is actually much, much less. And it will be interesting, right? I mean, I think some of the business model uncertainty, right? Some of their costs look pretty high right now because of those short-term short funds. term funding deals. I think that will become more evident in time and it will de-risk a little
35:44
Unidentified speaker (speaker_1): bit the level of commitments that they have. Yeah.
35:46
Sid Ahl: I mean, SpaceX is charging forty billion and the hyperscalers, if they had the ability to provide that to Anthropic or the willingness to do it, that would be like twelve. So it's you know basically three X as much. I think to your point, there were some people who said, Oh God, the cost of this compute is 40 billion now, you know, multiply
36:06
Sid Ahl: that times all the compute that they're trying to bring online, that's a lot of cost. Whereas it perhaps is just as likely that in 12 months they cancel their contracts with SpaceX and move it to cheaper compute. That's possible. And that to your point changes the economics pretty considerably. I think it's also important to
36:26
Sid Ahl: remember. they have a big number of compute commitments, but half of that number is beyond 2030. And something like half of the number before 2030 is cancelable. So versus that $200 billion war chest, it's a more reasonable amount. But maybe we could shift gears. We've talked a
36:46
Sid Ahl: lot about AI, and we could talk a little bit about credit. markets right now. Credit markets have been struggling a bit this year, and there's this really interesting bifurcation of the highest quality credits still being pretty tight and priced nicely, and the lower quality credits really struggling. We've seen,
37:06
Sid Ahl: perhaps surprisingly, in credit markets, some headlines of managers, a little bit like we saw from AI managers who struggled in July that are struggling.
37:15
Sid Ahl: in this environment. What is happening in high yield and distress credit markets this year? How are managers responding? Are we getting a more attractive opportunity set?
37:24
Unidentified speaker (UNKNOWN): I
37:24
Unidentified speaker (speaker_1): thought you wanted to move on from AI.
37:26
Sid Ahl: Well,
37:27
Unidentified speaker (UNKNOWN): I you know, I guess this is
37:29
Unidentified speaker (UNKNOWN): tell me how related.
37:30
Unidentified speaker (speaker_1): It's still very related. Well, I would look in the very recent past, you've seen a pretty big divergence from the weakest credits, triple C credits, or are widing widening credits. considerably, whereas slightly less stress credits, double B's have not. And some of that is higher rate environment taxes borrowers. They're
37:50
Unidentified speaker (speaker_1): gonna have a real impact on cash, and that's gonna weaken businesses. And we have seen a pretty sharp move up in rates. The part of this that ties to AI is you've just had a lot of issuance, and you're gonna going have to more as that build out we've been talking about is move from operating cash flow to the credit markets. And so whether that's long
38:09
Unidentified speaker (speaker_1): duration IG, which is funding a good portion of this, but maybe doesn't have as many natural buyers to bespoke chip finance deals and other kinds of clever financing, there's a you know incrementally more supply. And when you have a lot of supply and the expectation of more, there's kind of a you
38:28
Unidentified speaker (UNKNOWN): know
38:28
Unidentified speaker (speaker_1): couple dynamics. One, our managers are pretty attracted to some of these opportunities that come out, and that pulls capital away from some existing corporate investments that maybe are similar spreads and yields, but less attractive credit. And so that's putting pressure on parts of the market. And the other part is the expectation
38:48
Unidentified speaker (speaker_1): that there's so much more yet to come, is that you want to be patient with what you commit to today, because there'll be more coming tomorrow. And as lenders get filled up, there'll be even less ability to absorb that. So being patient because the better deal may be tomorrow. And that's meaning
39:06
Unidentified speaker (UNKNOWN): you
39:06
Unidentified speaker (speaker_1): have to offer more and more attractive returns. So that's part of it. Credit markets over the last several years have been, particularly in the stuff that our hedge fund managers do, right? The stressed and distressed market have been beset by what is euphemistically called liability management exercises, but these are what we used to call creditor on creditor violence. And this
39:26
Unidentified speaker (speaker_1): has made visibility in credit investing much harder. There's ways to deprive value from one group of creditors to benefit another. And I think it's changing a little bit the math on what makes that market interesting.
39:39
Unidentified speaker (speaker_1): And I think it's just getting a little bit harder when it's not the same set of rules or the same set of criteria, which is part credit analysis and valuation, part understanding how restructuring might play out. And there's always been negotiation in that. But then the power dynamics of how lenders are behaving is making this environment really, really tough. So then I'll
39:59
Unidentified speaker (speaker_1): add on another factor, which we've seen some drawdowns and we're seeing accelerated deleveraging in credit. And we should actually probably come to this because it's happening just as much in equities, if not more so. But we've seen some credit hedge funds that have faced pressure and their taking down risk has pressured other credit hedge funds.
40:19
Unidentified speaker (speaker_1): And there are some headlines about this. And then we've seen where some of the really large multi-manager multi manager hedge funds have allocated money to. some credit shops and have started pulling back that capital, which creates another cycle of forced liquidation and delevering. So I think this is probably creates a pretty interesting opportunity, but we certainly have
40:39
Unidentified speaker (speaker_1): managers where historically they've been pretty low vol and they've been pretty consistent performers where this has been over the last few months a bit of a headwind and probably creates a pretty good opportunity. going forward, but drawdowns in the space have been worse than typical.
40:55
Sid Ahl: Yeah, it's interesting. We talk a lot about the influence of levered hedge funds on the equity market. It feels newer that this is happening in credit. Maybe we could talk about that broader question that you brought up. I mean, I think one of the defining characteristics of this year has been all of these storms under the surface of the broader market.
41:15
Sid Ahl: both in equities and in credit. So the kind of index volatility has been pretty subdued, but we've seen really high volatility at the stock or the credit level or the so-called factor level, momentum being probably the key factor in the stock market that is really driving things. I mean, there's a great statistic that captures this. We
41:35
Sid Ahl: talked about it the last few days. There's been 65. I believe, trading days where Morgan Stanley's high momentum technology index has moved by more than 5%.
41:48
Sid Ahl: That's more than we had in 1999, 2000, and 2001 combined. And this is not in a bear market. This is in a market where the stock market's going up and index level volatility has been normal. normal-ish.
42:04
Sid Ahl: So, you know, clearly the market is struggling to process the wide range of potential AI outcomes that you talked about, the build out, but there's also something about market structure that really seems to have changed. There's a lot of money that's been raised by multi-managers in the hedge fund space, quants. There's a real
42:23
Sid Ahl: resurgence of retail trading and a surge. in options trading. What do you make of these storms under the surface? Are they here to stay? Does it make it a better or a worse environment for active managers? What's your take?
42:38
Unidentified speaker (speaker_1): So I think by and large, it's here to stay. And I think it's partly because of at least the multi-manager part of it, I think is probably going to be a persistent influence. And I think this is an area that there's been just tremendous innovation.
42:51
Unidentified speaker (speaker_1): and to the benefit of investors around how do you constrain certain types of risk and this kind of heavy factor analysis, heavy factor neutralization. And they built a very impressive model that isolates what's idiosyncratic, strips out factor risks long and short, and
43:11
Unidentified speaker (speaker_1): then applies leverage to this sort of low vol idiosyncratic. piece to generate a very attractive return. And I think that has a lot of scary terms in it, the leverage part particularly, but we've seen the model work pretty well. And I think intuitively, when you get rid of all of those things that drive stock movements but
43:31
Unidentified speaker (speaker_1): aren't the kinds of risk you want to take, there's a lot of benefit. However, because it's been attractive, you've seen a tremendous inflow of capital in. into the space. And the assets that these types of strategies are managing have ramped up quite a bit. And it just creates a very short-term orientation for a lot of market participants, right? You take all the capital they have
43:51
Unidentified speaker (speaker_1): and all the leverage. When you're in that construct, one of the things is many of them manage risk by having pure drawdown limits. So stop losses on a portfolio or a team that's running capital. So as there's unexpected drawdowns, they cut risk very aggressively. And it's that risk cutting, or what we call delevering, right? Because the gross
44:11
Unidentified speaker (speaker_1): leverage in the system is coming down, that creates losses through the system. So you're selling your longs and covering your shorts, and the offset That is not completely symmetrical. Like most people in the market have somewhat overlap. overlapping biases in what makes a good investment in a good company
44:31
Unidentified speaker (speaker_1): and what makes a bad one in a good short. And so as they're selling longs and they're pressuring longs, other people with those types of constructs are facing losses and they need to de-risk. They're covering their shorts and paying a higher, higher price for those. And you see negative PL. Now they're set up not to lose a tremendous amount of money in those settings, but it pressures things, but it also
44:51
Unidentified speaker (speaker_1): means that a lot of those investors are turning their portfolios 10, 12 times a year and they're focused on how the market might move the stock the next 5, 10, 15% over the coming month to three months. And that creates a really short-term orientation. So is this good or bad? Well, it can be both. On
45:11
Unidentified speaker (speaker_1): the one hand, it's very frustrating when a long-term manager manager's portfolio acts in a kind of unpredictable way in the short term. And you've seen that a lot of late. On the other hand, it creates pretty meaningful dislocations that a long-term investor can take advantage of. And so the key is really having managers structurally able
45:31
Unidentified speaker (speaker_1): to take advantage of those. And that's both a mindset of hey, we want to invest for the long term, but it's also how they build their portfolios, right? They have to have enough staying power, right? Which you might accomplish through diversification or the time you enter these positions, right? So diversification can be lots of forms, but where you are in the value
45:51
Unidentified speaker (speaker_1): realization path. And then you need to be aligned with your investors, right? The most important thing for those managers is they need patient capital. They need to be able to have that kind of volatility. So I think you can get very good and attractive returns, but I think you have to be patient for near-term performance to be more
46:10
Unidentified speaker (speaker_1): frustrating. So one question might be: well, how do you deal with that? First of all, it's like manager selection, we think is incredibly important, finding people who are skilled at that, thinking about how not just how they underwrite stocks, but how they build portfolios and whether they're well aligned with their managers. But it's also for our clients, right? Some of this is we've just seen an increase in how
46:30
Unidentified speaker (speaker_1): portfolios will differ from, say, benchmarks. And we've tried to rein that in a little bit. So it's not a hundred percent bottom-up active management the way it used to be for a lot of our clients. We'll use passive, we'll use passive linked strategies, whether that's portable alpha, alpha extension, as ways to free
46:50
Unidentified speaker (speaker_1): up some. of the capital to perform better in the long run while taking more tracking error, some of that volatility. So there's a whole number of things that have to feed into capitalizing on this, but I think some of that short termism, which can be frustrating, does create pretty good long-term investments for our clients.
47:08
Sid Ahl: Yeah, I think your points on portfolio construction, having the staying power. I mean, you know, this has definitely been an advertisement for that, right? We've seen some people get way too over their skis in AI or in credit or what have you. And the volatility is basically taking them out of the trade, which is obviously we talk
47:28
Sid Ahl: about this a lot. The number one goal for us working with our clients is building a portfolio that will allow them and us to play offense when there are great opportunities. to be greedy when others are fearful. It's even more important with all this volatility under the surface that we're giving money to managers who are adhering to
47:48
Sid Ahl: those same rules. And again, great advertisement in 2026 for that. I think it's interesting, as you mentioned, this kind of deleveraging. We've seen a pretty big deleveraging the last few weeks, right? By some measures, a few of these weeks were. kind of the biggest moments of taking risk off for the hedge fund industry in
48:07
Sid Ahl: since COVID. We saw last week the biggest day of retail selling since COVID. The prime brokers who deal with all the hedge funds have been telling us that hedge funds are now the most underweight tech and the most underweight AI and the lowest net exposures they've had basically since COVID.
48:27
Sid Ahl: So thinking about all of that, it certainly suggests that we may be in one of those moments where there are some great opportunities for active managers with a longer term time horizon. And if we are to get positive data points in broader economy markets, AR, what have you, there might be a lot of catching up to do for some of these managers who've
48:47
Sid Ahl: cut a lot of their risk. And you say the same thing about credit markets. I guess maybe that leads to we could do a couple of closing questions. What are managers most excited about today across all these markets? Is it Japanese activism as we've talked about a lot? Is it some of these bombat AI infrastructure stocks? Is it something else? What are you hearing?
49:07
Unidentified speaker (speaker_1): Hard to generalize. Yeah. I mean, we like those things, right? Japan's been an area we've really built up a pretty meaningful allocation to uh on the corporate reform story and the again, the reflexivity of that improving the economy and making hopefully spurring higher valuations. Activism has been great because
49:27
Unidentified speaker (speaker_1): whereas it used to be met with tremendous resistance, the corporate culture is changing. And now you're seeing management teams welcoming input from investors, welcoming activists. And there's a lot still to do. So that's an area where we have some specialists in that space and we have some generalists who have
49:47
Unidentified speaker (speaker_1): done more there. It's a pretty interesting one. You know, I think part of the delevering, right, has not only made some of these AI and technology stocks interesting, but biotech, we've seen a pretty big delevering in. And that was something we talked about a year ago. That's been a great market. And I think the opportunity remains pretty interesting. But
50:07
Unidentified speaker (speaker_1): also with this delevering, I think there's some pretty good alpha opportunities. One of the drivers in biotech is going to be continued MA. This industry is one in which large pharma companies generally kind of outsource their RD. You know, they buy promising new therapeutics and put them through their distribution pipeline.
50:27
Unidentified speaker (speaker_1): The big pharma companies have benefited from existing drugs, but a lot of those are facing a patent cliff at the end of the decade. And combine that with big revenue and cash flow surge from some of the GLP ones. And you see a kind of need to go out and replace these revenues as they hit patent cliffs and a lot of cash to do so and
50:47
Unidentified speaker (speaker_1): buying power, including leverage. But I think with the election last week, there's probably a bit of a near-term focus on what will the permissiveness, what will the environment be for MA post 2028? And I would have to believe that people are pulling forward their plans to do MA pretty rapidly. And
51:07
Unidentified speaker (speaker_1): biotech's one area where we think that will benefit people. I think that managers are probably a little frustrated by the volatility, but tend to look at that as an opportunity.
51:15
Unidentified speaker (speaker_1): We've seen some pullback in AI from July through certainly October, probably into the beginning of this month. But you and I were on a call with a manager that said, Hey, we used the term playing offense for the first time in a meeting in a while. We haven't actually talked about that, right? So I think we're getting through the kind of de risking, delevering d-levering
51:35
Unidentified speaker (speaker_1): portion. There do seem to be some pretty attractive investments. I And think there's still a lot on the short side. And I think we've seen that in gross exposure from managers on the short side, is pretty healthy right now. And market multiples come down a bit. Like there seems to be a lot to do. And I guess some of this is contingent on the economic backdrop, which has been pretty healthy despite a
51:55
Unidentified speaker (speaker_1): rise in rates. Probably Usually all bets are off if that slows down. If that slows down, because everything we talked about at the beginning of the call, the uncertainty around. the AI build-out, you know, that because rates get so high that they pressure the economy. But right now we're seeing very, very healthy earnings growth. 2026 has been a rare year in which earnings
52:14
Unidentified speaker (speaker_1): expectations have risen throughout the year. Normally you come into a year and expectations are pretty high. Analysts like to set high forecasts, and then those expectations kind of moderate throughout the year. We've seen the opposite. So the backdrop's really healthy. And I don't keep saying how wide the range of outcomes can be and how uncertain things are, that can in
52:34
Unidentified speaker (speaker_1): fact all change. And there's probably some real risks developing. But with that backdrop, I think there's a lot of excitement about investing right now.
52:42
Sid Ahl: That seems like a great place to close it out, Jordan. A little bit of optimism on the opportunity set, but with the requisite doses. dose of humility. I appreciate that. Thanks so much for joining today, Jordan.
52:54
Unidentified speaker (speaker_1): Great to be here. Thanks, Sid.
52:55
Sid Ahl: And thanks everybody for listening. We'll see you next
52:58
Unidentified speaker (UNKNOWN): time.
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.
All insights