William Taylor discusses AI infrastructure spending, data center power demands, and the DTCR ETF
Summary
William Taylor argues that the AI market is in an execution phase with hyperscalers spending hundreds of billions on infrastructure, creating opportunities in data centers and power management rather than just chips. He highlights that Australia is well-positioned to benefit due to land, renewables, and connectivity, with Goodman Group pivoting heavily to data centers. Taylor excludes neoclouds from his new ETF (DTCR) due to financing risks, focusing instead on physical infrastructure like Vertiv and Goodman. He identifies grid power access and regulatory delays as the primary bottlenecks for the build-out.
Key takeaways
Episode sections
- 0:00 – 3:01
Introduction and Market Context
Speakers introduce the topic of AI data centers, noting recent volatility with Firmus pulling its float and NEXTDC falling 42% in a day, despite trillions in hyperscaler spending.
- 3:01 – 5:29
State of AI Infrastructure and ETF Launch
Will Taylor assesses the market as being in an execution phase with healthy price corrections, launching the ETF Shares Global AI Infrastructure and Data Centres ETF (DTCR) to target infrastructure rather than just chips.
- 5:29 – 9:49
Hyperscaler Spending Scale
Hyperscalers have spent ~$750 billion this year with Goldman Sachs estimating over $1 trillion next year, a scale described as three times the Apollo space program.
- 9:49 – 11:41
Australia's Role in Data Centers
Australia is positioned to benefit due to land availability, renewable potential, stable government, and subsea cabling to Asia, with Goodman Group pivoting 80% of its future book to data centers.
- 11:41 – 16:04
Bubble Assessment and Valuations
Taylor argues valuations are high but not a bubble because earnings growth supports them, contrasting with the dot-com era where valuations lacked profit backing.
- 16:04 – 17:40
Neoclouds and Circular Financing
Neoclouds are a small part of the story; their financing is often collateralized against GPUs and contracts, reducing systemic risk if they fail.
- 17:40 – 24:08
Data Center Power and Cooling Mechanics
AI data centers consume power equivalent to 100,000 households; cooling is critical, with liquid cooling using water towers or air-cooled radiators, each with distinct energy and water trade-offs.
- 24:08 – 27:39
AI Capex Allocation Breakdown
Of every $100 in AI CapEx, $50 goes to chips, $15 to networking, $20 to power, and $15 split between construction and cooling, explaining the focus on infrastructure.
- 27:39 – 28:53
Lesser-known companies behind the racks
Speaker_3 identifies Vertiv as a key player in power management and thermal regulation, and Goodman Group as a strong Australian representation with 80% of future revenue prospects tied to data centers. NEXTDC is also included but in a small portion of the portfolio.
- 28:53 – 29:45
Exclusion of neoclouds
The speaker explains that neoclouds like Firmus are excluded from the ETF because they are viewed as a financing layer rather than pure infrastructure, involving heavy borrowing to buy and rent GPUs.
- 29:45 – 31:12
Bottlenecks slowing the build-out
The primary bottleneck is securing power from the grid, followed by government scrutiny and regulatory delays, particularly in Australia, which can extend project timelines to two to four years.
- 31:12 – 33:31
Investment options: ETFs vs stocks
ETFs are recommended for investors who lack time for deep research or access to international markets like Taiwan, though they may result in double-counting if investors already hold underlying companies.
- 33:31 – 36:59
Are we too late?
The speaker notes the index has annualized 35% over the last five years but argues the story is not over due to increasing CapEx and long-term contracts, advising dollar-cost averaging for long-term horizons.
Assets discussed
| Asset | Kind | Stance | Context |
|---|---|---|---|
| AlphabetGOOGL | equity | neutral | We just had Alphabet's quarterly results. Their cloud revenue more than doubled year on year, and yet the backlog still grew. |
| MicrosoftMSFT | equity | neutral | I think between what Microsoft, Amazon, Oracle and Alphabet, the cloud backlog is now what, two and a half trillion dollars. |
| AmazonAMZN | equity | neutral | I think between what Microsoft, Amazon, Oracle and Alphabet, the cloud backlog is now what, two and a half trillion dollars. |
| Meta PlatformsMETA | equity | neutral | And then you've also recently seen Google and Meta actually raise capital through selling or issuing more equity. |
| S&P 500 | index | neutral | Look at the S P 500. |
| NVIDIANVDA | equity | bullish | Um and NVIDIA's obviously become you know the most valuable company in the world as a result, but roughly half of the money actually doesn't go into chips. |
“Stance” is how the speakers talked about the asset, not a recommendation.
Claims worth verifying
NEXTDC fell 42% in a day.
Unidentified speaker (speaker_0) ·1:18
How to verify: Check NEXTDC (ASX: NXT) share price history for the date of the 'wobble' mentioned in the episode context (likely late 2026/early 2026 based on transcript timing relative to release). For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Firmus pulled its float.
Unidentified speaker (speaker_0) ·1:18
How to verify: Search ASIC or ASX announcements for Firmus IPO withdrawal. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Hyperscalers have spent around $750 billion this year.
Unidentified speaker (UNKNOWN) ·6:16
How to verify: Sum of quarterly CapEx filings (10-Q/10-K) for Alphabet, Microsoft, Amazon, and Meta for the relevant fiscal year. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Primary sourceGoldman Sachs estimates hyperscaler spending will be over a trillion next year.
Unidentified speaker (speaker_3) ·6:38
How to verify: Search for Goldman Sachs research reports on AI CapEx forecasts. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Anthropic is looking to spend $850 billion in the next couple of years.
Unidentified speaker (speaker_3) ·7:44
How to verify: Check Anthropic press releases or official announcements regarding capital expenditure plans. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Goodman Group has 80% of its future book going to the data center space.
Unidentified speaker (speaker_3) ·10:32
How to verify: Review Goodman Group (ASX: GMG) investor presentations or annual reports. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Microsoft increased its accounting lease life for data center assets from 15 to 25 years.
Unidentified speaker (speaker_3) ·26:40
How to verify: Check Microsoft (NASDAQ: MSFT) 10-K or 10-Q filings for changes in useful life estimates for property and equipment. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Primary sourceOne AI rack generates heat equivalent to 55 portable heaters.
Unidentified speaker (UNKNOWN) ·21:15
How to verify: This is a speaker analogy; verify via technical specifications of AI server racks (e.g., NVIDIA HGX) and thermal output data. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
An AI data center uses the same amount of power as 100,000 households.
Unidentified speaker (UNKNOWN) ·18:25
How to verify: Compare average household electricity consumption data with reported power draw of typical hyperscale data centers. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Taiwan is the main country to supply chips, with TSMC acting as the foundry for many GPUs and having infrastructure for Apple.
Unidentified speaker (speaker_3) ·27:39
How to verify: Check TSMC's investor relations page or annual reports for customer concentration and manufacturing capabilities; verify via semiconductor industry reports. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Primary sourceVertiv is a company focused on power management and thermal regulation for data centers.
Unidentified speaker (speaker_3) ·27:59
How to verify: Check Vertiv's (VRT) company website or SEC filings for business segment descriptions. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Primary sourceGoodman Group has 80% of its future revenue or prospects in data centers.
Unidentified speaker (speaker_3) ·28:19
How to verify: Check Goodman Group's (GMG) latest annual report or investor presentations for revenue breakdowns and future outlook. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Neoclouds like Firmus are excluded from the ETF because they are a financing layer, borrowing heavily to buy and rent GPUs.
Unidentified speaker (speaker_3) ·28:55
How to verify: Check the ETF's prospectus or fact sheet for exclusion criteria and holdings; verify Firmus's business model via news articles or company announcements. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Primary sourceThe biggest bottleneck in the AI build-out is getting power from the grid.
Unidentified speaker (speaker_3) ·30:18
How to verify: Review industry reports on data center power demand and grid capacity constraints from energy or tech consultancies. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Government scrutiny is a bottleneck in Australia, slowing down data center projects.
Unidentified speaker (UNKNOWN) ·30:46
How to verify: Check Australian government planning approvals or news reports on specific data center projects facing regulatory delays. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
The ETF's index has annualized 35% over the last five years.
Unidentified speaker (speaker_1) ·34:05
How to verify: Check the performance data of the ETF's underlying index or the ETF itself on the ETF Shares website or financial data providers. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Primary sourceHyperscalers have contracts locked in for the next couple of years.
Unidentified speaker (speaker_0) ·34:48
How to verify: Check hyperscaler earnings calls or investor presentations for guidance on capital expenditure and contract durations. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.
Do your own research
- Verify hyperscaler CapEx forecasts against Goldman Sachs research reports and company earnings calls.
- Check Goodman Group's annual report for the specific breakdown of its 'future book' allocation to data centers.
- Review Vertiv's financial filings for revenue exposure to AI data center power and thermal management.
- Assess regulatory approval timelines for new data centers in Australia via state government planning portals.
- Compare DTCR ETF holdings to avoid double-counting if you already hold individual AI infrastructure stocks.
- Investigate the debt levels and collateral structures of neocloud firms like Firmus to understand systemic risk.
- Check water and energy usage disclosures for data centers to assess environmental and operational risks.
Full transcriptexpand
0:00Unidentified speaker (UNKNOWN): So
0:00Unidentified speaker (speaker_0): I'm just gonna ask it straight up,
0:01Unidentified speaker (UNKNOWN): are
0:02Unidentified speaker (speaker_0): we in a bubble?
0:02Unidentified speaker (speaker_1): An AI data center uses the same amount of power as a hundred thousand households. So that's like the size of Hobart. And that's just one of the smaller ones. They get built regardless of whose model wins, and they need to get paid up front to build the data centers before anything happens anyway. If it's all giving you the same answer, then where does the pricing power go? One of these AI racks is let's say a size of a the fridge. It's kicking out heat, which is equivalent. to 55 portable heaters. Oh, I'm sure there's some spare electricity going around. I'll plug in from a neighbour. I'm sure he doesn't mind. Are we too
0:36Unidentified speaker (speaker_0): loud?
0:38Unidentified speaker (speaker_2): Welcome to another episode of Equity Mates, a show where we explore what's possible in the world of investing. If you've just joined us for the first time, I'm a huge welcome to our community. My name is Bryce. And
0:47Unidentified speaker (speaker_0): I'm Ren, and today we're talking all things data center. centers. Bryce, sometimes I think about a simpler time 10 years ago when no one was talking about data centers.
0:56Unidentified speaker (speaker_2): No one.
0:56Unidentified speaker (UNKNOWN): Five years ago. Yes.
0:58Unidentified speaker (speaker_0): But data centers now are front of mind for markets. They are the narrative driving markets. Some of the best performers in markets around the world are data center companies. But we're talking about them at a really interesting time. There's been a bit of a wobble from some of the smaller players here in Australia. Thermos couldn't IPO. Next DC fell 42% in a day. There are questions being asked about the sustainability of the data center story. And yet, there is trillions of dollars going towards these sheds with servers and racks in them. And today's guest thinks we may still be early in the story.
1:38Unidentified speaker (UNKNOWN): Joining
1:38Unidentified speaker (speaker_2): us in the studio. today, we're joining Ren. I wasn't actually part of this interview, but Will Taylor came in. He is the portfolio manager at ETF Shares.
1:46Unidentified speaker (speaker_0): Yeah, and they've just launched an ETF uh on the ASX. It's the ETF Shares Global AI Infrastructure and Data Centers ETF. The ticker is DTCR. So we talked all things data centers and AI infrastructure. And as I said, uh this story is capturing my market's attention. Will thinks we may still be early in the story, largely because the flood of money that is coming. So we asked the big question, are we in a bubble? We asked the other question, how early or late are we in this trade? But that's not all we talk about. We don't just talk about how the market is responding to data centers. I actually wanted to get into like how should we understand data centers beyond power go. goes in, tokens come out, GPUs sit in the middle. Uh, and actually, like, what are some of the components uh that make up a data center? And so we speak about the huge power needs, the cooling, the different options there, uh, and some of the obscure companies that are making a hell of a lot of money supplying like bits and bobs to data centers around the world.
2:48Unidentified speaker (UNKNOWN): A
2:48Unidentified speaker (speaker_2): massive thank you goes to ETF shares for supporting. today's episode and for helping us keep all of our content free. Well, with that said, let's get into it. The conversation with Will Taylor and Ren.
3:01Unidentified speaker (speaker_0): Will, welcome back to Equity Mates.
3:02Unidentified speaker (speaker_1): Thanks for having me back.
3:03Unidentified speaker (speaker_0): We're talking today about the topic that's front of mind for every investor AI, infrastructure, and data centers. It seems to be, well, in many ways, the theme that is driving markets at the moment. And you guys have stepped right into the middle of that conversation, launching an ETF, an AI infrastructure and data centers ATF on the ASX. So let's just talk about what's happening today. And then we're going to talk about like data centers more generally and how investors can approach it. But for people who maybe don't have their finger on the pulse as much, it's been an interesting month. We had Firmus Fermous call off what would have been Australia's biggest float in 20 years, core weave in the US. their bonds fell into distressed territory. Next DC couldn't get power, uh, to one of their data centers dropped 40% in a day. With all that going on, how do you assess the state of the AI infrastructure story at the moment?
3:51Unidentified speaker (UNKNOWN): Yeah,
3:51Unidentified speaker (speaker_3): you know, you're not really selling it there, are
3:53Unidentified speaker (UNKNOWN): you?
3:54Unidentified speaker (speaker_3): But no, it's been an interesting month or so. There's definitely been a few um areas of the story which have come under a little bit of screen. scrutiny, and rightly so. I think it's just a case of sometimes we get all get excited. These companies in particular maybe take on a little bit too much debt, and probably healthy,
4:13Unidentified speaker (UNKNOWN): it's
4:14Unidentified speaker (speaker_3): they've had a kind of a bit of a price correction and a bit of a pullback. But in terms of where we are in the story, I would say we're kind of in that execution phase. Like at the start of the year, everyone was talking about the hyperscalers, how much money they were spending. And now that's happening. Now they're actually spending that money. Um, and you're not you're seeing opportunities and potential cracks. And that's where we've seen
4:36Unidentified speaker (UNKNOWN): a
4:37Unidentified speaker (speaker_3): company like Fermos try to get on a part of the a part of the game, and it's probably not worked out the way it should. Um, but the reason why we're offering this ETF and we're coming to market is we actually want wanna to give give investors investors unique exposure. The companies that are in our ETF are actually aren't in the Nasdaq. Everyone thinks
4:57Unidentified speaker (UNKNOWN): you're
4:57Unidentified speaker (speaker_3): semiconductors, so you're Nvidia's of the world, the hyperscalers. But what we want to do and give investors opportunity is invest in those companies which are actually building the networking, the power connections to the data centers, the cooling. So it's it's a really good time to actually potentially buy it because the market has had this kind of pullback and it's no longer potentially launching at where markets are at their most enthusiastic. There's actually had a bit of doubt. And as an investor, that's a great time.
5:29Unidentified speaker (speaker_0): Yeah. Well if you said my last question wasn't selling it, maybe this uh the setup for this question will be because you know whilst the market you know there's a bit of a wobble and maybe people are a bit nervous about some of the data center operators, the there is an ungodly amount of money being made and being spent. We just had Alphabet's quarterly results. Their cloud revenue more than doubled year on year, and yet the backlog still grew. I think between what Microsoft, Amazon, Oracle and Alphabet, the cloud backlog is now what, two and a half trillion dollars.
6:02Unidentified speaker (UNKNOWN): That's
6:02Unidentified speaker (speaker_0): just it's unprecedented amounts of money. So how do we reconcile these two stories? Like the story of the wobble and investors being nervous with this money that's still being spent?
6:13Unidentified speaker (speaker_3): Yeah, I mean,
6:14Unidentified speaker (speaker_1): that's a
6:15Unidentified speaker (UNKNOWN): better
6:15Unidentified speaker (speaker_3): question.
6:16Unidentified speaker (UNKNOWN): I'd rather answer that one. Yeah, yeah. Yeah. So
6:19Unidentified speaker (speaker_3): the at the start of this year, like they said they were gonna spend a lot of money, and so far they have. They've spent around seven 750 hundred and fifty billion this year. Um and it's expected to go up. Markets have kind of struggled with that concept because it seems like such a large amount, and it is, but what we're starting to see now is just that growth of the revenue side uh come through. Goldman's kind of estimate over a trillion next year, which again is seems like a crazy amount when you put that in perspective, kind of the US highway interstate, that's more money
6:54Unidentified speaker (UNKNOWN): that
6:54Unidentified speaker (speaker_3): they're spending in the next couple of years, is more than the the money they've spent in the last thirty-five years just on the highway. And then if you look at like the Apollo space program, it's three times as much as what they spent there. And that's in just a couple of years. Investors are asking questions, which they should, which is what which is what investing is. And you've actually seen like a few of the hyperscalers ha be in negative territory year to date. And versus the Nasdaq, the the four hyperscalers have actually underperformed. So it's kind of shown you that maybe the place to be is not the hyperscalers where they're spending the money. And that in part is actually fifth like half of where they're spending the money is going in the the infrastructure space.
7:34Unidentified speaker (speaker_0): Yeah, yeah, yeah. Real picks and shovels.
7:36Unidentified speaker (UNKNOWN): Yeah, exactly.
7:37Unidentified speaker (speaker_0): Yeah.
7:38Unidentified speaker (UNKNOWN): You know,
7:38Unidentified speaker (speaker_0): you said the Goldman are estimating over a trillion dollars next year just from the hyperscalers spending
7:43Unidentified speaker (UNKNOWN): on data
7:43Unidentified speaker (speaker_0): centers.
7:44Unidentified speaker (speaker_3): That's excluding Anthropic coming out overnight at the time we recorded, saying they're gonna they're looking to spend $850 billion in the next couple of years. So yeah, like these numbers are just keep going up. So
7:55Unidentified speaker (speaker_0): I guess the big question for investors then is like, how sustainable is this level of spending? And that might be an impossible question to answer at the moment. It might depend. on, you know, the value that businesses and individuals get from AI and how much we're willing to pay on AI subscriptions and, you know, all like just like as we sit here today with the information we have, like how sustainable is it? How many years can hyperscalers be spending trillions of dollars on data centers?
8:19Unidentified speaker (UNKNOWN): They're
8:20Unidentified speaker (speaker_3): not gonna be doing it every year, put it that way. Yeah, look, it's gone from them using internal capital, so kind of excess cas. cash that's sitting there, to raising money externally. So you've come you've seen that in them uh selling bonds. And then you've also recently seen Google and Meta actually raise capital through selling or issuing more equity. So there is a little little bit of concern, or I think it's just more of a a change that people have been okay with it once they that was it just internal. cash, and then now they're less comfortable that it's through debt. And we've kind of drawn parallels with like the the telecommunications rollout back in the dot-com, and people are echoing, is this going to be the same? But I think one distinction is
9:05Unidentified speaker (UNKNOWN): the
9:05Unidentified speaker (speaker_3): hyperscalers are, I would say the the trades or the companies of the last 10 years, these are enormous profitable companies that have got a big balance sheet and and can do this. I d I don't know what's gonna happen, but the market thinks this will be peaking in the next year or so. So this isn't something that's required every year to to deliver the the AI rollout. This is just one part of it and getting that kind of kind of the building and the infrastructure in place to really scale what they're trying to do.
9:36Unidentified speaker (speaker_0): Alphabet's the most profitable company in the world and I th we we saw a chart recently, it's got the second most cash on hand behind Berkshire Hathaway. So if any company is gonna spend this much, they're gonna
9:48Unidentified speaker (UNKNOWN): explain it. Yeah, yeah,
9:49Unidentified speaker (speaker_3): yeah.
9:50Unidentified speaker (speaker_0): So this is a very American story, and we're we'll get back to the American side of it. And I'm gonna ask you the big question: are we in a bubble? But I just want to like ground this in the c in context for Australian listeners. How relevant are we to this story? Like um should uh obviously there's been a few conversations about data centres in Australia, but like do you think about Australia at all in the the context of this this story?
10:12Unidentified speaker (speaker_3): Yeah, I actually think Australia can do really well here. So I mean historically Australia have done really well with having commercial property management companies doing really well. So like a Goodman, for example, um, there's a lot of um success in that. area and there's definitely going to be some crossover and then pivoting. So for example, Goodman, eighty percent of their future book is going to be more on the data center space. If we look at Australia in general, there's actually lots of um advantages of how we're set up to do well here. We've got a lot of land, which is which is great. So we can build them. There's a lot of renewable potential and that's important. And I'll I'll touch on that a little bit later when it comes to the power side. Our government is stable ish, I would say, um, which is important because you need strong data uh security. There's subsea uh cabling into Singapore. So then you you potentially got a huge market in Asia, which is a which is really fast and growing. In addition to that, you've got the some of the hyperscalers are already set up here. in terms of just like traditional cloud based um data centers.
11:20Unidentified speaker (UNKNOWN): So
11:20Unidentified speaker (speaker_3): there's lots of reasons why this could um be rolled out here and it's something which from an investment point of view and capital's already coming overseas into Australia for these reasons.
11:33Unidentified speaker (speaker_0): Yeah. So there there is some of this, you know, trillion dollars or whatever it is, some of that money
11:37Unidentified speaker (UNKNOWN): is actually
11:37Unidentified speaker (speaker_0): already flowing to Australia. the
11:38Unidentified speaker (UNKNOWN): street. Yeah, yeah.
11:39Unidentified speaker (speaker_3): So that's just not a US number, that's a global number.
11:41Unidentified speaker (speaker_0): So Australia's got potential to you know benefit from this uh and to capture some of this money. Um the hyperscalers are planning to spend a lot. Um if they continue to spend a lot, a lot of the sort of picks and shovel plays will be big beneficiaries. That's the setup. And then it leads to the ultimate trillion dollar question are we in a bubble? How long is this trade going to last? Where are we at in this story? So I'm just going to ask it straight up,
12:08Unidentified speaker (UNKNOWN): are
12:09Unidentified speaker (speaker_0): we in a bubble?
12:09Unidentified speaker (speaker_3): If I had to describe the the investing buzzword of this year, it would be bubble. And look, there's there's many parts of it that do seem bubble like. You might say I'm cherry picking here, but let's go back from COVID to now. Look at the S P 500.
12:24Unidentified speaker (UNKNOWN): I
12:25Unidentified speaker (speaker_3): don't know, it's up around 150. 160%. 120% of that is from earnings
12:31Unidentified speaker (speaker_2): growth.
12:32Unidentified speaker (speaker_3): When 15, 20% is dividends, and then the rest is that multiples expansion. So that kind of means like valuations are becoming more expensive. So, yes, valuations are higher than what they might have been historically, but not at the core because for me, a bubble is when the value of the asset drifts significantly from the underlying kind of cash flows and profits, like the the fair value or the intrinsic value. And yes, it might be slightly above,
13:04Unidentified speaker (UNKNOWN): but
13:04Unidentified speaker (speaker_3): I wouldn't say it's it's drifted so much that we're now kind of in bubble territory. I I know that people like to draw the comparison to like the telecommunications under. and and other scenarios where where we've been in in bubbles, but the companies behind it are in a lot stronger position than they were in other other market um cycles.
13:26Unidentified speaker (speaker_0): Yeah. There's this great chart that gets rolled out and it compares um the dot com bubble and uh today and it's just a line that has like valuations and earnings and earnings have followed uh valuations this time around very different in
13:41Unidentified speaker (UNKNOWN): dot
13:41Unidentified speaker (speaker_0): com
13:42Unidentified speaker (UNKNOWN): where
13:42Unidentified speaker (speaker_0): we were valuing companies by page views. Um the question then is like is it an earnings bubble and like how sus sustainable are these earnings and for me the question is like as I said earlier it's like how much value we all get from AI
13:56Unidentified speaker (UNKNOWN): and and
13:57Unidentified speaker (speaker_0): like where how much people are willing to spend on AI
13:59Unidentified speaker (UNKNOWN): and
14:00Unidentified speaker (speaker_0): sort of how that flows
14:01Unidentified speaker (UNKNOWN): back up
14:02Unidentified speaker (speaker_0): the uh pipeline.
14:03Unidentified speaker (speaker_3): Do you use it?
14:05Unidentified speaker (UNKNOWN): Yeah, yeah, all the time. Yeah, yeah, yeah. Yeah. Are you are you
14:08Unidentified speaker (speaker_3): OpenAI? Are you Claude? Are you
14:09Unidentified speaker (UNKNOWN): Gemini? But not
14:10Unidentified speaker (speaker_0): Gemini. Gemini's really dropped off for me. Yeah.
14:12Unidentified speaker (UNKNOWN): Yeah.
14:13Unidentified speaker (speaker_3): Yeah. Yeah.
14:13Unidentified speaker (speaker_0): So I guess like but like if if we say that it's about like how much value AI creates for society and that then like how much we're willing to pay for it, like what are you watching to sort of like what are the factors that would indicate it is creating enough value to justify its earnings and not
14:30Unidentified speaker (UNKNOWN): be a
14:30Unidentified speaker (speaker_0): bubble? Or vice versa, it's looking a little bubbly?
14:33Unidentified speaker (speaker_3): For me it's it all comes down to earnings, right? If we look at the market long term, everything is based off earnings growth. And what would concern me is if the numbers that the all the like the the kind of the AI labs, the hyperscalers, all the the projections are coming short in terms of their estimates.
14:55Unidentified speaker (UNKNOWN): And
14:55Unidentified speaker (speaker_3): you're not seeing that yet. Um if if we look at Anthropic, for example, last September their quarterly revenue was uh just under two billion. This year's one quarter, uh September quarter, and it was up it I think it increased tenfold. So I I get I get it I get why it feels bubbly because it's a new technology there's a lot of hype, but there's revenue being backed by it. I guess the concern for me, or uh one cr critique, is does it become a commodity? Because if it when it does, and like a model is just a standard, doesn't matter if you go Gemini, Claude, OpenAI, if it's all giving you the same answer, then then where does the pricing power go?
15:37Unidentified speaker (UNKNOWN): Because you
15:38Unidentified speaker (speaker_3): don't know who's gonna going to win.
15:39Unidentified speaker (UNKNOWN): And
15:39Unidentified speaker (speaker_3): then that's probably one of the reasons why we've gone down the this AI infrastructure way, because they get built regardless of whose model wins. Yeah. And they need to get paid up front to build the data the data centers before anything happens anyway.
15:53Unidentified speaker (UNKNOWN): From
15:54Unidentified speaker (speaker_3): our perspective, we're bullish on the long term, but we don't know who's going to gonna win and we don't know how it plays out. But we just know that this AI computational, like the models, they just need a huge amount of power.
16:04Unidentified speaker (speaker_0): Yeah, it's uh God, it's it's fascinating. The bears in this story often point to the circular financing deals there and they talk about you know how much NVIDIA is funding a lot of the neo clouds and you know how much debt some of these neo clouds are are raising. How concerned would you be about that part of the story? Um and I guess particularly you you know you're focused on the picks and shovels, the
16:26Unidentified speaker (UNKNOWN): racks
16:26Unidentified speaker (speaker_0): and the cooling systems
16:27Unidentified speaker (UNKNOWN): and all of
16:28Unidentified speaker (speaker_0): that. How concerned are you about these Neo Clouds getting in trouble?
16:32Unidentified speaker (speaker_3): Look, they're they're one part of the story. I wouldn't say they're the core. I would say that they've they've entered because there's been a AI or like the demand part of the story has has significantly um outstripped outstrip supply. And that's provided opportunities for companies like NeoClouds. to get financing from someone, buy the GPUs, put them in a data center, get at least the data center out, and then rent the capacity to when someone needs it. If we look at the Neo Cloud spending versus the overall picture, it's quite a small amount. There is internal, I call Of course, circular financing, so it would have knock-on effects if they they did kind of leave the market. But an interesting part of that financing is a lot of it's actually collateralized against the GPUs and also the contracts themselves. So if they did fall out, there would be it would be painful for shareholders, potentially bondholders. But I would say someone's gonna pick up those kind of that additional demand, whether it's a hyperscaler and they might actually it might be opportunistic and they could get a discount. on those contracts.
17:40Unidentified speaker (UNKNOWN): Yeah,
17:40Unidentified speaker (speaker_0): like Amazon or Alphabet stuff in. Yeah,
17:42Unidentified speaker (UNKNOWN): yeah,
17:42Unidentified speaker (speaker_0): yeah. It's a fascinating story. We're not gonna have the answer for it today, but yeah, I think what you said there around like uh everything follows earnings and as long as the the earnings story continues, this story will continue.
17:53Unidentified speaker (speaker_1): Yeah.
17:54Unidentified speaker (speaker_0): Let's turn to how a data center actually works because I think most of us now have a surface level understanding of it. Uh you need a hell of a a lot of electricity uh to power NVIDIA JPUs and that creates um comp oh that generates compute and and then we all use AI on the back of it.
18:13Unidentified speaker (speaker_3): Yeah.
18:14Unidentified speaker (speaker_0): Actually step us through in more detail like how how we should understand a data center and like some of the key components in there.
18:20Unidentified speaker (speaker_3): Yeah, yeah. An AI data center uses the same amount of power as
18:25Unidentified speaker (UNKNOWN): a
18:25Unidentified speaker (speaker_3): hundred thousand house households. So that's like the size of Hobart. And that's just one of the smaller ones. Like some in the US are considerably bigger. The power's essentially scarce, and it's not a case of look, I want to build a data center, I'm sure there's some spare electricity going around, I'll plug in from a neighbor, I'm sure he doesn't Don't mind. It's actually a really lengthy process.
18:48Unidentified speaker (UNKNOWN): And
18:48Unidentified speaker (speaker_3): if if you go through that, it's kind of applying to the local network. They kind of understand what you what you how much you want, whether that's going to impact the stability of the grid. They'll then come back and there'll be some negotiations of of how much the data center actually needs to pay to do the connection. There might be some other fees. And then it's potentially going gonna to go to the local council or government. They're going to review. gonna review it. And that's just the the approval process. Then if if it does go through, you're then looking at how to design the like the transformers, which is a big part of it because there's a huge amount of voltage that's coming from the grid to the data center. And then that has to be kind of like stepped down and transformed into the right power or voltage to then go into like the servers and
19:36Unidentified speaker (UNKNOWN): et cetera.
19:37Unidentified speaker (speaker_3): Building those can take 18 months or so. And in even in the U the US to kind of come out now and no one's getting any more power from the grid, I think it's beyond after twenty eight, twenty nine. So people are looking elsewhere to how how to generate power.
19:52Unidentified speaker (UNKNOWN): When
19:53Unidentified speaker (speaker_0): you like the government have said they're not getting any
19:55Unidentified speaker (UNKNOWN): But it's just not there.
19:55Unidentified speaker (speaker_3): It's just not there because for example in Australia, there's around 225 projects that are there for submission. And the amount of power that they require is double the peak demand of the entire grid. So they're not going to all get approved because it's just not there. And which is why the government, rightly so, has has pushed back and said, Well, you're not stealing or essentially not stealing all this electricity, and normal households potentially have to pay higher prices for you to make money. So what they've pushed. they push back and said, well, they actually need to be more potentially renewable and actually source your own power or approach like a hybrid approach of some from the from the grid, some um kind of like in not internally but on site.
20:40Unidentified speaker (UNKNOWN): Yeah,
20:40Unidentified speaker (speaker_0): yeah.
20:41Unidentified speaker (UNKNOWN): So
20:41Unidentified speaker (speaker_3): that's really a key part of it. And then another really, really interesting part is the cooling side. So the way in which it's changed is on the AI data center. side is the power consumption is 10 times as much as a like a traditional just data center. And I I think people don't appreciate actually how much heat comes out of these. So again, a a nice kind of comparison is one of these AI racks is let's say a size of a fridge. It's kicking out heat, which is equivalent to 55 fifty-five portable heaters coming out of that fridge.
21:15Unidentified speaker (UNKNOWN): Right?
21:15Unidentified speaker (speaker_3): So it's pretty hot.
21:16Unidentified speaker (speaker_0): Yes.
21:17Unidentified speaker (speaker_3): You somehow need to remove that heat. So before the AI side, uh computational side, it used to just be like air cons
21:25Unidentified speaker (UNKNOWN): and
21:25Unidentified speaker (speaker_3): j and chilled. But because there's now so much heat coming out, it's just not enough.
21:30Unidentified speaker (UNKNOWN): Um
21:30Unidentified speaker (speaker_3): so you've now seen that people have have gone to liquid cooling,
21:34Unidentified speaker (UNKNOWN): but
21:35Unidentified speaker (speaker_3): that's caused a lot of controversy because there's basically two ways to to do it. You you bring the cooling in, but you also this heat comes out and then it needs to go somewhere. And you've kind of got two ways to do it. You either
21:47Unidentified speaker (UNKNOWN): put
21:47Unidentified speaker (speaker_3): it into a tower and you let essentially kind of physics do do it. So the heat will go into the bottom of the cooling tower, which is like a big tower of water, and it will essentially evaporate into the atmosphere or out of the data center. And that's a lot more efficient because water absorbs heat four times as much as air.
22:08Unidentified speaker (UNKNOWN): So
22:09Unidentified speaker (speaker_3): it's a better way to do it. But in terms of water, is that they're constantly having to put water into this towel because it's evaporating. So in a year, it's estimated that it's going to be around one percent of Sydney water is it's would go into these. So that's why it's faced a lot of scrutiny. So then the other alternative. alternate alternative is a little bit like a a car um radiator, is the heat comes out of the cooling in like a pipe, and then you have essentially like cold air blowing on the pipes to cool them down, then it goes back through.
22:41Unidentified speaker (speaker_0): Yeah.
22:41Unidentified speaker (speaker_3): But that requires a lot more electricity.
22:45Unidentified speaker (UNKNOWN): So
22:45Unidentified speaker (speaker_3): that's cost that's that would then increase the costs of the data center between like around five percent. So you've got this trade-off of which one um which one to do. And for there's some environmental questions that have got to be asked there because it's it's quite important.
23:00Unidentified speaker (UNKNOWN): So
23:00Unidentified speaker (speaker_0): that's a lot about I guess like some of the key concerns around um data centers today, electricity and water usage. Yeah. If we actually go under the hood inside the shed, what's actually what do we need to know about what's happening in there in terms of like the servers and the racks and the GPUs and everything?
23:16Unidentified speaker (speaker_3): Yeah. So in terms of actually what happens is yeah you have a a large amount of power that comes to the the data center that then just kind of gets is set called step down or transformers is or basically just like reducing the the voltage of the electricity. And then they're going into into the the server through the racks into the GPUs um and then powering the GPUs to answer like our day-to-day kind of AI queries, unless we're kind of getting a bit more complex and people are running actual like genuine systems in their companies, for example, through AI. And that's that's 24-7. They're they're not going home at 5 p.m. They're constantly running. So it's quite um it's a very complicated process, but yeah, at a high level. Um uh power goes in, computation happens, heat goes out, and we need to get rid of that heat.
24:08Unidentified speaker (speaker_0): And then uh when it comes to like the you know the trillions that we were talking about earlier that's being spent on data centers, people often think about chips as like where the money
24:17Unidentified speaker (UNKNOWN): is
24:17Unidentified speaker (speaker_0): going. Um and NVIDIA's obviously become you know the most valuable company in the world as a result, but roughly half of the money actually doesn't go into chips. So can you guys give us give us give us a breakdown of like where the money actually flows?
24:29Unidentified speaker (UNKNOWN): Yeah,
24:30Unidentified speaker (speaker_3): yeah, yeah. So if $100 goes into AI CapEx, $50 yeah, it'll go into the chips. And then with that remaining $50, around $15 goes into networking. So that's kind of like cabling, switches, transceivers, things like that. $20 will go
24:49Unidentified speaker (UNKNOWN): uh
24:49Unidentified speaker (speaker_3): into power. So that's the grid connection, that's the generators, um, that's the backup energy supply, like batteries, things like that. Um, and then the remaining uh $15 is split between construction, so actually building uh the center, and then the the other half is through cooling. So, yeah, how do we how do we cool the chips down whilst they're whilst they're running?
25:13Unidentified speaker (speaker_0): Mm. And so then thinking about that pie chart and like where the money goes, when you were designing this ETF that you've just brought to market, what were like the pockets of spending that you you've focused on?
25:23Unidentified speaker (speaker_3): I think everyone's heard of Nvidia. Everyone it's done really well and I think it will carry on and do really well. Um but no one really talks about the boring stuff uh happens behind the scenes and what it all relies on. So yeah, we kind of want to put an ETF together to really focus on that. So like
25:40Unidentified speaker (UNKNOWN): the
25:41Unidentified speaker (speaker_3): fourth sub-themes of how we describe it is power and cooling, the AI racks, which are the to kind of um assemble them together. The data center REITs. So there's actually kind of structured as a fund, and the fund will actually invest in the property itself and buy the property, and it will receive money from. rent. And then the actual data center operators, so they're companies that actually get paid to manage the property day to day. So we think that gives a really nice kind of um broad um supply of like value chain of of the investment theme.
26:18Unidentified speaker (speaker_0): So why not chips?
26:20Unidentified speaker (speaker_3): Good question. Uh I just think that they've done they've done really well. But from a economics point of view, you haven't they're having to change. them every five years versus a building that's gonna last. I think Microsoft have have got it on their like on their from an accounting perspective as a long le long lease asset from they've increased it from fifteen to twenty-five years. We'd rather be in that space, the the kind of the hard assets, That's the core like components that need to be built before any of this happens. Again, no no, it's not really been covered that much either. And yeah, like I said at the start, is these the companies that are in our ETF, none of them are on the Nasdaq. Um, whether they're too small, whether they're in other countries. So you're actually getting something unique.
27:06Unidentified speaker (speaker_0): Yeah. So you said they're the boring parts of the ecosystem, but I don't think I uh I love finding out about companies that I didn't know before. I don't think it's boring. Give us a sense of like some of these companies that you're talking
27:17Unidentified speaker (UNKNOWN): about
27:17Unidentified speaker (speaker_0): and kind of like what they do.
27:19Unidentified speaker (speaker_3): Be surprised if people have heard of it, but uh one's called He Hon, it's in Taiwan. So people probably don't know, but Taiwan actually dominate in terms of AI racks and is again no one talks about it, but it's so important because they're the ones actually assembling and putting all the GPUs in the servers in the racks and combining them, and they have a partnership. with Nvidia. They're the main country to supply them. Like a little bit like obviously T T SMC is like the foundry for a lot of the chips. Taiwan seems to be very good at that manufacturing side of the GPUs. They also used to, or they still do, is they're a big manufacturer for Apple as well. So they already have that kind of infrastructure in play. There's a company called Vertiv, which again are, well, which are different. They are more on the power management side and third. thermal um regulation. So again, when them when they connect to the grid and the power comes through, they're responsible for like the step downs and etc. So that's more on on there. Um and then you've got companies like we've all heard of, like Goodman. We thought it'd be great to have some Australia representation in there, like um 80% of their their future kind of uh revenue or
28:26Unidentified speaker (UNKNOWN): um
28:27Unidentified speaker (speaker_3): prospects are in data centers. So I I think they're they're a good good one to be in. Um we also have a small part of next DC is in there. Not uh well it's a good thing I guess because we we've not launched the ETF. Well
28:39Unidentified speaker (UNKNOWN): even before the forty percent drop yeah yeah
28:42Unidentified speaker (speaker_3): but again they're they're a small part of the portfolio. No, It's uh it it the m the bigger biggest part really is the um is the power, the cooling the racks. That's that the the the core of the portfolio.
28:53Unidentified speaker (speaker_0): Would Fermos have made it in?
28:55Unidentified speaker (speaker_3): No, they wouldn't. No. So we so we've purposely excluded the neo clouds. Oh
28:59Unidentified speaker (speaker_0): okay, yeah. So
29:00Unidentified speaker (speaker_3): yeah, from our perspective,
29:02Unidentified speaker (UNKNOWN): look,
29:02Unidentified speaker (speaker_3): they're they're legitimate companies, but from from uh a representation of what the AI infrastructure is, we don't feel think it's appropriate. We see it as more of a essentially like a a financing layer
29:14Unidentified speaker (UNKNOWN): Yeah.
29:14Unidentified speaker (speaker_0): that they
29:14Unidentified speaker (speaker_3): they've just emerge because of excess demand. They're borrowing heavily, buying GPUs to then rent out. Whereas I don't I don't I don't see that personally or or the guys um in the team don't think that that should be a part of the AI pure infrastructure story.
29:30Unidentified speaker (speaker_0): Yeah, yeah. I mean like you think about someone like Iron, you know, they are building their own data centers. Like maybe at some point they're not considered neo clouds and they transition. But yeah, I understand what you're saying if you're just borrowing money from here and renting
29:44Unidentified speaker (UNKNOWN): over there. Yeah.
29:45Unidentified speaker (speaker_0): As we talk about how investors can approach this as a theme, a big um way that people have made money investing in AI has been focus on the bottlenecks. Yeah. Um and as those bottlenecks have kind of moved around, people have kind of moved money around and you know, like GPUs was the classic bottleneck and then there was other there's been others.
30:03Unidentified speaker (speaker_3): Yeah.
30:03Unidentified speaker (speaker_0): When we think about some of these like lesser discussed parts of a data center, server racks and stuff, uh is it similar there? Are there like is there a massive supply imbalance, um, long order backlogs um for some of these uh less discussed items?
30:18Unidentified speaker (speaker_3): Yeah, yeah. So I would say the biggest um bottleneck is power, is getting power from the grid. And if you're not able to do that, how else do you source it? Whether you go the renewables route, whether you go gas, potentially they're doing that in the US.
30:35Unidentified speaker (UNKNOWN): And
30:35Unidentified speaker (speaker_3): then also, as we said earlier, is especially in Australia, the the government scrutiny is essentially a bottleneck because it's slowing things down, because things are getting reviewed,
30:46Unidentified speaker (UNKNOWN): um,
30:46Unidentified speaker (speaker_3): because it's quite a new technology. We're where we're we're still struggling of what the right way to go. is obviously the
30:52Unidentified speaker (UNKNOWN): the
30:52Unidentified speaker (speaker_3): Lane Cove pullback from Goodman the scrutiny in Melbourne over the the power poles being too close to communities so there's a few bottlenecks and they they want to go now they want the power and the compute capacity right now but realistically it's a it's a two to two to four year process.
31:12Unidentified speaker (speaker_0): Yeah. Now as it's investors are thinking about the ways to invest in it, you know, we can obviously go some of the big Nasdaq uh listed hyperscalers. We can do an ETF like like the one you guys have brought to market. We can go individual stocks. Um, obviously you've got a dog in the race. Uh what how would you counsel investors to think about the different options and like weigh up the different options and which ones are right for different investors?
31:37Unidentified speaker (speaker_3): Yeah, there's lots of ways to to play. this theme. I think ETFs are great when
31:42Unidentified speaker (UNKNOWN): you
31:42Unidentified speaker (speaker_1): potentially don't have
31:44Unidentified speaker (speaker_3): you have a day job and you don't you can't spend hours upon hours researching. I mean the good thing about well AI can help now and assist but if you if you really like a theme but you're not 100% sure on which company's going to win then an ETF is a great way to do
31:59Unidentified speaker (UNKNOWN): it.
32:00Unidentified speaker (speaker_3): Also access so as a as a retail investor, you're not potentially able to invest in all countries. So I know most kind of broking platforms allow you to invest in Australian stocks, US stocks, which is great. But as I said earlier, a Taiwanese stock, probably not. We know we know firsthand how painful that process is is getting a trading account set up. So it's a good way to get access that you wouldn't be able to do yourself. But there are also some circumstances where it might not be the best fit for you. You might actually really like a certain company, but there's a part of the ETF which you don't want exposure to. Or the ETF actually might give you a double up in some of the underlying companies that you hold elsewhere. So yeah, there's there's lots of reasons to consider.
32:44Unidentified speaker (speaker_0): What about the decision between like the the big end of town, the hyperscalers and uh an ETF ATF like this? I
32:50Unidentified speaker (UNKNOWN): guess you
32:51Unidentified speaker (speaker_3): you need to have a view on where you think the growth is gonna be. Um for us we really like the concept of picks and shovels, hard assets, because they get built first. They get built regardless of whose model or whose chip is going to win. They're here for a while. Our product actually has some of the uh some of those REITs in which you get paid a steady or the the company itself gets paid rent from renting the data center out, so you get a nice steady income. So our ETF kind of provides a nice little balance between growth in a few of these areas, such as like the the AI racks, the power, but there's also some some income on the on the REITs and the operators.
33:31Unidentified speaker (speaker_0): Now, uh earlier I asked the biggest question we're getting about AI, which is are we in a bubble? I would say the second big biggest question we get is, Am I too late? So I'm gonna ask it uh to you. You know, like this story's Chat JPT launched what four years ago, a bit over four years ago now. Um this story has played out, a lot of money has been made, a lot of these companies have run very hard. Um, when you think about the story in 2026, are we too late?
33:59Unidentified speaker (speaker_3): Look,
33:59Unidentified speaker (speaker_1): our our index, which the funds tracking has done annualized
34:04Unidentified speaker (UNKNOWN): the
34:05Unidentified speaker (speaker_1): last five years, 35%
34:06Unidentified speaker (speaker_3): a
34:07Unidentified speaker (speaker_1): year, which is pretty amazing.
34:08Unidentified speaker (UNKNOWN): Why
34:09Unidentified speaker (speaker_1): didn't you release
34:09Unidentified speaker (UNKNOWN): the 685 years ago?
34:11Unidentified speaker (speaker_3): Obviously. Um, which doesn't mean it's going to continue. That's not how markets work. But yeah, a lot of these companies have have been beneficiaries of this
34:18Unidentified speaker (speaker_1): of
34:18Unidentified speaker (UNKNOWN): this
34:19Unidentified speaker (speaker_3): rollout. But the CapEx spend is increasing. So this isn't the end
34:24Unidentified speaker (UNKNOWN): of
34:24Unidentified speaker (speaker_3): the story here.
34:25Unidentified speaker (UNKNOWN): Um,
34:25Unidentified speaker (speaker_1): and we're
34:26Unidentified speaker (speaker_3): believers that. this has got a long way to run. And we're we're far from having the pure like everything set up for these hyperscalers to to really um scale their business. I think that they've they've run really run pretty hard. Um but I think yeah these these these companies or the ones in our ETF, Yeah. they're the ones where half of all this money is going.
34:48Unidentified speaker (speaker_0): And
34:48Unidentified speaker (speaker_3): that's continuing. That that there's contracts locked in for the next couple of years, which this is why the hyperscalers are building these these projects. Um, so that it's not a case of the taps are gonna get turned off anytime soon.
35:01Unidentified speaker (speaker_0): So for people who hear that answer, get excited, wanna
35:05Unidentified speaker (UNKNOWN): continue
35:05Unidentified speaker (speaker_0): riding this technology boom, but also want to remain sensible long term investors, what advice would you leave everyone here with today? Don't
35:14Unidentified speaker (speaker_3): get excited.
35:14Unidentified speaker (UNKNOWN): ruin one.
35:16Unidentified speaker (speaker_3): Um, look at this space. Um, this isn't gonna be a
35:20Unidentified speaker (UNKNOWN): a
35:21Unidentified speaker (speaker_3): s a replacement for, I don't know, a core holding like a VGS or a a VAS of the worlds. Obviously there's other ETFs. This is gonna be more on that kind of smaller allocation in your portfolio because it's more of a kind of a a growth,
35:35Unidentified speaker (UNKNOWN): um,
35:35Unidentified speaker (speaker_3): a pure growth position.
35:37Unidentified speaker (UNKNOWN): But
35:37Unidentified speaker (speaker_3): if you're investing over the next 10, 15 years, dollar cost average and take opportunities when when they come to you in terms of market pullbacks. Low prices are a good thing if you're if you're long term buying, but also don't think the world's going to explode just because the markets have pulled back 10%.
35:54Unidentified speaker (speaker_0): Yeah. Well, Will, it's a fascinating conversation. Uh it's great that you know products like this come to market and give us access to obscure Taiwanese companies that we know. haven't heard of before. So thank you for joining us today. Looking forward to watching this ETF playing out and I guess watching the broader AI infrastructure story continue to play out. Awesome. Thank you, V.
36:18Unidentified speaker (speaker_1): You've been listening to an Equity Made's media production. This podcast is intended for education and entertainment purposes. Any advice is general and advice only and has not taken into account your personal financial circumstances, needs, or objectives. Before acting on general advice, you should consider if it is relevant to your needs and read the relevant product disclosure statement. If you're unsure, please speak to a financial professional. The host of this podcast and their guests may have positions in the companies mentioned. Equity Mates Media operates under Australian Financial Services Licence 545. 540697.