Artemis Live - Insurance-linked securities (ILS), catastrophe bonds (cat bonds), reinsurance

Panelists debate whether ILS can fill capacity gaps for data center construction and operational risks amid AI boom

1h 1m Listen to the original Episode page
Insurance-Linked SecuritiesData CentersReinsuranceCatastrophe BondsAI InfrastructureRisk Management

Summary

Moderator Andy Palmer and panelists Michael Knecht, Niklaus Hilti, Mick Moloney, and Abderrahim Elamrani discuss the insurance capacity gap for data centers. They identify construction phase risks, business interruption, and cyber threats as areas where traditional insurers face constraints. Speakers argue that ILS offers diversification benefits as physical perils are uncorrelated with financial markets. However, challenges remain in standardizing risks and pricing them competitively against existing asset classes.

Episode sections

  1. 0:05 – 1:56

    Introduction and Market Scale

    Moderator Andy Palmer sets the scene with projections from McKinsey, the IEA, and Swiss Re Institute regarding the massive capital expenditure and energy consumption expected from data centers by 2030. He highlights the insurance capacity gap, citing a $14 billion Texas project and comments from AIG CEO Eric Anderson about limits maxing out the P&C industry.

  2. 1:56 – 3:16

    Panelist Introductions and Core Question

    Palmer introduces the panelists: Michael Knecht (Cambridge Associates), Niklaus Hilti (Euler ILS Partners), Mick Moloney (Oliver Wyman), and Abderrahim Elamrani (Descartes Underwriting). He poses a provocative question: given triple net leases and chip amortization, is there actually a capacity shortage or an insurance opportunity?

  3. 3:16 – 9:01

    Mick Moloney on Data Center Economics and Constraints

    Moloney details the capital stack of a gigawatt data center, noting the $10-15 billion shell cost, $20-25 billion in chips, and the critical role of Power Purchase Agreements (PPAs) with utilities. He explains the 'five nines' reliability requirement in leases and the scarcity of power infrastructure as a major constraint.

  4. 9:01 – 12:30

    Abderrahim Elamrani on Construction Insurance Gaps

    Elamrani identifies that construction projects are insufficiently insured due to size, NATCAT sub-limits, and poorly covered indirect losses like contingent business interruption. He notes that parametric NATCAT solutions are being deployed to address these gaps and are easier to repackage into ILS.

  5. 12:30 – 16:35

    Michael Knecht on Lender Requirements and Market Fluidity

    Knecht describes the market as fluid, with some projects fully financed and others scrambling for builder's risk coverage. He warns that while builder's risk is manageable, constraints may hit sooner in business interruption and chip obsolescence coverage, which are not yet fully priced or structured.

  6. 16:35 – 23:33

    Niklaus Hilti on Reinsurance Capacity and ILS Fit

    Hilti argues that insurers are doubling their books but struggling to find reinsurance for large, concentrated data center portfolios. He suggests ILS is needed for peak risk capacity, particularly for natural catastrophes where traditional insurers limit coverage, though he cautions against overestimating hurricane risk for inland facilities.

  7. 23:33 – 30:42

    Niklaus Hilti on ILS Investor Psychology and Structure

    Hilti emphasizes that ILS investors require clear, modelable risks and dislike surprises or uncertainty. He questions whether data center risks (cyber, business interruption, secondary perils) can be sliced into a format suitable for classical ILS investors who prefer transparent, diversified, multi-year contracts without hidden tail risks.

  8. 30:42 – 34:14

    Michael Knecht on Portfolio Exposure and Risk Transparency

    Knecht argues that investors are already massively exposed to data center risks through equities, utilities, and debt. He suggests the primary need is not just capacity but transparency to price these risks, noting that grid fluctuations and shared infrastructure failures create cat-like tail risks that are currently opaque.

  9. 34:14 – 37:16

    Correlation and Diversification Benefits

    Speakers discuss whether data center risks correlate with financial markets. One speaker argues that natural perils like fires and earthquakes are uncorrelated with stock market signals, while another notes that while stock prices may react to disasters, the impact is short-lived and does not trigger recessions.

  10. 37:16 – 39:40

    Risk Premium Extraction and Portfolio Exposure

    The panel discusses how institutional investors are already heavily exposed to AI and data center risks through equity and credit. They suggest that ILS allows these investors to extract a premium for specific physical risks (like fire) that are not priced into their existing investment holdings.

  11. 39:40 – 43:32

    Market Segmentation: Existing vs. Mega-Projects

    A speaker distinguishes between existing data centers, which have adequate traditional coverage, and new mega-projects. For new builds, traditional insurers struggle with capacity and construction phase risks, creating an opening for ILS solutions to provide supplemental capacity.

  12. 43:32 – 50:04

    Pricing Benchmarks and Return Hurdles

    Speakers debate whether returns are sufficient to attract capital. One speaker suggests ILS allocations should be priced comparably to US wind cat bonds to justify the diversification benefit, while others note that premiums for data centers are significantly higher than standard industrial property due to the risk profile.

  13. 50:04 – 57:34

    Space-Based Data Centers and Future Perils

    The moderator raises the topic of Google/Alphabet launching space-based data centers to utilize solar power. Panelists discuss the feasibility, noting that while space perils (solar storms, debris) exist, the latency issues make space data centers unsuitable for real-time applications, and insurance solutions for this are years away.

  14. 57:34 – 1:01:10

    Risk Slicing and Standardization Challenges

    The discussion turns to how to slice complex data center risks (property, cyber, liability) for ILS markets. Speakers agree that a lack of standardization and the difficulty of separating perils (e.g., cyber vs. property) is a major hurdle, though traditional property risks are already somewhat sliced.

Claims worth verifying

McKinsey estimates that data centers could require $6.7 trillion of investment by 2030.

Unidentified speaker (speaker_0) ·0:31

How to verify: Search for the McKinsey report 'The economic potential of generative AI' or subsequent data center infrastructure forecasts published by McKinsey & Company. For recent events, verify 2026-08-10 through 2026-11-10. Use another period only when the speaker explicitly names it.

The International Energy Agency expects data center electricity consumption to more than double from 2024 to 2030.

Unidentified speaker (speaker_0) ·0:41

How to verify: Check the International Energy Agency (IEA) report 'Data centres and data transmission networks' or their latest electricity demand forecasts for the AI sector. For recent events, verify 2026-08-10 through 2026-11-10. Use another period only when the speaker explicitly names it.

Swiss Re Institute estimates that capital spending by the five largest hyperscalers could exceed $600 billion in 2026.

Unidentified speaker (speaker_0) ·0:51

How to verify: Search for the Swiss Re Institute Sigma report or their specific publications on 'Data centres and AI' released around 2026-2026. For recent events, verify 2026-08-10 through 2026-11-10. Use another period only when the speaker explicitly names it.

AIG CEO Eric Anderson stated that insurance limits required for data centers are maxing out the P&C insurance industry.

Unidentified speaker (speaker_0) ·1:49

How to verify: Search for news articles or transcripts from 2026-2026 citing AIG CEO Eric Anderson's comments on data center insurance capacity. For recent events, verify 2026-08-10 through 2026-11-10. Use another period only when the speaker explicitly names it.

Primary source

Building a gigawatt of power for a data center requires a local utility to invest approximately $2.5 billion.

Unidentified speaker (speaker_0) ·0:41

How to verify: Cross-reference with utility capital expenditure reports or industry analyses on grid infrastructure costs per megawatt/gigawatt. For recent events, verify 2026-08-10 through 2026-11-10. Use another period only when the speaker explicitly names it.

Building the powered shell of a gigawatt data center costs between $10 to $15 billion.

Unidentified speaker (speaker_0) ·0:51

How to verify: Review construction cost benchmarks from real estate firms like JLL, CBRE, or specialized data center construction firms for 2026-2026. For recent events, verify 2026-08-10 through 2026-11-10. Use another period only when the speaker explicitly names it.

Chips in a data center amortize over four to six years.

Unidentified speaker (speaker_0) ·0:51

How to verify: Check accounting standards (ASC 360) or financial reports of major hyperscalers (e.g., Microsoft, Amazon, Google) for their depreciation schedules for compute equipment. For recent events, verify 2026-08-10 through 2026-11-10. Use another period only when the speaker explicitly names it.

Insurance companies are doubling the size of their data center insurance books roughly on an annual basis.

Unidentified speaker (speaker_3) ·18:36

How to verify: Analyze quarterly earnings reports or public disclosures from major specialty insurers (e.g., AIG, Chubb, Zurich) regarding their property/casualty or cyber insurance growth segments. For recent events, verify 2026-08-10 through 2026-11-10. Use another period only when the speaker explicitly names it.

One large insurance company has an annual budget for NATCAT risk of $1 billion.

Unidentified speaker (speaker_3) ·21:25

How to verify: This is a specific claim about an unnamed entity; verify by searching for public disclosures or industry reports identifying major insurers' reinsurance/NATCAT retention limits in 2026-2026. For recent events, verify 2026-08-10 through 2026-11-10. Use another period only when the speaker explicitly names it.

Speaker 1 claims that fires and earthquakes are not caused by financial market signals, and while stock prices may react to disasters, the effect typically washes out after 1-2 weeks and does not trigger a recession.

Unidentified speaker (speaker_1) ·35:50

How to verify: Check historical stock market data (e.g., S&P 500) against major natural disaster dates (e.g., 2026 Japan Earthquake) to measure duration of price impact and subsequent macroeconomic indicators (GDP growth) for recession signals. For recent events, verify 2026-08-10 through 2026-11-10. Use another period only when the speaker explicitly names it.

Speaker 1 states that a 1% fluctuation in the S&P 500 is significantly larger than the insurance losses discussed for data centers over the last decades.

Unidentified speaker (speaker_1) ·36:53

How to verify: Compare S&P 500 daily volatility data with aggregated industry loss data from sources like Swiss Re Sigma or Munich Re NatCatSERVICE for the relevant period. For recent events, verify 2026-08-10 through 2026-11-10. Use another period only when the speaker explicitly names it.

Speaker 2 claims that data center values have doubled or tripled in the past couple of years, making upgrades to existing sites more common than new builds.

Unidentified speaker (speaker_2) ·40:24

How to verify: Review real estate valuation reports from major data center REITs or commercial property indices (e.g., JLL, CBRE) for data center asset values between 2026-2026. For recent events, verify 2026-08-10 through 2026-11-10. Use another period only when the speaker explicitly names it.

Speaker 2 states that Descartes Underwriting has the only product licensed to cover a construction project on a parametric basis for earthquake in Japan.

Unidentified speaker (speaker_2) ·42:53

How to verify: Check the Financial Services Agency (FSA) Japan or local insurance regulatory filings for licensed parametric products and their scope of coverage for construction perils. For recent events, verify 2026-08-10 through 2026-11-10. Use another period only when the speaker explicitly names it.

Primary source

Speaker 3 states that typical institutional ILS allocations are between 2% and a maximum of 6-7% of the portfolio.

Unidentified speaker (UNKNOWN) ·43:37

How to verify: Review institutional investor surveys on alternative investments (e.g., Preqin, Cambridge Associates Global Alternative Investment Trends) for ILS/cat bond allocation percentages. For recent events, verify 2026-08-10 through 2026-11-10. Use another period only when the speaker explicitly names it.

Speaker 1 estimates primary insurance rates for data centers are 10-20 times higher per unit of risk compared to normal large industrial property.

Unidentified speaker (speaker_1) ·47:06

How to verify: Request sample insurance quotes or rate sheets from major property insurers (e.g., Chubb, Zurich) for industrial vs. data center properties, or review industry rate surveys from brokers (e.g., Aon, Willis Towers Watson). For recent events, verify 2026-08-10 through 2026-11-10. Use another period only when the speaker explicitly names it.

Speaker 1 claims pricing for data center ILS is roughly two to three points above the 2023 US Hurricane cap market rates.

Unidentified speaker (speaker_1) ·49:11

How to verify: Compare current ILS transaction pricing data (from brokers like Guy Carpenter or RenaissanceRe) for data center risks against historical 2026 US Hurricane cat bond issuance yields. For recent events, verify 2026-08-10 through 2026-11-10. Use another period only when the speaker explicitly names it.

Speaker 0 mentions Google/Alphabet is launching a pilot to push data centers into space next year (2027).

Unidentified speaker (speaker_0) ·50:27

How to verify: Check Google/Alphabet press releases or SEC filings (Form 8-K or 10-K) for announcements regarding space-based data center pilots scheduled for 2026. For recent events, verify 2026-08-10 through 2026-11-10. Use another period only when the speaker explicitly names it.

Primary source

Speaker 1 argues that cyber risk is often sublimited or excluded in data center property policies because the accumulation risk with cyber cap bonds could lead to double losses.

Unidentified speaker (speaker_1) ·59:58

How to verify: Review standard data center insurance policy wordings (e.g., from ISO or major carriers) for exclusions or sublimits on cyber perils, and compare with cyber cat bond model assumptions for correlated losses. For recent events, verify 2026-08-10 through 2026-11-10. Use another period only when the speaker explicitly names it.

Speaker 2 states that they provide accumulation control services for Generali, including accumulation with risks outside their portfolio.

Unidentified speaker (speaker_2) ·57:17

How to verify: Review Generali's investor presentations or press releases detailing their partnership with Descartes Underwriting and the scope of their risk management services. For recent events, verify 2026-08-10 through 2026-11-10. Use another period only when the speaker explicitly names it.

Do your own research

  • Verify McKinsey, IEA, and Swiss Re Institute projections on data center capex and energy consumption by 2030.
  • Review AIG's public filings or earnings calls for comments on P&C industry capacity limits.
  • Check Oliver Wyman or Cambridge Associates reports on data center capital stacks and PPA structures.
  • Analyze Descartes Underwriting's licensed parametric earthquake products in Japan for market penetration.
  • Compare current data center ILS pricing benchmarks against US wind cat bond rates.
  • Assess the correlation between natural disaster impacts and stock market performance using historical data.
  • Investigate the specific terms of 'five nines' reliability clauses in data center leases.
Full transcriptexpand
0:05
Unidentified speaker (speaker_0): Excellent. So maybe I propose to do just a quick introduction and background to the to the topic and then introduce the um the the panelists, I suppose, just to set the scene.
0:16
Unidentified speaker (UNKNOWN): So
0:16
Unidentified speaker (speaker_0): I don't think anyone in the room would be particularly surprised about the
0:23
Unidentified speaker (speaker_0): burgeoning amount of infrastructure development that we read. about and we understand from AI and cloud investment.
0:31
Unidentified speaker (speaker_0): McKinsey estimates that data centers could require $6.7 trillion of investment by 2030.
0:41
Unidentified speaker (speaker_0): The International Energy Agency expects their electricity consumption to more than double from 2024 to 2030.
0:50
Unidentified speaker (UNKNOWN): And
0:51
Unidentified speaker (speaker_0): Swiss reinstitute estimates that the capital spending by the five largest hyperscalers could exceed $600 billion in 2026.
1:02
Unidentified speaker (UNKNOWN): And
1:02
Unidentified speaker (speaker_0): I note, of course, Mick,
1:04
Unidentified speaker (speaker_0): you're sort of expecting that to be far exceeded even in 2027 as well.
1:09
Unidentified speaker (UNKNOWN): And
1:09
Unidentified speaker (speaker_0): the insurance challenge is equally striking. So the FT recently reported on a specific project in Texas, a $14 billion project, and describes specific limits of about half a billion dollars during construction and then when operational, alongside separate coverage for the likes of
1:29
Unidentified speaker (speaker_0): terrorism and liability. Really just illustrating that the gap can emerge, of course, between what's insured and then the actual true economic exposure. And recently the AIG CEO, Eric Anderson, called the build-out a great opportunity while saying that the limits required are absolutely maxing out the P and C
1:49
Unidentified speaker (speaker_0): insurance industry and pointed to the need spanning construction, cyber, property, and liability.
1:56
Unidentified speaker (UNKNOWN): For
1:56
Unidentified speaker (speaker_0): ILS, Unless, however, the demand is only the beginning. The existence of a protection cap does not gap does not itself make a risk suitable for investors. So we need to ask ourselves three practical questions. Can we clearly define and model the risk? Can it be appropriately structured? And is the
2:16
Unidentified speaker (speaker_0): return sufficient? So with that backdrop, I'm very pleased to present my panel today. We have Michael uh fr starting from here, Michael Necht, the Managing Director of Investment at Cambridge. Associates, Nicholas Hilte, the CEO of Euler ILS, Mick Maloney, partner at Oliver Wyman, and Abdurim El
2:35
Unidentified speaker (speaker_0): Renrani, sorry Abdrim, uh head of reinsurance at Descartes Underwriting. So maybe I can sort of ask this question somewhat provocatively. So someone described Describe to me that data centers are glorified warehouses with cooling systems. And inside the value of the chips are amortized down to zero in a
2:55
Unidentified speaker (speaker_0): matter of years.
2:57
Unidentified speaker (speaker_0): Meanwhile, you have triple net lease securitization structures that eliminate the insurance risk and the ultimate coverage provided by hyperscalers, meaning that maybe there isn't an insurance opportunity at all. So I ask you this: is there actually a capacity shortage?
3:16
Unidentified speaker (UNKNOWN): So
3:16
Unidentified speaker (speaker_0): I don't know whether Mick, you want to sort of jump in here or
3:20
Unidentified speaker (speaker_1): Yeah, I and I'll I'll take a couple of minutes uh to
3:22
Unidentified speaker (UNKNOWN): do this if
3:23
Unidentified speaker (speaker_1): if it's all right. Um
3:27
Unidentified speaker (speaker_1): and maybe just to to dimensionalize it for everybody in case we have folks in the audience who who aren't super close to the numbers, but e everybody talks about a gigawatt as a as a unit of scale for a data center.
3:39
Unidentified speaker (speaker_1): A gigawatt is enough power to uh to effectively power a city of about 700,000 people. Okay. When I'm building a gigawatt data center, and I'll use the US as a as an example, there isn't a spare gigawatt lying around, right? So what what I end up doing is I turn around to a local utility and I say, I want you
3:59
Unidentified speaker (speaker_1): to build me a gigawatt worth of power if I'm taking it through the grid. The local utility, in order to do that, is saying, I am regulated in the price I can charge my local household. So if you want me to build a gigawatt for you, I'm going to invest two and a half billion in building that. I want to know that you're going to pay me for
4:19
Unidentified speaker (speaker_1): that, whether you need the power or not. So it gets wrapped in what is called a power purchase agreement between the utility and the data center. Okay. That power purchase agreement oftentimes somebody looks at and goes, I have a capital stack sitting behind a data center, I'll get to the triple net lease in a moment, but I have a capital stack sitting behind the data center
4:39
Unidentified speaker (speaker_1): with an upstream liability to a hyperscaler, or I may not have an upstream liability to a hyperscaler. I the local utility, and we've seen this in a number of cases, say I want credit protection wrapped around the power purchase agreement. Okay, and that credit protection can be upwards of $10 billion in terms of what that looks like over a 20-year lease. I
4:59
Unidentified speaker (speaker_1): then have the data center itself, which is the the gigawatt, and this is the the um the powered shell uh piece of it. So a giga building a gigawatt data center, building the powered shell is about, you know, let's call it 10 to 15 billion dollars in investment on that on. that itself. I'm then putting 20 to 25 billion dollars worth of chips
5:19
Unidentified speaker (speaker_1): into it. Those chips amortize over, and again, the the chip amortization schedule is is something that we'll we'll talk about, but the chips amortize over, let's say, four to six years.
5:30
Unidentified speaker (UNKNOWN): So
5:31
Unidentified speaker (speaker_1): over a 20 to 25 year lease, I've got a hundred billion dollars worth of investment sitting in a data center. I I what I have sitting there behind the data center then is a capital stack that says that look, depending on who's putting the capital in, I'm either paying for the for the powered shell, or if I'm a neo cloud, I'm paying for the chips that are sitting in there
5:50
Unidentified speaker (speaker_1): as well. And it's up to me to suffer the economics on what happens. Okay. So what you have happened then is you have between the data center and somebody sitting upstream of it, you will often have a lease to a hyperscaler. Okay, and what the lease that this kind of triple net is effectively saying is that the hyperscaler is committing to cover all of the economics of the lease and
6:10
Unidentified speaker (speaker_1): give it downstream revenue stream there. The issue with with the these leases uh oftentimes have what they call the five nines in them, which is that I am committing to having reliability of 99.999% over the 20 or 25-year 25 year lease. If you look at that, that's less than 10 hours of downtime
6:30
Unidentified speaker (speaker_1): over that over that time horizon. So there there is no margin for error in the operation of the data center. If I trip into having downtime, then I'm I'm paying penalties, and I may actually hit a penalty that says that if you know I I I trigger a clause in the lease that would allow the hyperscaler to get out of the lease. Okay. Now if if
6:51
Unidentified speaker (speaker_1): you think about it, there are constraints lying all over this landscape. One one of the biggest constraints currently is the ability to get power on stream, okay? Because I if if I'm looking at it, it it takes three to five years for the local utility to build it. If I'm putting my own power behind it, there is a massive backlog on turbines, for example, on
7:10
Unidentified speaker (speaker_1): on gas generation. So th there is this assumption that power is going to be scarce for a significant period of time. If power is scarce and I'm sitting on, I am a hyperscaler, I'm going to suck any power I can get out of the system. So the idea that I would default on a lease and then cause a whole bunch of destruction in terms of the
7:30
Unidentified speaker (speaker_1): capital sitting downstream of that is a fairly remote possibility. But if you look at it, you can imagine lots of scenarios where I mean you just take the SpaceX IPO. if anybody read that, and you say, okay, Musk is saying he's going to put 100 gigawatts of power a year into uh into orbit uh you know over the next decade. So you can paint lots of scenarios where
7:50
Unidentified speaker (speaker_1): the the power piece goes away, but then look, if if I look at it, it is incredibly hard for me to imagine that there is not a very large opportunity for you know the folks represented in this room, whether it's ILS or sidecars, because the the this is $10 to $20 billion worth of premium arriving into the
8:10
Unidentified speaker (speaker_1): landscape over the a year over the course of the next five to ten years that is going to require a massive stack of capital to sit behind that, with massive amounts of concentration in terms of where these things are are occurring, and with very novel risks then in terms of that chain running from a part. power purchasing agreement into a lease sitting on top of a chip
8:30
Unidentified speaker (speaker_1): stack upstream of something that's very uncertain in terms of the the demand for AI at a certain price. So I uh very long-winded uh way of saying I can't tell you exactly what the opportunity is, but if we can't find an opportunity in that for uh for capital to flow in in the way it does through ILS or sidecars, I you know I don't know what uh
8:50
Unidentified speaker (speaker_1): you know. yeah, I I I I don't know what we would be doing.
8:53
Unidentified speaker (speaker_0): Hopefully we'll establish something here on today's panel. I don't know whether Abd or Nicholas you wanna jump in on the follow up.
9:01
Unidentified speaker (speaker_2): Um maybe um on a a couple of words on uh things you said. So um the the power and the reliability of uh the um energy supply is is a big issue and we see uh a lot of um
9:18
Unidentified speaker (UNKNOWN): like
9:19
Unidentified speaker (speaker_2): the we see the topic coming back a lot when it comes to construction projects. So construction projects for data centers are largely uh insufficiently insured today. Um there are various areas uh that are difficult to insure. ensure First, the sheer size of the investment, that's one thing.
9:37
Unidentified speaker (UNKNOWN): Um
9:38
Unidentified speaker (speaker_2): the second is
9:40
Unidentified speaker (speaker_2): very often a NAT cut sub limits.
9:45
Unidentified speaker (speaker_2): And
9:47
Unidentified speaker (speaker_2): the third is the indirect losses coming from the
9:55
Unidentified speaker (speaker_2): energy infrastructure that you can be exposed to. So usually these are contingent business interruptions interruption or non business. non-damaged business interruption kind of losses.
10:07
Unidentified speaker (speaker_2): And they are very poorly insured. So one way to bring in additional capacity that
10:15
Unidentified speaker (speaker_2): we are working on and that we uh deployed in uh in various uh instances is to uh is to look at different things. So first when it comes to the NATCAT experience. exposure, we express uh what's very often uh the PML peril uh being
10:35
Unidentified speaker (speaker_2): tornado or earthquake uh on a parametric basis. Uh this has different uh advantages for the corporate client that's uh looking for such protection. Um one is no litigation on business interruption. Uh
10:49
Unidentified speaker (UNKNOWN): um
10:50
Unidentified speaker (speaker_2): second is less discussion on the value of the chips. And third is the liquidity.
11:04
Unidentified speaker (speaker_2): And we provide a parametric offer for that. So it helps in two ways. So first that's driving the PML, and two, just again because of the sheer size of the investments, very very often there is insufficient demand. So any new capacity is welcome. And
11:23
Unidentified speaker (speaker_2): that's where we provide solutions. So linking to the ILS uh world, so by nature, uh parametric NATCA solutions are easier to uh
11:35
Unidentified speaker (UNKNOWN): uh
11:36
Unidentified speaker (speaker_2): to repackage into an ILS product and the traditional covers that uh uh we catch all risks that are out there.
11:44
Unidentified speaker (UNKNOWN): So
11:44
Unidentified speaker (speaker_0): so maybe let's come to the natural entry point for ILS in a in a in a second and sort of take a step back. So thank you obviously, Mick, for providing your your background and then after you sort of added to that with some some of your own experience.
12:00
Unidentified speaker (UNKNOWN): We to
12:01
Unidentified speaker (speaker_0): took we talk about this capacity gap. today. So what happens today in terms of the purchase of insurance? Is it just simply that investors, lenders are just happy to accept the risk today because it's simply not possible to purchase enough insurance? Is that is that today's scenario? How how does that play through,
12:21
Unidentified speaker (speaker_0): given that we know that lenders and many, many other environments simply would not accept that? um that is a condition.
12:30
Unidentified speaker (speaker_1): Yeah, I I mean I again th the the first thing I'd say is g given the scale of what we're describing here, there is just about every flavor of of situation happening. The the the the first piece, and again folks will be familiar with this, is and and you said it yourself, if if you look at the hyperscalers spending six, seven,
12:50
Unidentified speaker (speaker_1): eight hundred billion,
12:52
Unidentified speaker (UNKNOWN): a
12:52
Unidentified speaker (speaker_1): lot of that is they are financing these things and then operating them themselves. Up to now they have financed that from cash flow. They are only now beginning to finance it with debt or with equity. They are also then beginning to push it out into these leases. There there is a
13:12
Unidentified speaker (speaker_1): requirement.
13:22
Unidentified speaker (speaker_1): About where are capacity constraints and where will capacity constraints be going forward? Because if if you know if I talk to colleagues in Marsh Risk, it can be a scramble to find builders' risk coverage at at the levels that you you want.
13:38
Unidentified speaker (UNKNOWN): But
13:38
Unidentified speaker (speaker_1): it is possible to put it together, right? So that there isn't a view that capital behind builder's risk is a problem. I I do have to scramble to put you know the limits that I want to get sometimes together. We but what you need to remember is if you take that six or seven trillion dollars that's gonna flow into the landscape, let's call it a trillion has been committed
13:58
Unidentified speaker (speaker_1): already, right? So
14:00
Unidentified speaker (UNKNOWN): what
14:01
Unidentified speaker (speaker_1): what what's not clear to me is whether we will still think that's the case in a year's time or two years' time, given some of the concentrations that are building up behind these portfolios. And the other thing is a lot of these campuses are not yet operational. So the problem around business interruption has not yet come on stream. And I look at the commitments around business
14:20
Unidentified speaker (speaker_1): interruption, given some of the the kind of contingencies that are operating behind it, and I I think that is a place we will hit constraints maybe sooner than we'll hit constraints on the aggregate kind of builder's risk piece.
14:33
Unidentified speaker (UNKNOWN): And
14:33
Unidentified speaker (speaker_1): the piece that isn't even there yet is anything around chip obsolescence, right? And and what's what's sitting what's sitting behind that. So I I would say, is it apparently a problem where cover can't exist today, not yet. Now your your your question though was on what are lenders requiring kind of behind these
14:52
Unidentified speaker (speaker_1): portfolios. Lenders are requiring uh are requiring cover. It it you know I again you know and I I you know I I look at it and I go, this thing is moving so fast. I was having a conversation with with one of our brokers yesterday who was describing a situation where it was twenty, 20-30% thirty percent into the build uh on this
15:12
Unidentified speaker (speaker_1): and somebody had just figured out that they didn't have cover in place, and it was okay, hey, we need to turn around and and give me that certificate now. So th there is a a very fluid landscape sitting behind it, which goes from yes, covers are covers are are are kind of getting in place through to others are
15:32
Unidentified speaker (speaker_1): financing the thing. And
15:35
Unidentified speaker (UNKNOWN): you
15:35
Unidentified speaker (speaker_1): know, I I also look at it and go, I I wish worry a little bit from from purely from a um
15:43
Unidentified speaker (UNKNOWN): you know
15:43
Unidentified speaker (speaker_1): a financing standpoint, which is you know, are the are risks being realized and legible across the landscape properly yet? And I'm I'm not sure they
15:54
Unidentified speaker (UNKNOWN): are.
15:56
Unidentified speaker (UNKNOWN): Thank
15:56
Unidentified speaker (speaker_0): you for that. Um so men maybe then taking the conversation forward. So we've talked about possible
16:03
Unidentified speaker (speaker_0): possible capacity shortages across every single form of investment, every single form of risk, because it spans pretty much everything. I sort of want to get into the heart of the discussion, which is where where does ILS come into this? Where does ILS help? Now, after talked about obviously a parametric cover, particularly for BI.
16:24
Unidentified speaker (speaker_0): Niklaus, what's what's your view? Is it more NATCAT capacity? We know that ILS does that very well. Is there something beyond that?
16:33
Unidentified speaker (speaker_0): How would you approach that?
16:35
Unidentified speaker (speaker_3): I think we we've been a little bit busy this summer in looking into the the topic and and I think when you
16:46
Unidentified speaker (UNKNOWN): I
16:46
Unidentified speaker (speaker_3): think you dis described it
16:49
Unidentified speaker (UNKNOWN): if
16:49
Unidentified speaker (speaker_3): you look at the display the speed this is coming and on all scales
16:56
Unidentified speaker (speaker_3): it probably starts with the with the insurance market and the insurance market of course was insuring data centers
17:05
Unidentified speaker (speaker_3): probably more than a decade but I think the new element is that you have a huge pipeline
17:15
Unidentified speaker (speaker_3): of data centers being built where the question arises: is there enough capacity for that? That's one thing. The second thing is you should not forget some of the data centers are
17:30
Unidentified speaker (speaker_3): refitted. They basically get new chips
17:34
Unidentified speaker (speaker_3): into an existing data center, and just to give you an idea, we looked into that.
17:41
Unidentified speaker (speaker_3): We did the analysis. If you look at the square meter of
17:45
Unidentified speaker (speaker_3): a typical cloud data center or warehouse,
17:51
Unidentified speaker (speaker_3): and you look then at the modern AI
17:55
Unidentified speaker (speaker_3): data center square meter cost, it's easily between 10 to sometimes 30 times. the cost per square meter. And and what that means is for an insurer as well is maybe they gave insurance, but now they are facing new values and new
18:15
Unidentified speaker (speaker_3): demand on providing larger limits. And that is what we are seeing at the moment is that insurance companies they the the specialized insurance companies, I think they are around doubling doubling the size of their book roughly
18:36
Unidentified speaker (speaker_3): on an annual basis. This is what we what we are seeing in terms of premiums.
18:42
Unidentified speaker (speaker_3): And because it's direct insurance, you can think about the liability side being um
18:50
Unidentified speaker (speaker_3): much larger. So
18:53
Unidentified speaker (speaker_3): it's a gigantic wave which is coming in here. On the insurance side, they are growing, but they need, for example, the reinsurance market to grow with them. And reinsurers, they are or they were happy insuring reinsuring that risk within their normal
19:13
Unidentified speaker (speaker_3): treaties when these were small data centers or smaller data centers. Today, with these values,
19:21
Unidentified speaker (speaker_3): that doesn't fit anymore. Is you have you have a broad industrial or heavy commercial book which is protected, and all of a sudden you have these maximum lines across hundreds of risks, and I think that is where we see. that the insurers now
19:41
Unidentified speaker (speaker_3): are struggling to get reinsurance on data sent to specific portfolios. They need the reinsurers to get their head around that. And with the reinsurers, the natural reaction is that they look up to the IS IRS market and say,
20:01
Unidentified speaker (speaker_3): basically, we need big capacity to absorb those type of shock losses.
20:09
Unidentified speaker (speaker_3): That is what the capital market basically should do or should be involved in. And I think this is really what we're seeing. We see it on both.
20:21
Unidentified speaker (speaker_3): Side.
20:23
Unidentified speaker (speaker_3): There I would say we shouldn't overestimate the NAT NATCAT component. I think usually these data centers are not built on the beach, so Hurricane Typhoon should be less of an issue, not zero, but less.
20:41
Unidentified speaker (speaker_3): Usually the insurers, similar to cyber, they are limiting how much they want to pay. So maybe their clients are not getting full cover of earthquake, for example. So that is, I guess, why they're reaching out, for example, to you on the
21:00
Unidentified speaker (speaker_3): parametric side. So I would say there is
21:05
Unidentified speaker (speaker_3): definitely demand for natural catastrophe cover from the ILS industry, but the risk component as well. And just to put it into perspective, one of the larger insurance companies, they have an annual budget for NATCAT risk of a billion.
21:25
Unidentified speaker (speaker_3): And you can imagine that they're probably not going to put down a net line of a billion for one data center.
21:34
Unidentified speaker (speaker_3): I guess they they have smaller net uh limits on on data centers on the per risk basis. And I think that is really where the opportunity I would say arises for the capital market on a single per risk side, but also on the NATCAD side, because that cover is
21:54
Unidentified speaker (speaker_3): maybe not that available from the insurance market.
21:59
Unidentified speaker (UNKNOWN): So
22:00
Unidentified speaker (speaker_0): be b f before I come to Michael, maybe just a follow up to you. Niklaus. Um when we were preparing for this, you also, I suppose, drew an analogy to some of the existing chemical factories and other large infrastructure that exists today where ILS obviously does play a role.
22:17
Unidentified speaker (speaker_3): Yeah. Warehouses.
22:19
Unidentified speaker (speaker_3): As well.
22:20
Unidentified speaker (speaker_3): Um definitely I think it's I think it's a it's a good comparison is uh um it is to a certain degree a warehouse with extremely expensive um expensive equipment inside or or materials and and and it already exists so it's
22:40
Unidentified speaker (speaker_3): not it's not something which is not not being covered if you look at the large petrochemical plants which are inshore insured today they are all also in the region, I guess, of the 10 billion size, highly complex. They need cooling as well, some of them. And so you have a lot of
23:00
Unidentified speaker (speaker_3): elements which are similar in the existing market on the heavy industrial, um, industrial side. So uh from that perspective, it's nothing really new for the insurance market. What is new is the speed, the scale,
23:18
Unidentified speaker (speaker_3): the scale of it, I would say.
23:22
Unidentified speaker (speaker_0): So maybe then coming to Michael from where you sit in the market
23:28
Unidentified speaker (speaker_0): and the institutional investor view on this. What's your take?
23:33
Unidentified speaker (UNKNOWN): It's
23:33
Unidentified speaker (speaker_1): quite funny. I just had that
23:34
Unidentified speaker (speaker_3): thought and to imagine myself sitting into a pension plan and trying to sell that.
23:39
Unidentified speaker (speaker_3): And um
23:40
Unidentified speaker (UNKNOWN): um
23:41
Unidentified speaker (speaker_3): I was not I was not flirting with anyone and I took myself some notes and one of the notes I took myself is how do we fit this data center opportunity into the ILS bottle? Because the ILS bottle has a has a history and it has specific needs. You know, you mentioned multi multi-year year contracts,
24:01
Unidentified speaker (speaker_3): you m you will list listed it down a lot of different risks and a lot of different pockets. So if I come from the ILS perspective and where we are coming from, ILS alternative risk transfer, mainly having the focus on risk and not uncertainty. So we are collecting a risk premium for something
24:21
Unidentified speaker (speaker_3): which we can model, we can understand, and that's what we can easily put into an investor's portfolio. also understanding the diversification benefits and its liquidity need. So important thing, and this is always the challenge between reinsurance market and asset management, we do not have a balance sheet. Right? So we
24:41
Unidentified speaker (speaker_3): do have to return back the money. So all the risks which you mentioned and listed down is here, it's really like how do we
24:49
Unidentified speaker (speaker_3): you know. for me, not every supply, every supply gap sounds like an opportunity, but not every opportunity is for us.
24:58
Unidentified speaker (speaker_3): And so what I would have to need is to understand the risks which are involved. And please let me spend one or two seconds here. Um, if I say the ILS market has its history, what people can't stand is surprises. Surprises coming from uncertainty. from risks they don't understand or they did not know that they are
25:18
Unidentified speaker (speaker_3): caring. And the worst case is for risks that they have not been paid for.
25:23
Unidentified speaker (UNKNOWN): I'll
25:23
Unidentified speaker (speaker_3): give you a classic example: COVID, PC business. Some of the contracts have been triggered. So here we have a variety. I understand the capital need. We have a P and C book, we have a P and C business piece here, and just given where. the the data centers are, this is reintroducing second tier
25:43
Unidentified speaker (speaker_3): perils, floods, wildfires, hail, everything which we try to squeeze out because we had a lot of volatility, you guys try to bring in with the data center now. We have questions like on the man might side business interruption. Um we have maybe cyber, you can tell me I'm not expert the here. How much
26:03
Unidentified speaker (speaker_3): cyber we introduce here? So there's potentially a lot of noise, which we normally try to cut out because this has been spooking a lot of investors away. So the question really is: I do fully understand where you guys are coming from. I fully understand the need, and the capital market is here to address peak risk. That's something that there we agree at all.
26:23
Unidentified speaker (speaker_3): But how can we slice and dice this into a form that's a part that's apparently suitable for the classical ILS investor? Does it have to come in a multi year Lloyd structure? Does it have to come in a closed end format? I mean has it has it come in a cap on format? And how can we
26:43
Unidentified speaker (speaker_3): address the risk transparency point to give a lot um of comfort to the underlying investors that we're not ending up in the bin there dumb that so this is not very constructive well it's just kind of like I try to imagine how to explain that to people who have already been heard with
27:02
Unidentified speaker (speaker_3): uncertainty and they ask themselves is it's really worth it? So a lot of questions back then sorry.
27:09
Unidentified speaker (speaker_0): Yeah so I'm gonna bounce around one straight throw it back to you.
27:11
Unidentified speaker (UNKNOWN): Okay.
27:12
Unidentified speaker (speaker_0): So that's so so of course the the premise that's that so so the I ILS market is there for for peak risks. I think indeed it the role that it plays is for peak risks it adds that additional capacity in genuine scale to support the existing market and and the risk taking. It does it very, very well and very efficiently and we've seen that play out over time. And
27:31
Unidentified speaker (speaker_0): you're absolutely right. What what spooks investors indeed is just uh surprises uncertainty. It's fine if you're paid for something that you understand. It's not fine if either you didn't understand it, weren't paid, or didn't know about it, right? But here, from the sounds of it, the size of the
27:51
Unidentified speaker (speaker_0): opportunity, if we can call it that, is so large that it is the peak. It will become the peak, meaning that you talked about some of the perils which we're exposed to, probably the main one, so severe. convective storm tornado risk in the regions where a lot of this is being constructed, suddenly becomes a peak risk. So then does it not fall on
28:11
Unidentified speaker (speaker_0): the ILS market to address that and say, okay, this is going to be the new peak or one of the new peaks? I need to find a way to provide that capacity to the market because quite frankly, that's where all the economic development is.
28:27
Unidentified speaker (speaker_3): It's an interesting one. Um I'm I'm glad I didn't interrupt you. I wanted to interrupt you and give you an example. But I mean for us it's super easy if an earthquake in San Francisco happens, you have uh minus 20%. Everyone is prepared for that. The difficult thing over the last 10 years, which we since we did that, uh since we are messing in ILS, is
28:47
Unidentified speaker (speaker_3): like, why are we down 75 seventy five basis points in July? Nothing happened? And then I would have to go back and say, Oh, you know what, there was this hacking attack on one of the data centers and you know that caused the fire and this ten billion data center burned down and um we have a five hundred million loss of one of the contracts, and that's the reason why you're down. And then the question
29:07
Unidentified speaker (speaker_3): immediately is, why is that in my ILS portfolio? It's just from the mindset. I'm not sure if the investor the journey you made now, is absolutely logical. But the way how ILS is understood in the institutional investor
29:25
Unidentified speaker (speaker_3): landscape, I don't know if people will be ready for that.
29:29
Unidentified speaker (UNKNOWN): I
29:29
Unidentified speaker (speaker_3): don't know. I ask myself if you go into a hedge fund strategy, I mean some of the big multi-threat Multi strat hedge funds, they're super active in the ILS space. I don't know if that's an extension of a reinsurance business. extension of a hedge fund business, where they feel comfortable also dealing with that uncertainty. I'm not sure if that classical cap on peak risk ILS investor is the
29:49
Unidentified speaker (speaker_3): right person to deal with risks like that. But maybe I'm just too old and too conservative.
29:54
Unidentified speaker (UNKNOWN): I mean
29:55
Unidentified speaker (speaker_0): you we we can talk about it in a in a second because you also talked about the correct format of the investment.
30:00
Unidentified speaker (speaker_3): Yeah.
30:01
Unidentified speaker (speaker_0): Which one can imagine.
30:04
Unidentified speaker (speaker_0): investment structures that clearly isolate and say, look, this is data center, so you need to get into it, and we'll we'll talk about that. But one of the concerns, I suppose, that I think you're is the speed, it's simply the speed, because educating investors is a m multi-year, if not you know, multi-decade
30:24
Unidentified speaker (speaker_0): job, as you know. It takes a lot long, long time for people to understand what this market is about and what drives it. And I think, Niklaus, you pointed to one of the concerns here is not just the size, it's the speed at which
30:37
Unidentified speaker (speaker_0): this is emerging.
30:40
Unidentified speaker (speaker_1): Just a a um
30:42
Unidentified speaker (UNKNOWN): uh
30:42
Unidentified speaker (speaker_1): a a point to see if everybody agrees with it as as we're talking about it here. I mean, I I see a fundamental difference between
30:51
Unidentified speaker (UNKNOWN): cat
30:51
Unidentified speaker (speaker_1): risk structured through ILS and the risks we're talking about here, right? The the you know the there isn't an investment portfolio that I'm aware of that isn't already exposed massively to the things we're talking about here, right? The if if I'm indexing to US equities, I I have got the hyperscaler risk
31:11
Unidentified speaker (speaker_1): like sitting front and center, right? If I'm investing in utilities, I've I've got this risk sitting in there. If I have investment grade debt coming onto onto a balance sheet, chances are a lot of that investment grade debt going forward is going to be sitting be behind these data centers. So I I have this risk in almost all portfolios that I'm looking at in
31:31
Unidentified speaker (speaker_1): a way that isn't the case with hurricane risk, for example, on on other things or anywhere near the same extent. So
31:38
Unidentified speaker (UNKNOWN): you're
31:38
Unidentified speaker (speaker_3): asking
31:44
Unidentified speaker (speaker_1): Yeah, but but I but I I I think there are a couple of points there,
31:47
Unidentified speaker (UNKNOWN): right?
31:48
Unidentified speaker (speaker_1): W w one is I I think there and I'm not saying this is an answer, but th there is a need for something that gives transparency into the To price of the risk. Whether I am choosing to offset it or not, I don't have a means of judging what that risk looks like today and quantifying it across a portfolio.
32:07
Unidentified speaker (speaker_1): Your point then is then I face a choice.
32:11
Unidentified speaker (speaker_1): Am I in too much of it? Am I a good counterparty for doing something else outside of it? And I I maybe I'll make one more point and hand it back to you guys. Like uh for me it's helpful to abstract up from ILS and come back to the conversation that I think is at the core of this, which is are there natural counterparties for this that are happy to hold some of it and be paid a
32:30
Unidentified speaker (speaker_1): premium for it? Right? And and I think the the features of what we're talking about here that are kind of parallel to other things is have I got a catastrophic event that can lead to a tail risk? That I have a set of parties holding that tail tailorish risk today that are willing to pay to get rid of it, and a the set of counterparties that are willing to accept a premium and at
32:51
Unidentified speaker (speaker_1): a level they think are. And I look at that and I go,
32:54
Unidentified speaker (UNKNOWN): the
32:54
Unidentified speaker (speaker_1): concentration risks building up in here I think are I mean the the warehouse is a good parallel, but the the the the consequence of a power failure to to one of these things that extends beyond whatever the a kind of an emergency thing is, are huge. And there are all sorts of interesting things beginning to happen with with grids.
33:15
Unidentified speaker (speaker_1): and utilities in the US, where there was a recent example where a fluctuation in the grid got all of the data centers in the area to shut down and go to remote power in a way that then caused a big problem for the grid itself, in terms that the problem was fixed within two minutes, but they didn't come on stream for a period of time. It causes
33:34
Unidentified speaker (speaker_1): massive fluctuations in the grid itself. So that there are all sorts. of risks accumulating in the system that look like cat risks. It's not a hurricane, it's not a wind, it's not a fire, but it's it's fluctuations in a shared infrastructure that if something goes wrong is catastrophic. Yeah. And I so I I I
33:54
Unidentified speaker (speaker_1): look at it and I go, look, our job, I think, is is providing legibility, facilitating a market in risk transfer, and then it i finding finding an investors that that that are are natural counterparties for it. Those characteristics I think this has. And I I'm not I'm not disagreeing with anything you're saying in terms of, hey, I I don't want to double down on risk. I I think in
34:14
Unidentified speaker (speaker_1): lots of cases you go, if a pension fund is sitting on an equity portfolio and an investment great debt portfolio, chances are it's wrong way risk to begin to accept more of this in a way that they might be happy to do with with ILS. But I I think the fact that you know there are these kinds of risks, capital will need to come in behind them if it's going to flow and get and kind of get diversified through the
34:34
Unidentified speaker (speaker_1): system. I mean that that feels you know adjacent to our our current job.
34:40
Unidentified speaker (speaker_0): So there's a couple of points sort of I want to draw out here. So so one of them is effectively the correlation question because it clearly there is some some degree of correlation between these assets and the the wider financial markets, which you point out it's exposed pretty much across every single kind of
35:00
Unidentified speaker (speaker_0): investment you can possibly imagine. So maybe we'll we'll talk to you that shortly. The other one which you sort of allude to the fact that effectively this becomes critical infrastructure, or already is critical infrastructure, which clearly is in most people's minds not an insurable risk because of the size, the scale, and the
35:19
Unidentified speaker (speaker_0): nature of what it represents. Maybe on the first one, the correlation point. We discussed this before, Nikolaas. I don't know if you had a particular perspective on
35:29
Unidentified speaker (UNKNOWN): that.
35:30
Unidentified speaker (speaker_1): I think correlation for us always is a flow of information in two directions, and that means let's say it's stock market, there's an issue, and then information flows over to another asset class, asset class reacts, correlates.
35:49
Unidentified speaker (UNKNOWN): Um
35:50
Unidentified speaker (speaker_1): so this direct direction I think we we don't have here. Um
35:55
Unidentified speaker (speaker_1): because fires, earthquakes are not caused by financial market um signals. The other way around, it's also something we have seen over the last 20 years.
36:11
Unidentified speaker (speaker_1): Or there is some information flowing the other direction. If you have a big earthquake in Japan, yes,
36:20
Unidentified speaker (speaker_1): certain companies will be affected and you will see it in their stock price or even the entire market is reacting. But I think what we've seen is it's not triggering a recession. So typically these
36:36
Unidentified speaker (speaker_1): signals they have an effect for a day, for a week, maybe for two weeks, but typically they wash out Um after after that because other things are more important. We shouldn't forget it's also. a different scale of the market.
36:53
Unidentified speaker (speaker_1): Fluctuation on on I don't know the SP five hundred of one percent is probably significantly larger than the losses we have been talking here on the insurance side over the last decades. So so I think from that perspective, I think from our perspective
37:13
Unidentified speaker (speaker_1): it is uncorrupted. uncorrelated.
37:16
Unidentified speaker (speaker_1): Another aspect, I think, going back to
37:20
Unidentified speaker (speaker_1): that element is it is uncorrelated, but it's also a different way of extracting a return. So we see and agree with you, there is probably no institutional investor which is not exposed to AI data center and the whole
37:39
Unidentified speaker (speaker_1): technology sector and what we are seeing as a development, whether they are through credit funds on the equity side, anyhow. But if you look at the list of risk they are running with these investments, they're already holding in a significant way.
37:59
Unidentified speaker (speaker_1): Is we heard about chip depreciation.
38:04
Unidentified speaker (speaker_1): And there it is a massive question whether if if Elon Musk is faster
38:11
Unidentified speaker (speaker_1): basically bringing the stuff into space or whatever, there's a lot of uncertainty around the business case of some of these investments. And what I think is the interesting thing, and this is also where we see a lot of investor appetite. is
38:28
Unidentified speaker (speaker_1): we give them we give them the risk of a warehouse burning down fundamentally.
38:36
Unidentified speaker (speaker_1): So they can extract a premium from something which is a very small list as compared to what they're already holding on on their investment side. It's very specific and they can extract a very attractive premium um if they compare that to what they
38:56
Unidentified speaker (speaker_1): are getting on their investments in certain cases um with a lot of risk involved.
39:02
Unidentified speaker (UNKNOWN): So it
39:02
Unidentified speaker (speaker_0): in a
39:03
Unidentified speaker (UNKNOWN): different
39:03
Unidentified speaker (speaker_1): aspect.
39:04
Unidentified speaker (speaker_0): Yeah. So in in a sense what you're saying is of course those other investments simply are not pricing in this element of the risk which alludes to what Mick was getting, you know, you're getting at as well. The lack of transparency means it's sort of forgotten amongst the overall investment thesis in AI and data center and
39:24
Unidentified speaker (speaker_0): so on and so forth. Can I ask them maybe we'll come to to Abda?
39:28
Unidentified speaker (speaker_0): Is the return sufficient today for what you're seeing? Do you think it, you know, does it need to actually increase to attract capital in to make Niklaus's argument whole? hold true?
39:40
Unidentified speaker (speaker_2): Yeah, so uh
39:42
Unidentified speaker (speaker_2): definitively so we have data centers in our portfolio and the returns are adequate, very adequate. Um so uh um the premium is commensurate to the NATCAT risk that we uh that we cover. And our products uh so we cover NATCAT, we pay we price and structure.
40:03
Unidentified speaker (speaker_2): coverage for each peril individually, and therefore we have full control of the perils that are covered. And um and I would uh so I would tend to distinguish two things. So there is uh there are the um there is the infrastructure which is already out there. Um so the stuff that is running and that's been running for years now. So
40:24
Unidentified speaker (speaker_2): um so this was covered by traditional. players, still covered today. And what happened is that in the past couple of years, the values doubled, sometimes tripled, on these sites. So it's much easier to take a site that's operating already, to put in there the newest chips, or actually to add a couple
40:43
Unidentified speaker (speaker_2): of floors uh than to build like a 10 ten uh uh 10 ten billion uh infrastructure and uh data center. So I would tend to distinguish the two things. So the first the the ones that are up and running, they've been there, they've been covered. So we we provide their uh additional capacity uh for the pig
41:03
Unidentified speaker (speaker_2): perils because uh uh there was sufficient capacity with traditional players in the past and the stuff got too big for the the incumbents and they they they built um like some some a smarter kind of coverage or a more efficient coverage and that's where we we provide um as I said
41:23
Unidentified speaker (speaker_2): earlier tornado earthquake um kind of coverage and I would distinguish this uh to the construct uh the newer kind of projects the mega megaprojects projects as for these things they've been running they operate with an insurance coverage that's uh that's more or less covering uh the full uh
41:43
Unidentified speaker (speaker_2): PML exposure uh that used to be the case that's still the case so the setup is a bit different it's bigger and somehow probably some of it is flowing into uh the ILS world uh via the exposure of uh traditional players uh and then there is the mega projects and on the mega project side it's uh it's it's taking a lot
42:03
Unidentified speaker (speaker_2): of time for traditional insurers uh to build a product that meets the needs of the buyers. Uh we provide part of the solution. Again, it's uh
42:15
Unidentified speaker (UNKNOWN): um
42:16
Unidentified speaker (speaker_2): weather related and that cat.
42:18
Unidentified speaker (speaker_2): And uh the construction phase is particularly uh unsatisfactory for for the buyers, uh so it's insufficient in uh in size and in terms of uh sub limits uh and we provide part of the solution for our nat cat bi things that are uh the limitations of uh the the traditional products that's
42:38
Unidentified speaker (speaker_2): yet to take uh a complete or standard shape
42:43
Unidentified speaker (UNKNOWN): um
42:44
Unidentified speaker (speaker_2): and there we have we have plenty of uh of things that uh we are looking at so for instance in in japan there is lots of investment
42:53
Unidentified speaker (speaker_2): to build new data centers. We have the only product that's licensed to cover a construction project on a parametric basis for earthquake in Japan. So that's an area where we're looking at different things. But by nature, our or offering is complementary to the
43:13
Unidentified speaker (speaker_2): traditional market. and the traditional market is yet to find a way to cover uh the risk or to uh to convince the the buyers that are out
43:26
Unidentified speaker (UNKNOWN): there.
43:27
Unidentified speaker (UNKNOWN): Do
43:27
Unidentified speaker (speaker_3): you allow me a comment on the relative active attractiveness pricing?
43:32
Unidentified speaker (UNKNOWN): I mean if you
43:32
Unidentified speaker (speaker_3): take the normal ILIS eyeless allocation of an institutional, it's maybe between
43:37
Unidentified speaker (UNKNOWN): two
43:37
Unidentified speaker (speaker_3): to maximum. max six, seven percent.
43:40
Unidentified speaker (UNKNOWN): I hardly
43:40
Unidentified speaker (speaker_3): see more than
43:41
Unidentified speaker (UNKNOWN): that
43:41
Unidentified speaker (speaker_3): that figure.
43:43
Unidentified speaker (UNKNOWN): And
43:43
Unidentified speaker (speaker_3): if you have people who are experienced in the field, they say this goes a little bit in the last um uh panel. I don't need diversification, right? Um if I have two percent in something, I get the thing which brings me the biggest bang for the dollar. I'm fine with US wind, I have a diversify portfolio. So that introduces
44:03
Unidentified speaker (speaker_3): somehow a little bit the benchmark of the pricing of this new opportunity here. Because if we agree that we we might gonna reintroduce a little bit of diversification with the rest of the book, it needs to be priced as least as git as US wind from a multiple risk perspective. So you didn't really mention figures. The question would be are we
44:23
Unidentified speaker (speaker_3): getting there? Also accepting all the different risks and the second part of the question is in case of a loss, how does the tail look?
44:33
Unidentified speaker (UNKNOWN): Um
44:33
Unidentified speaker (speaker_3): are we more in a liability, casualty business here, depending on the claim? Are we have do we have to wait five, ten years? Trap collateral question, or is that a pretty clean and quick cut? Because th that those things are important for nihilist investors as well.
44:50
Unidentified speaker (speaker_2): Just to react on this, so what we do, which is partial,
44:55
Unidentified speaker (speaker_2): just a part of the insurance solutions that exist out there, is relatively short type by nature. So uh
45:03
Unidentified speaker (speaker_2): so from that perspective that's easier to manage. Um and the second point is that uh to your question the economics would only work for an ILS investor if you
45:17
Unidentified speaker (UNKNOWN): have
45:17
Unidentified speaker (speaker_2): many risks that are pulled together. Because uh very often uh we cover a single risk in a single location um against uh like one or two perils, and the the this single peril would not be enough to meet your hurdles. So uh you you'd need to pull together lots loads
45:37
Unidentified speaker (speaker_2): of risks. in order to make it attractive and like slice
45:42
Unidentified speaker (speaker_2): the risk.
45:43
Unidentified speaker (speaker_0): But in the end, that's no different from what we see in the existing commercial insurance markets.
45:49
Unidentified speaker (UNKNOWN): Is
45:50
Unidentified speaker (speaker_0): that an individual facility, some sort of chemical plant, you know, Niklaus mentioned,
45:58
Unidentified speaker (speaker_0): they would not meet the return hurdles of say a cat bond investor, but when pooled as part of a treaty portfolio, there a purchase obviously of um
46:10
Unidentified speaker (UNKNOWN): cat
46:11
Unidentified speaker (speaker_0): cat cover can can indeed meet it. And what what the issue a bit here that I'm sort of hearing is how do you aggregate it up to get that treaty portfolio view? And in the meantime, time who's who's got big enough pockets to warehouse that risk? Because the individual size of the these things are just so large that it's one thing
46:31
Unidentified speaker (speaker_0): for a relatively smaller individual to you know c commercial operation for a number of insurers to swallow that but here they simply can't do this and this and this and then manage to aggregate and and seed that and that just just getting to that point is also a challenge.
46:46
Unidentified speaker (speaker_1): I think this is I I think it's important is when you look at an individual data center, compare it with a warehouse. If you look at the primary insurance rates, I would guess they are ten to twenty times higher for a basic
47:06
Unidentified speaker (speaker_1): per
47:07
Unidentified speaker (speaker_1): per unit of risk compared to normal normal large um
47:15
Unidentified speaker (speaker_1): large insured values on the property side, on the industrial side. So on the primary side you have definitely more, whether it's on the property side, whether it's on the engineering side, we shouldn't forget it's a large part as well.
47:31
Unidentified speaker (speaker_1): We see much higher premiums. When you then get and that is the problem is so that portfolio or that growth only can happen if the insurers have reinsurers and potentially have ILS market because you correctly said, and it it's too big
47:51
Unidentified speaker (speaker_1): for them to grow into it, and and they they underwrite it, and they have the ability to underwrite these risks, but they need on the back. end they need the hedge in place to be able to basically grow into it with the pace we
48:11
Unidentified speaker (speaker_1): see. Having said that, what we see a lot is a bit different, I would say, rather than a single single policy, we we see very defined treaty contracts from insurers. basically re-insurance contracts. And and
48:31
Unidentified speaker (speaker_1): I think there if we compare it is and I think this is I agree with you, Michael, is
48:39
Unidentified speaker (speaker_1): this this is this is not about the diversifier. We don't bring a diversifier Higher here. Here we bring something which grows in a year the size of the entire Florida cap market.
48:55
Unidentified speaker (speaker_1): This is different, but this is also what we are seeing at the Treaty side. Pricing is roughly two to three points above what we've seen on the cap on market in 2023 on US Hurricane. So this is
49:11
Unidentified speaker (speaker_1): the insurers, in order to grow that business, they need
49:15
Unidentified speaker (speaker_1): capacity and they are willing to pay peak peak for it
49:22
Unidentified speaker (speaker_1): because for them it's a huge growth opportunity as well. So that is where we see the market is
49:32
Unidentified speaker (UNKNOWN): the
49:33
Unidentified speaker (speaker_1): cat market becomes the diversifier, diversify, or Florida becomes the diversifier to that market if it grows. grows with the same pace at some point.
49:42
Unidentified speaker (speaker_0): There's your quote, Steve.
49:44
Unidentified speaker (UNKNOWN): Um
49:44
Unidentified speaker (speaker_0): good. We've got about ten minutes left, so I want to turn to the room. I'm sure there'd be plenty of questions for the panel. We've covered a lot uh here today.
49:52
Unidentified speaker (speaker_0): Why not?
49:54
Unidentified speaker (speaker_0): Anyone got a question? Otherwise, I'll
49:58
Unidentified speaker (speaker_0): I'll carry on then. So there was a mention earlier in the panel discussion on of
50:04
Unidentified speaker (speaker_0): obviously moving data centers up into space because
50:09
Unidentified speaker (speaker_0): the issue around, as you say, the compute, right? That is the limit, the key limitation of supporting the data sensors today is the energy availability, the efficiency of a whole bunch of solar panels or small small nuclear power stations which are being established right next to these data centers.
50:27
Unidentified speaker (speaker_0): doesn't really help. And so the concept and you know I was I was reading about uh Google Alphabet effectively who next year is launching a pilot to push a couple of data centers up into space, which means that you can get 100% solar to support those. It's way more efficient. Um, And
50:47
Unidentified speaker (speaker_0): that's the way in which the hyperscaler hyperscalers sort of believed believe that the future uh is going, and it's very real because this they are putting that out next year. So I guess the question is: is this is this something where you can imagine that the ILS market will literally follow this into space, or you know, is this
51:04
Unidentified speaker (speaker_0): somewhat theoretical and let's wait and see, and right now we have such a big problem to deal with that we should kind of stay grounded.
51:15
Unidentified speaker (speaker_1): Solar storms. So there are enough perils. There is, I'm sure, there is enough space garbage flying around threatening these data centers in the future. So I think generally values are big
51:31
Unidentified speaker (speaker_1): and probably they
51:34
Unidentified speaker (speaker_1): in the end they want to have insurance. But I think one of the problems is with all these things: a nuclear plant or these data centers, this is something which is something for five to ten years from now.
51:51
Unidentified speaker (speaker_1): A nuclear plant takes a long time
51:54
Unidentified speaker (UNKNOWN): to
51:54
Unidentified speaker (speaker_1): get approved and to build. And
51:58
Unidentified speaker (speaker_1): I think we see we see this market with a pace.
52:03
Unidentified speaker (speaker_1): we think is interesting over the next five years. We don't know what comes then. Um whether this is still something which is interesting, I don't know, but but um yes it will develop and I'm sure there is need for protection against perils Parallels wherever they are.
52:26
Unidentified speaker (speaker_0): Views from the end of the panel there?
52:28
Unidentified speaker (speaker_2): On this.
52:31
Unidentified speaker (UNKNOWN): Yeah,
52:31
Unidentified speaker (speaker_2): it's um it's quite interesting. Actually, we work with um
52:37
Unidentified speaker (speaker_2): with uh planet that's uh one of our data providers for some of our parametric product, and they they're actually starting to work uh to work into uh a data center in space solution. There are plenty of technology. technical. So the for me, I'm rather on the camp that it's a we still have plenty of time to think about this
52:57
Unidentified speaker (speaker_2): because it's not gonna emerge tomorrow. But let's see. And there are plenty of complexities around it.
53:05
Unidentified speaker (speaker_2): So uh so there is a big latency. Uh you have loads of data transfers uh constraints between uh Earth and space. So you wanna when you're asking a question to your Chat GPT or cloud or whatever
53:18
Unidentified speaker (speaker_2): whichever tool you're using, you wanna reply directly. So this infrastructure is not going to be used for uh it's rather going to be used for training or for things where you don't need um
53:30
Unidentified speaker (UNKNOWN): uh
53:30
Unidentified speaker (speaker_2): to send directly a reply. So in terms of how it can be used and whether it can be done more efficiently than what we have on Earth.
53:39
Unidentified speaker (UNKNOWN): uh
53:39
Unidentified speaker (speaker_2): for me it's yet to be proven. Um
53:43
Unidentified speaker (speaker_0): the message I'm hearing from the two of you is let's just focus on the what we know, what's on the ground, and let's see what happens in the future. And there's plenty plenty to consider today.
53:54
Unidentified speaker (UNKNOWN): I
53:54
Unidentified speaker (speaker_0): do have more questions. I wonder if there's any questions on the floor, just in case. There's there's this one over here.
54:09
Unidentified speaker (speaker_1): Thanks a lot for the panel. Very interesting.
54:11
Unidentified speaker (UNKNOWN): And
54:11
Unidentified speaker (speaker_1): I must say fully agree with Michael or Michael that it's really a concern also from my perspective when you build a portfolio how to look at data center. But I think
54:23
Unidentified speaker (speaker_3): what is clear is as you're also saying is that uh it's coming, so the speed and the scale is immense. So my point is even if we would like to avoid it, because it's not really a diversified, is coming, so it's too big to be avoided. And considering this, I think, and the question to you: there
54:43
Unidentified speaker (speaker_3): will be a selective approach from ILS
54:47
Unidentified speaker (UNKNOWN): to
54:47
Unidentified speaker (speaker_3): data center, or not at all? Because I see I'm
54:51
Unidentified speaker (speaker_0): hearing
54:52
Unidentified speaker (UNKNOWN): you, there are
54:53
Unidentified speaker (speaker_0): completely different ways
54:54
Unidentified speaker (UNKNOWN): to look at it. And my point is is coming,
54:57
Unidentified speaker (speaker_0): it
54:57
Unidentified speaker (UNKNOWN): cannot
54:58
Unidentified speaker (speaker_0): be avoided,
54:58
Unidentified speaker (UNKNOWN): it's too big to a so
54:59
Unidentified speaker (speaker_0): there
55:00
Unidentified speaker (UNKNOWN): will be a
55:00
Unidentified speaker (speaker_0): selective approach.
55:01
Unidentified speaker (UNKNOWN): So I'm also coming back also on the parametric space and the I for and work in parametric, so I it's it's a solution and will
55:11
Unidentified speaker (speaker_3): be a solution,
55:12
Unidentified speaker (UNKNOWN): but as also Abderraim up there I said is a partial solution. So because cannot be fixing
55:18
Unidentified speaker (speaker_3): currently the main made component and the non-at cut one. And I'm happy about that. But then uh so the point is really do you think it's a selective approach or in the end all the ILS market and education will bring to a conversion
55:33
Unidentified speaker (UNKNOWN): to
55:33
Unidentified speaker (speaker_3): say okay, in the end we do it because it's too big to be avoided.
55:38
Unidentified speaker (speaker_0): Maybe if I can just add a quick comment from my own experiences, even today I would say that every single investor in the market has what they consider to be a selective approach. And everyone will take a look at a transaction for its merits, they will have a certain underwriting philosophy, and they will invest or not invest based
55:58
Unidentified speaker (speaker_0): on the structure, the terms and conditions, and the pricing that goes with that. I personally don't particularly see a difference in the way that an investor would would view this, but as always the return needs to be sufficient. And I think that was a little bit where we were sort of leaning. But I mean there was a comment on parametric directly. I d I don't know if you want to talk to the partial solution I
56:18
Unidentified speaker (speaker_0): think it was that was referenced.
56:21
Unidentified speaker (UNKNOWN): Um
56:21
Unidentified speaker (speaker_2): yeah, so um
56:24
Unidentified speaker (speaker_2): in terms of uh
56:27
Unidentified speaker (UNKNOWN): um
56:28
Unidentified speaker (speaker_2): in terms of how to select and build a portfolio so that's it's actually the job that we do for Generali uh
56:36
Unidentified speaker (UNKNOWN): uh
56:36
Unidentified speaker (speaker_2): today. So we uh we have a strong partnership together and our job is actually to uh
56:43
Unidentified speaker (speaker_2): pool all these risks together and create diversification to to build a healthy portfolio for for Generali. Uh
56:54
Unidentified speaker (UNKNOWN): So
56:54
Unidentified speaker (speaker_2): actually just like we package things
56:57
Unidentified speaker (speaker_2): for uh for generali so provide solutions to uh a number of uh client needs and data center needs uh uh to cover that their nat cat we we then pool pull everything together uh and that's the that's the way generali looks at it and that's also why uh they um
57:17
Unidentified speaker (speaker_2): we we provide a number of services in terms of uh accumulation control
57:23
Unidentified speaker (speaker_2): uh uh including accumulation with things that are outside our portfolio.
57:28
Unidentified speaker (speaker_3): Would it be already helpful for you I mean if you have that book and that portfolio on Um if you then
57:34
Unidentified speaker (speaker_1): slice and dice it say you know what a P and C part
57:38
Unidentified speaker (UNKNOWN): I
57:38
Unidentified speaker (speaker_1): could put out and get a coverage for my P and C losses of that book. The cyber risk part, I can do a cap bond. Because we haven't answered the question, how do we gonna trans transform the different contingent claims which we have to the market? And the question is, if you can take it apart and you have
57:58
Unidentified speaker (speaker_1): dialess market is quite suitable for certain risks, which we all agree. P and C part could be easily transformed. Maybe even the cyber part could be transformed. Or maybe the stuff which are not so suitable, which would be maybe the liability part, casualty part, could maybe stay on the book. Would that be already
58:17
Unidentified speaker (speaker_1): helpful? Or does it have to be the quota share all or nothing?
58:24
Unidentified speaker (UNKNOWN): I
58:25
Unidentified speaker (speaker_2): think it's part of the challenge of broken. brokers today to try to find this kind of uh solutions to slice the risk in order to have things that can be digested by each individual risk carrier. And that's also coming back to the lack of standardization in the
58:44
Unidentified speaker (speaker_2): current market, where yes, there is a massive investment into the field, there is demand for insurance.
58:51
Unidentified speaker (speaker_2): but somehow there is yet to uh so a standard is yet to be found. Because in order to do what you're describing, you need
59:00
Unidentified speaker (UNKNOWN): a
59:00
Unidentified speaker (speaker_2): number of risks that are placed that you can then pull together, slice, and redistribute. And that's yet to happen at the ground level for most of the very large uh data centers that are being built.
59:15
Unidentified speaker (UNKNOWN): And
59:15
Unidentified speaker (speaker_1): I think it is already sliced to a large degree because insurers reinsurers have the same problem is they
59:23
Unidentified speaker (speaker_1): I think the slicing is that you know
59:28
Unidentified speaker (UNKNOWN): I
59:28
Unidentified speaker (speaker_1): think it's a very traditional property policy. It's um it's really it's property loss from
59:36
Unidentified speaker (speaker_1): from things like fire.
59:39
Unidentified speaker (speaker_1): Um and and it is with that also I think very short-tail ish. Um
59:48
Unidentified speaker (speaker_1): I think things like cyber we see very sublimited and limited or excluded in these policies because
59:58
Unidentified speaker (speaker_1): the problem the problem is is complex and I think this is also why capacity is probably missing is think about the reinsurer having cyber on the book as well as data centre. If the data s if you look at the cyber take the cyber cap bonds, if you look there in the models
1:00:19
Unidentified speaker (speaker_1): what is what is the reason why you get a total loss to the to the cap bond or the cyber bond is because a data center fails.
1:00:30
Unidentified speaker (speaker_1): So in theory you can lose twice, and I think this accumulation is another constraint. So it is important to really to slice the risk and make it controllable
1:00:44
Unidentified speaker (speaker_1): from downstream to upstream.
1:00:47
Unidentified speaker (UNKNOWN): Good.
1:00:47
Unidentified speaker (speaker_0): So we're we're a little over time. I think it's a good note to finish on. So the opportunity is very real. It exists. There's work being done on the ground, but there's even opportunity today. With that, I'd hopefully like to thank the panelists today.
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