PFI Data Centre Roundtables 2025 - Structured Finance Transcript
Richard Leong: Thank you to the first panel— that was very insightful. It sets us up well to discuss securitisation and term financing for the digital economy.
A key theme from the first panel focused on the construction phase, whereas we’re looking at terming out that financing into securitised markets. The recent Blue Owl Capital/Meta Platforms transaction is an example of long-term capital entering the space.
To start, with the size and structure of the Blue Owl/Meta deal, what are the broader implications for securitisation? Will we see more deals of that scale, or a different path?
Jon Salzinger: There is roughly US$50bn of outstanding securitised data centre debt across CMBS and ABS. As these mega transactions currently under construction mature, they’ll require term financing. The capital needs will be enormous and – as in construction financing – the investor base for term financing will have to broaden. The Meta deal demonstrates that shift.
We expect many large projects to be financed in pieces across multiple markets — CMBS, ABS, self-amortising project bonds and corporate JV structures. The scale of capital required means we’ll see all of the above.
Brian O’Hara: Agreed. A large conduit CMBS deal today is around US$1bn and that can be difficult to place. The largest single asset, single borrower data centre CMBS deal was US$3.5bn. We will need multiple channels: CMBS, ABS and private markets.
Some CMBS investors still won’t touch data centres because they don’t understand the asset. Education takes time but as more deals come, comfort increases and the investor base expands. That supports wider deal flow across both markets.
Jian (Jay) Hu: Stepping back, the data centre financing boom is driven by three forces: artificial intelligence, cloud computing and digital transformation. These forces brought on strong momentum in the supply and demand mix, refinancing/renewal needs and technology/regulation advances. Sustaining the momentum and the degree of success will depend on the characteristics of assets, variation of structural innovation and the quality of the data centre manager.
Richard Leong: Before we go on, it would be helpful to outline how ABS and CMBS formats are being used to term out data centre financing.
Why might a sponsor choose one format over the other — or use both?
Jian (Jay) Hu: ABS has been extremely active – issuance has grown more than 500% over the last five years, albeit from a small base, and this year’s issuance to date has already exceeded last year. We (Moody’s) published a methodology earlier this year and since then have rated deals including a US$1.1bn DataBank transaction, and there is a strong pipeline behind it.
ABS is attractive because of structural flexibility: you can tranche risk and diversify collateral within a master trust structure, enabling smaller, bite-sized transactions across many data centre assets. It provides an alternative to issuers and investors beyond large single asset, single borrower CMBS data centre deals. We expect continued strong growth in ABS. The potential is enormous.
Richard Leong: Jonathan, could deals on the scale of Blue Owl/Meta ever be broken up and executed in securitised form? How are clients approaching this?
Jon Salzinger: Large projects are increasingly being segmented into multiple buildings with multiple leases – sometimes with identical structures – so that term financing can be completed in smaller bite-sized pieces over time. A US$25bn project might be structured as 10 US$2.5bn projects that can be financed over a staggered period of time.
Traditionally CMBS handled the largest single asset deals – up to US$3.5bn. ABS deals have been smaller and in master trust format, but where the total quantum in the master trust can exceed individual CMBS volumes. There are pros and cons.
Historically, single tenant long-term lease assets were a CMBS fit. Over the past year, however, we’ve seen successful single tenant ABS executions – Compass, QTS, EdgeCore – so both markets are viable.
Choice often reflects sponsor background: real estate funds are CMBS-comfortable; infrastructure funds often prefer ABS. As capital needs scale up, issuers will use all markets.
Jian (Jay) Hu: From a ratings perspective, ABS and CMBS structures require different methodologies because their risk profiles differ. CMBS has more refinancing and event risk; ABS benefits from diversification and amortisation and associated statistical techniques for analysing long-dated risk exposures.
Brian O’Hara: CMBS analysis is looked at from a real estate underwriting perspective: leases, exit debt yields, tenant credit and the ability to refinance. It’s non-recourse, so we must be comfortable with the assets and market liquidity at the right kind of debt yield at maturity. ABS has master trust support and reinvestment ability; CMBS doesn’t – it is a single entity (asset or portfolio) that can be handed back. The investor base also differs – insurers favour fixed-rate ABS while money managers and hedge funds are active across CMBS capital stacks because it goes down below investment grade.
Investors are up to speed with both structures and for data centres, it’s a mixture of the two types of asset classes and we’ll continue to see both for data centres. It really depends on the issuer.
Jon Salzinger: Cross-market financing isn’t new – net lease assets have long used both CMBS and ABS. In data centres, on a representative single asset, single tenant, 15-year triple net lease to a rated hyperscale tenant, we believe ABS can achieve a 75% advance rate versus a 65% advance rate in CMBS.
And as we think about forming the capital stack and what that pricing looks like, the overall cost of debt capital, even at the higher advance, tends to be lower in ABS. In this hypothetical example, that’s worth around 30bp. So, 10% higher leverage and 30bp lower cost of debt in ABS. That’s a key driver of ABS growth.
Jian (Jay) Hu: ABS also employs structural triggers that add and tailor investor protection.
Richard Leong: Let’s touch on refinancing risk. Some of the initial wave of ABS deals are approaching anticipated repayment dates. How do you assess the refinancing risk for some of these deals, even though they are backed by relatively long underlying leases on the data centres?
Jon Salzinger: Leases originated several years ago are now below market so there is significant embedded rental uplift. That’s supportive of refinancing.
The bigger question is the projects in development today — seven years from now, when you’re five years or so into a lease with 10 years left and you’re in a different interest rate environment, the calculus may change. But near-term refinancing risk is manageable.
Brian O’Hara: From a CMBS perspective, with leases rolling at 40% below market you can be comfortable that it will get taken out. Near-term refinancings look healthy given sector growth. The more uncertain period is longer term: smaller tenants with more optionality require strong operator capability and power availability.
Jian (Jay) Hu: Agreed – near-term risk is low. Five to seven years out will be less certain as more debt matures.
Jon Salzinger: In traditional CRE, assets nearing lease maturity often move back to bank balance sheets before reentering securitisation. The question is whether banks can absorb assets of this size – they are so large. We’re not at that point yet.
Brian O’Hara: If distress occurs in CMBS, the special servicer comes in and takes over the property; the receiver comes in.
With data centre management, expertise will be critical, especially with smaller tenants where the management is highly intensive. We have great operators right now but if they run into trouble, does the receiver have the right expertise to keep the centre going and maintain the value there?
Richard Leong: How do you factor geography into securitisation and structuring? Not every asset is in a tier-one market.
Jon Salzinger: Location is a major driver of ratings in CMBS and ABS and in bank financing. With tier-two and tier-three markets, there is typically a difference in development yield on cost, and then, from a sizing perspective, that manifests into healthier debt yields on those transactions, so stronger debt-coverage metrics. That is going to help facilitate the term financing in securitised markets.
You tend to have better cashflow margin in weaker markets, and that’s the mitigant we’re looking for.
Brian O’Hara: Lower-tier markets need higher debt yields. Agencies stress test those markets more, so tranching is different than in tier-one markets.
Lease terms and power access all feed into exit debt yields.
Jian (Jay) Hu: We group data centres into five tiers in our ABS methodology. Location matters – customer access is essential; power and water constraints are real – but facility quality and modernity are equally important. Some data centres being built now have highly advanced facilities. Given data centre financing is long term, over a 10 to 15-year horizon, we assess cashflow volatility by tier, with lower volatility assumptions for the highest-tier locations.
Richard Leong: Regarding the discussion on tier-one, tier-two and tier-three markets, we understand these concepts in commercial real estate, but given the constraints of building a data centre – power, water, land – is the tier classification the same as in standard commercial real estate or is it different?
Jon Salzinger: It’s different. It also ties into fibre routes because connectivity between data centres is crucial. Northern Virginia is the primary market in the United States, but there are other strong markets such as Chicago, Atlanta and Phoenix. That does not mean that New York is a bad market – it just isn’t a core market like Northern Virginia due to factors such as power availability and access to fibre. The critical mass in northern Virginia compounds to make the market stronger.
Brian O’Hara: It’s not typical commercial real estate. Northern Virginia is not the greatest office market, nor is Chicago or Atlanta, in traditional CRE terms. These are tier one because of the factors discussed, not for CRE price appreciation.
Jian (Jay) Hu: This is an important question as it differentiates CMBS from ABS. When analysing ABS structures, we focus more on cashflow rather than the property itself. Our data centre grouping considers more than just the property location and power access, but also tenant quality, lease term and manager experience.
Richard Leong: Considering the Blue Owl/Meta megadeal discussed in the previous panel, what takeout solutions might exist, given the leverage and direct loans? It doesn’t seem like an ABS.
Jian (Jay) Hu: A US$20bn deal is not going to be an ABS.
Large deals carry their distinct risks such as event risks. ABS structures diversify and divide risks and cashflows and transform them into more agile and customised tranches for wider investor bases. Whether they are larger deals or smaller deals, they are just different; not necessarily better or worse and they help the market grow more robustly in different ways.
Jon Salzinger: That structure is a long-term bond. Refinancing isn’t imminent so the cashflow story just needs to play out. There will likely be innovation before it is refinanced.
Richard Leong: In the securitisation market, what developments might we expect over the next 12–18 months in terms of maturities, structures or rating methodology changes?
Jian (Jay) Hu: Project finance CLOs are increasing. Data centre loans are a growing component. Additionally, the market is exploring hybrid structures between ABS and CMBS. These structures combine benefits from both markets.
Jon Salzinger: Innovation will likely include investment-grade JV structures, like the Meta transaction, providing access to bond markets for hyperscalers.
Brian O’Hara: Investors remain open to different structures if pricing is appropriate. We don’t shy away from looking at these.
Richard Leong: Sustainability was raised in the previous panel. Many data centre ABS deals are green. Do you expect changes in reporting or green label requirements?
Jian (Jay) Hu: ESG remains a feature. Some clients highlight the environmental aspect as beneficial. Certain deals have achieved lower bond spreads due to environmentally friendly structures.
Brian O’Hara: ESG continues to be important to Janus Henderson, particularly for our UCITS funds. ESG is important for us but for other funds it doesn’t matter.
Jon Salzinger: Key constituents – issuers, banks, investors – still care about ESG so it remains meaningful despite political changes.
Richard Leong: What about financing supportive assets, like graphics processing units or data centre-adjacent infrastructure?
Jon Salzinger: GPUs are analogous to data centre ABS or equipment ABS. Currently, most financings are private loans sitting on bank and private credit balance sheets. ABS takeouts haven’t emerged yet.
Jian (Jay) Hu: We are seeing enquiries for chip and GPU securitisations, likely in ABS format. Data infrastructure offers abundant opportunities, potentially in the billions of dollars.
Brian O’Hara: From a CMBS perspective, this isn’t our area, but ABS teams are actively considering these opportunities.
Richard Leong: What about financing for cooling systems and fibre infrastructure to support data centres?
Jon Salzinger: GPUs face technological obsolescence, cooling less so, and fibre even less. The key is a long-term lease with an investment-grade offtaker. ABS fibre financing already exists; other ancillary assets should also be financeable.
Richard Leong: Regarding tenors, ABS deals are typically five years. What about CMBS and amortising bonds?
Brian O’Hara: Generally floating rate, two years with three one-year extensions, no amortisation, with flexibility to extend using debt yields and interest rate caps.
Richard Leong: For a 10-year CMBS on a 15-year lease, how would you characterise investor appetite?
Brian O’Hara: Ten years is long. Shorter tenor and floating-rate assets are preferred. Fixed-rate assets around five years are more attractive.
Jon Salzinger: For 15-year leases, bonds would typically amortise over the lease term.
Richard Leong: As we wrap up, what do you expect in the sector over the next year?
Jon Salzinger: US$38bn deals will seem small.
Brian O’Hara: Hyperscalers will continue spending; we’ll discuss similar topics next year.
Jian (Jay) Hu: Expect more structural innovation; new structures will likely emerge.
Richard Leong: Thank you all for your comments.