Subscribe
Share
Share
Embed
TCW is a leading global asset management firm with over 50 years of investment experience and a broad range of products across fixed income, equities, emerging markets, and alternative investments. In each episode of TCW Investment Perspectives, professionals from the firm share their insights on global trends and events impacting markets and the investment landscape.
AI is clearly the topic of the day, and the topic has numerous facets, from the
incredible performance of the MAG7 stocks, to the massive recent debt issuance
of debt from the hyperscalers, to the challenges of building and powering data
centers and other critical infrastructure necessary for the AI-enabled future.
Even so, much of the public conversation is focused on
semiconductors, software, and the largest technology companies.
But as the demand for capital permeates markets around the world,
securitized markets provide a different set of opportunities and risks.
Today, we're going to look at how CMBS and ABS markets have
been impacted by the AI buildup, what's happened to investors
thus far, and how we see that market developing over time.
Welcome to the TCW Investment Perspectives Podcast.
I'm your host, David Vick.
Joining me today are two members of TCW's Securitized Products team.
Liza Crawford is a Managing Director, Specialist Portfolio Manager in
TCW's Fixed Income Group, and a Co-Head of our Global Securitized Team.
Dominic Bea is a Senior Vice President on the Fixed Income
Team, and a Senior Analyst Specializing in ABS Securities.
Liza, Dom, welcome to the podcast.
Thanks for joining me.
Thanks so much for having us.
So let's get started here.
AI is obviously creating massive demand for data centers, for
servers, for network equipment, related infrastructure, and the
financing for that has spread across a whole bunch of markets.
From a securitized credit standpoint, how should investors think about
the financing of this build-out and the needs that are required for that?
Yeah, so I think it's worthwhile to start with some context.
Data centers have been around for a while, and there's
been significant investment in them over time.
So really, the evolution is twofold.
In the ABS market, you've seen data center financing since 2018.
In the single-assist, single-borrower CMBS
market, you've seen financing since 2021.
But there has been significant increase.
So right now, you've got around $50 billion
outstanding in ABS and CMBS and heavy issuance.
So from an investment perspective, managing securitized
portfolios, it's become even more relevant.
There's more to take advantage of, and there's more to come.
And then from a credit underwriting perspective, to
your point earlier, AI has become incredibly topical.
That's another demand driver for data center.
And so that's part of our analysis when we're underwriting
these credits and also anticipating the supply to come.
So I think with that context, it's important
to understand that data centers predate AI.
And a lot of data centers aren't running AI.
And so before we have this conversation, I think a little bit of context
about what these different types of deployments are is important.
So there's five kind of key types of
deployments you can have with the data center.
You can have on-prem, co-location, cloud, AI training or AI inference.
And each one of these have a sort of a different business perspective and they
impact the type of assets that are financed in their commercial arrangements.
So the first is on-prem and this is really the legacy business.
This is your data, your servers in your data center.
Oftentimes that meant the whole closet at the office.
This is a runoff business, right?
And this isn't financed in ABS or CMBS or really elsewhere in capital markets.
What began to replace on-prem was co-location.
This is your data on your servers in someone else's data center.
These assets are typically less technically demanding, but they are
latency sensitive and you typically get a diversified rent roll.
With different quality of tenants from investment grade down
to small, medium price enterprises with shorter lease terms.
Co-location is a fine business.
It's growing single digits.
What really superseded co-location is cloud service, right?
This is you hosting your data with Microsoft or Amazon or Google.
So this is your data, third-party servers in a third-party data center.
These are highly demanding assets.
They are highly efficient.
They're latency sensitive.
They have really rigorous uptime requirements.
Typically, these are leased by a single investment grade tenant on a five
to 15-year lease term, fixed rent, fiscal escalator, that sort of thing.
This is your classic real estate lease.
And until 2023, these were the two types of data centers that were out there.
And with the advent of AI, you now have sort
of two types of deployments related to AI.
There's AI training and AI inference.
AI training is about cheap power, cheap land,
high density, and latency and sensitivity.
That is really big, powerful data centers, far remote from populations.
This is just where you can get the compute turning to train the model.
And the last type, the AI inference.
This is also similarly a very high-powered deployment,
but it needs to be closer to the load centers.
And so these are kind of the five types of deployments you see in data center.
And it's really important to understand because these tend to gravitate
towards different structures, towards different parts of the market,
and will be financed either on balance sheet by the tenant, financed in
the corporate credit markets, or financed in the securitized markets.
And even more, the different types of data
centers will be financed in ABS or CMBS.
Got it.
So we've talked a lot about data centers to this point, but there's
a lot of other parts that go into, you know, building a data center.
Obviously there's power and cooling, capacity, you
know, all of the other infrastructure that's involved.
What's been the impact of that secondary demand on securitized markets?
What role has that played?
So if we think about the securitized market, when you've got
data center being financed in ABS or CMBS, it is stabilized,
it's energized, it's built, it's got its tenants.
And so you're really underwriting something that's very
different than, say, a syndicated construction loan.
And with that, you have the benefit of certainty of execution of that power.
But you need to make sure you've got redundancy,
you need to make sure that you've got backup, etc.
And then also we benefit from some really accessible developers and issuers.
They will let you tour their property, they will
talk to you about the future proofing of the asset.
So I'd emphasize that when you're accessing data center exposure through
the securitized market, for the most part, you're mitigating a lot of that
access to power risk, but you need to make sure you have no interruptions.
And that extends, as you all know, in our securitized market, if you were
to have a default, can you make sure that you can take over that asset?
You have access to everything you need to keep it running.
And then additionally, as we all know, that the demand might evolve.
We are talking about AI to start the podcast.
Absolutely, you're seeing more future proofing to make sure
that you can accommodate tenants' different needs, whether
it's AI or it's the more traditional cloud service.
And that type of effort should be ongoing for the developers and
these assets to make sure they stay relevant over the long run.
So, we've talked a little bit about CMBS and ABS.
Why would an issuer use the CMBS market, the Finance
One deal, the ABS market for a different deal?
Like, what drives that decision, and what are the differences between those two?
So, at a high level, the financing is going to
come down every single capital markets pipe.
As we know, financing is in high demand.
So, aside from the diversification and tapping two
different markets, the asset-backed security side is
going to be issued typically off of a master trust.
It's going to be some smaller issuance sizes, whereas the
commercial mortgage-backed security side accommodates
issuers that are looking to execute on a billion-plus.
Additionally, you're seeing evolution on the structure side.
So, in the securitized market, you've got to be an expert in the
underlying collateral, the cash flows, and also the structure.
And then you think about the relative value when
you're constructing portfolios and trading.
On the CMBS side, we saw some initial prints that
were longer fixed-rate, strong prepayment protection.
Then the terms evolved closer to floating rate,
kind of two-year plus three one-year extensions.
And then now we're seeing a broader mix of five-year,
ten-year fixed-rate, anticipated repayment date structures
with longer ten-year terms until the final maturity.
And we've also seen fixed-rate that basically replicate the ABS model.
The ABS issuance tends to be five-year term with three years of call protection.
So one of the things that used to be true of the ABS markets
was that they were meant for highly diversified trust.
So you'd have assets from Chicago, New York,
Atlanta, Texas, Phoenix, all in the same trust.
And so there was a real differentiation in
terms of the underlying commercial risk, right?
You'd get these big, chunky data centers, sometimes with big tenant
concentrations in the CMBS market, and you'd get the more diversified,
well underwritten, kind of primo collateral via the ABS market.
And so the ABS market would price tighter than the CMBS market.
What we've seen over time is that the structures
and collateral types have kind of melded together.
So a lot more now in ABS, you'll see one market within an ABS trust.
There are some issuers out there that have only one market.
So they have a Phoenix trust and they have an
Atlanta trust and they have a Portland trust.
And so what they're doing there is they've sort of shifted the
reason for the master trust structure, which used to be to
finance their entire platform, to now finance a campus.
They'll have a campus, they'll have 15 pads in Dallas, Texas.
They'll build out five that issue ABS and then they'll...
As they build out each pad, they'll issue another series out of the ABS.
And so the market's evolved in that way, but you've definitely
seen what used to be sort of two very distinct products with
similar underlying collateral really start to meld together.
Interesting.
Thanks.
So, Vlaja, you talked a little bit about some of the
things you don't have to worry about necessarily.
The power's already set up.
Those things are in place already.
What are the key risks that investors, whether on the CMBS or ABS
or maybe both sides, what are the key risks that investors need
to think about when they're buying data center back securitized?
So we'll look at it multiple ways.
You shouldn't buy anything where you think the credit's an issue, but
bigger picture, you are typically facing some better quality tenants.
And then as Dom was highlighting, you have different use cases.
So you've got your traditional co-location
cloud service in addition to the AI side.
We're going to see more of the cloud service in
the securitized market, more of the co-location.
What you're thinking about is basically the credit's
going to make sense, and then where do I see the value?
We've been able to pick up some really
attractive spreads for enterprise co-location.
So co-location for enterprise clients.
That's a really great profile with really strong operator developers.
They're future-proofing the assets.
They've got intentionality in their locations, etc.
And then additionally, we've been mindful, especially
in the commercial mortgage-backed securities market,
where you've seen more evolution in the structure.
You don't have the commoditized structure of the asset-backed market.
Mindful that you're going to get some liquidity give up, because everyone's
going to be underwriting the specific asset, and then they also need to review
the structure and the potential extension risk, etc., from those profiles.
So we're mindful there.
Additionally, we've been very thoughtful overall in our portfolios.
If ABS is trading cheap to CMBS or vice versa, we'll take advantage of that.
In the CMBS market, you typically have a steeper credit curve
to take advantage of in those larger deals, for example.
We'll see some opportunities to trade.
And then overall, our exposure has been really shorter duration, shorter spread
duration, mitigating the risk of potential widening on increased supply, whether
that's directly into the securitized market or that tends to be a knock-on
effect from significant hyperscaler issuance
in the corporate credit market, for example.
Additionally, TCW is very focused on making sure across fixed income, we're
thoughtful about the debt coming through at all levels of development of data
center and making sure that issuers that are very active across markets are not
able to pick off any fixed income investors that might be operating in a silo.
So that's another element.
And the last thing I should mention is we are
seeing some more significant deltas in leverage.
So that's something we're keeping an eye on if we've
got some issuers taking advantage of too much leverage.
So I think that Liza picked up on something really important there, which
is the types of data centers that are financed via the securitized products
market and the types that are financed via the corporate credit market.
If you look at it from the CFO seat at a hyperscaler, the CFO can finance a data
center build on balance sheet with an unsecured issuance, or they can lease.
And if they lease, all those assets typically got
financed through the securitized products market.
And those were on five to 15 year leases,
fixed lease price with a fixed escalator.
And that was an underwrite for us as structured products investors
of both the tenant and the lease, and then the underlying asset.
Because you're with a five to 15 year lease, you're necessarily
underwriting and internalizing a releasing experience for that asset.
What we've seen come up in the corporate credit market are
full life cycle leases, or what we call cost recovery leases.
This is where the lease price is not a fixed rate,
but rather a yield on cost for the landlord.
And then all of the operational requirements, construction requirements, all
those are put back onto the tenant, meaning the tenant can't cancel the lease.
So really what those become are a look through to the underlying tenant credit.
And so those structures can finance through the corporate credit market.
And typically those are going to be those
AI training data centers we talked about.
They're remote, they're highly specialized, they're very expensive,
and you couldn't finance those via the structured products market,
given our structures and given the way that people like us think.
What we get are the more traditional commercial real estate
leases and the assets that are appropriate for cloud service
or co-location, which really means that the assets are more
flexible, they're more desirable for different workloads.
You've both touched a little bit on sort of the differences
between corporate credit and securitized credit.
Maybe talk a little bit about how the opportunities
set, like what do the valuations look like?
What it looked like a couple years ago?
How has it evolved over time?
And how do you see that value proposition, the
opportunity set changing as we move forward from here?
The structured products markets have evolved a lot over the last few years.
The first is the proliferation.
So we used to have a handful of issuers, and these guys all priced wide of 200.
It was a specialized sector.
And where we had three or four issuers back in 2019, we now have 27
different tickers in our market in ABS, and then another dozen in CNBS.
So it's a much larger market, it's much
deeper, a lot more liquidity, tighter pricing.
You've also seen evolutions in structure.
So ABS, which used to be senior single A only, has now structured down.
You typically see a triple B bond in those structures.
They've also structured up, meaning they've taken that single A, that's
senior single A, and chopped it up into a triple A, double A, and single A.
And so investors are getting more choices there, and pricing is getting tighter.
And then you've also seen a collateral evolution,
which relates to the underlying assets.
Lease terms are a lot longer.
So when we started out doing this, lease terms were
five to seven years on the underlying collateral.
And you were underwriting a really big lease role as an investor.
Now you're seeing leases come out 10 to 15 years
more typical in your securitized products and deals.
And to emphasize a couple of points that Dom made, and by the way, Dom's a great
example of why you need to have the credit discipline in these investments.
And so we're just really grateful to have Dom
on my team and not on any opponent's teams.
We just want to emphasize that with $50 billion outstanding
in ABS and CMBS, we are still so small compared to the amount
of issuance coming from corporate credit market, et cetera.
But we welcome any growth in different sectors.
And especially if there's credit differentiation, term
differentiation, structure differentiation, and that
allows us more opportunities to trade that relative value.
Bigger picture, year to date in 2026, the main kind of action has been in
rates, and we haven't seen that translate into much credit spread movement.
In fact, it's been a buoy for some sectors.
So the main kind of credit widening we've seen in the broader
securitized market remains CLOs down the cap stack on software weakness.
And seeing a little more widening on the digital infrastructure side, thanks
to more issuance on the corporate credit side, headlines, supply concerns.
So we're grateful for the opportunity to take
advantage of that, and hopefully we'll see even more.
All right.
So that's a great review of kind of where markets are at.
Let me change directions a little bit.
No conversation about AI would be complete without
looking at it from the other's perspective.
So how is AI influencing what you guys are doing every day?
How is it helping you or making things easier or harder?
Is it changing what you do and how you think about
markets and how you think about what you're doing?
Yeah.
So I appreciate TCW being such a supporter of making sure
talent has access to AI and we optimize around its use case.
And we've gone through the evolution of more
internal build to exploring more third parties.
And really, what I see it as is not a replacement
for any talent on the desk, but an enhancement.
We love using it, so it can be used to kind of extract
the data that you need for a lot of credit write-ups.
So you have an auto-populated and then you're going in with all the manual work.
It is also really helpful for taking unstructured data and organizing
it in a meaningful way, whether that's for internal dialogue
around risk budgeting or for client updates on their portfolios.
It's been really helpful.
I'm just nervous when everyone tells me it costs a lot.
So in the rare exception, we haven't developed any
processes that are absolutely 100% handled by AI.
In the interim, it's really deploying it to
the team and making sure they're using it.
As a manager, one thing I know too, it is amazing how
different the experiences people will have with AI.
I can give somebody a login and they won't touch it.
And then I'll just rotate it to somebody else.
I give somebody else a login and their productivity is doubled because
they're able to get to more of the projects they otherwise wouldn't get
to because that manual work is now much more efficient handled by AI.
It's an incredibly powerful technology.
The one caveat I think is you can't rely on it exclusively.
You can't become a crutch because the last thing you want to
do is become programmatic and predictable because Wall Street
will then figure that out and we'll optimize around that.
So AI enhances our investment process.
It does not replace our investment process.
You make such a good point, Dom.
And I think about this too, or we all should.
You'll see these credit notes come out from certain areas
and then you try to talk to the human to go through it.
They're like, "Oh, no, I didn't do that.
I just ran this information through this output and that's
what it is". So that human's not getting smarter, right?
Instead, what you want is the humans are really talented and smart.
They're constantly leveling up their knowledge, their skillset, and then
they are using the AI as a really smart analyst on their desk to just
handle a lot of the grunt work and make them even more efficient.
We'll hopefully see that play out.
Awesome.
Well, Liza, Dom, thank you both for joining me today and sharing your insights.
Thanks everybody for joining us on TCW's Investment Perspectives Podcast.
While the long-term implications obviously are still evolving, understanding
the nuances of all these details will be critical for investors seeking
to identify both opportunities and potential risks in the years ahead.
For more information on TCW strategies, please visit our website at tcw.com.
Thanks for listening, and we'll pick up next time exploring
the trends and opportunities shaping global markets.
Thank you for joining us today on TCW Investment Insights.
For more insights from TCW, please visit tcw.com/insights.
This material is for general information purposes only and does not
constitute an offer to sell or solicitation of an offer to buy any security.
TCW, its officers, directors, employees, or clients may have
positions in securities or investments mentioned in this
publication, which positions may change at any time without notice.
While the information and statistical data contained herein are based on sources
believed to be reliable, we do not represent that it is accurate and should
not be relied on as such or be the basis for an investment decision.
The information contained herein may include preliminary information
and/or "forward-looking statements". Due to numerous factors,
actual events may differ substantially from those presented.
TCW assumes no duty to update any forward-looking
statements or opinions in this document.
Any opinions expressed herein are current only as of
the time made and are subject to change without notice.
Past performance is no guarantee of results.