00:00Trey Parker, Sycamore Tree Capital Partners, CIO and co-founder, joining us here today.
00:05Thanks so much for being with us.
00:06Thank you for having me.
00:07You know, you just wrote a paper about AI data centers and how the build-out is increasingly credit-financed.
00:14When I went through it, it didn't seem that you were concerned about AI, but the credit financing of it.
00:21AI is a massive economic boom and trend, if you will, in the U.S. economy, in the Texas economy,
00:27in the global economy.
00:28But it's acquiring a lot of capital.
00:30Morgan Stanley estimates nearly $3 trillion of capital will be spent on AI over the next several years.
00:35Only half of that is being funded by cash flows out of hyperscalers and large companies.
00:40The other half is going to the credit markets, and that's where the focus of the white paper was.
00:44And you talked a little bit, too, about the public markets versus the private markets in this situation.
00:49Yeah, we've seen year-to-date already almost $500 billion of issuance of credit to fund the hyperscaler and AI
00:56and data center-driven CapEx build.
00:59The private markets have financed more than half of that.
01:03We've seen over $100 billion of issuance in the IG markets year-to-date, which is up 20x year-over
01:08-year.
01:08The high-yield markets, which is another very active market, has seen over $50 billion of issuance against zero last
01:14year.
01:15And certainly, asset-backed finance markets, private credit markets, and just private credit issuers have been big deployers of capital
01:23into this build-out.
01:24One thing you pointed out was the telecom boom, almost as a warning.
01:29So, we've seen massive capital builds in the past.
01:33We've seen it for power infrastructure.
01:35We've seen it for, as you point out, or I pointed out in the paper, the telecom boom.
01:40Massive amounts of capital are being spent for a demand equation that presumably is going to come to bear.
01:47Everybody likes AI.
01:48Everybody's using AI in their business models.
01:50It creates efficiency.
01:50It creates opportunity.
01:52But all of this is requiring a certain level of data center construction, GPU construction.
01:59Just like the fiber was built in the telecom boom back in the late 90s and early 2000s, the demand
02:06ultimately materialized.
02:07Ultimately, internet traffic and demand came through.
02:11The challenge was the financing markets that were financing a lot of that build-out matured before the demand and
02:17the pricing of that demand ultimately came to bear.
02:19And that created a lot of disruption, a lot of default and issues in the credit markets.
02:24So, how do investors really shield themselves from the risk that you're talking about here?
02:28Look, I think, as I said, AI is going to be a prolific part of our economy.
02:32I think being discerning credit investors, picking your spots, making sure you're getting paid adequately for the credit risk you're
02:38taking is the most important thing.
02:41As we think about being credit investors or providing financing to AI and data center companies, we're staying relatively short
02:49in our maturities.
02:49We're focusing on getting excess spread relative to what the IG or below investment grade markets are otherwise paying.
02:59Strong tenants, obviously the folks that are leasing that data center capacity need to have strong tenant credit ratings.
03:08And then relatively high amortization, so where we're getting money back before the maturity because that's ultimately the issue.
03:14So, if the boom goes bust, what are the real threats in the credit market?
03:19Yeah, like I said, I don't think this is a credit bubble per se.
03:22I think the real issue is the timing of the credit maturities and the risk that's being taken on the
03:28development side of our markets vis-a-vis when the economics of AI ultimately play out.
03:34And so, people are taking cliff risk or bullet financing risk on a development deal is a much different equation
03:40than someone who's taking operating data center financing risk, if you will.
03:46So, how is Sycamore playing the dislocations we're seeing in the AI debt world?
03:51So, we're certainly taking advantage of some of the supply technical.
03:55So, right now, we're seeing a fair amount of supply, as I mentioned, up 20x in IG markets, up substantially
04:01in the below investment grade markets.
04:03A lot of that supply tension or that excess supply is leading to wider spreads.
04:07And so, we find a lot of those opportunities or many of those opportunities interesting, but we're being highly selective.
04:13Again, we're focused on high-quality tenants, short-duration facilities that have cash flows to pay us back as credit
04:19investors.
04:20So, you've looked very deeply into this with your research.
04:23What do you think are the main hurdles for AI going forward?
04:27Look, I think the capacity and the capital getting deployed efficiently and effectively, it's a lot of capital.
04:33$3 trillion deployed over a three- to four-year period in a market that didn't exist 24 months ago.
04:39Very, very challenging.
04:41These business models aren't tested over a cycle.
04:43We don't know what the collateral is going to look like if a data center doesn't work or if there's
04:48some sort of, you know, repricing, if you will, of AI.
04:51And I think that's really what we have to be careful about is this fast a deployment of capital, there's
04:56going to be mistakes made along the way.
04:58We sit here in Texas and you talk about capital.
05:01There is a lot of capital in data centers here in Texas.
05:05Look, Texas is a pro-business state.
05:07We've been very hospitable to business investment, capital investment.
05:10There's $90 billion of data center development underway in the state of Texas.
05:15You've seen it in the power markets.
05:17We haven't talked about power yet, but that's a big component of the AI build-out.
05:22Just in ERCOT, which is the Texas power market, interconnection demand has gone from 64 megawatts in the end of
05:292024 to 438 gigawatts, you know, at the end of June of this year, up 7x.
05:37Only about 1.5% of that has actually been approved and operating.
05:40So the conversion rate is really low.
05:42But the state of Texas has already, even as politically hospitable as it is, put constraints in place, made data
05:48center companies, put capital at risk to be able to be in that interconnection queue, to be able to ensure
05:54that they're paying for the interconnection risk.
05:57The consumer, the folks that are paying their electricity bill, don't want to see the adverse impact of data center
06:02and AI growth.
06:03And so I think states are having to step up and protect that.
06:06You know, you're seeing that in the states.
06:08What's your reaction, though?
06:09We're seeing the backlash here in Texas of the not-in-my-backyard.
06:12Yeah, it's a real issue.
06:13I think there's certainly some political issues with how people are positioning AI as a potential threat to jobs.
06:20There's a political issue that's being highlighted in that it could increase your electricity costs, your power costs to run
06:26your home.
06:27You know, at the end of the day, consumers are very focused on their wallet, on what they can spend
06:31and what they have discretion to spend.
06:32And as you have these things that potentially threaten jobs and create efficiencies for business, it puts the consumer in
06:39a really difficult position.
06:40And I think that's where that not-in-my-backyard is coming from, politically speaking.
06:43Well, we are seeing it, though, as an election issue already.
06:46And right now we're in the midterm elections, and all our state lawmakers, the high offices in the state, governor,
06:52senator, you're seeing all those on the ballot as well.
06:54Absolutely.
06:55It's a big issue.
06:57I do think it will be transformational for the economy.
07:00But, look, I think the other thing that's going to happen is we're going to be capacity-constrained as to
07:05how fast this capital gets deployed.
07:07I think whether it's power markets not being able to develop overnight or it's labor challenges with getting these data
07:13centers built,
07:14we've seen some experts point to the fact that they only think that there's going to be a capacity to
07:19build 5%, 6%, 7% power each year,
07:23which is not going to allow the full $3 trillion to get deployed in the next three years.
07:26And so I think there's going to be a natural governor of the capital deployment that's going to smooth some
07:31of the impact to the consumer.
07:32Before I let you go, we haven't gotten into the insurance part of all of this.
07:36So, look, as we talked about public markets versus private markets, a lot of the credit issuance is going into
07:42insurance companies.
07:43It's going into private credit investors.
07:46Because of the nature of the way insurance invests, they invest on a risk-based capital basis.
07:52So if an insurance company puts paper that's single-A rated on its balance sheet,
07:56it only has to post one hundredth of the capital at an equity level to finance that paper,
08:01which means insurance companies investing in private letter rated single-A paper have an insatiable amount of demand.
08:07So it's a home for all of this issuance in AI to go.
08:11All of that's contingent upon a certain rating.
08:14And so even if those insurance companies, these private credit investors, are not taking credit risk,
08:19they're taking rating designations risk.
08:21So we've already seen the NAIC, who's the insurance regulator for credit, start to look at private letter ratings,
08:28start to question or challenge, are these ratings appropriate?
08:31We think data center financing could be at the center of that.
08:34And that's something that we should keep a close eye on.
08:36I think if we saw some significant degradation in that ratings profile within the insurance companies,
08:42I don't think it's a failure issue.
08:44Again, it's not a bubble issue, but it is a capital issue, not a mark-to-market issue.
08:48Trey Parker, we appreciate your time today.
08:49Thanks so much for being with us.
08:51Thank you for having me.
08:52Appreciate it.
Comments