Friends,
Homeowners: do you remember the math that went into buying a house?
You have to consider how much income you have, what your interest rate will be, and -- perhaps most importantly -- what your monthly payments will look like.
Indeed, many people consider the monthly payment to be a more important number than the actual asking price for a house. And that makes sense: it's the part that affects your daily life the most.
Here's one number you probably don't think about too much: the total costs (interest plus principal plus insurance and taxes) you'll pay over 30 years.
If you did, you might never buy a home. And that'd be a shame, because if you multiply any number by 30, it's going to look huge.
Off-Balance-Sheet Debt
It's no secret that the biggest concern investors have about the AI buildout is the amount of debt companies are going into. This isn't just limited to neoclouds. Behemoths like Meta and Amazon are completely using up all of their cash flows to build data centers -- and issuing debt or equity to help pay the bills.
As if that's not scary enough, another term has been gripping investors lately: off-balance-sheet debt.
When people hear this word, they think of evil bankers in a back room taking on debt that will never touch them but will cost shareholders their nest egg.
While we're sure that scenario has played out before, it's important to clear up three key misconceptions about off-balance-sheet debt.
1 - It's not all "debt"
Let's use Meta as an example. Numbers are floating around that the company has nearly $900 billion in off-balance-sheet debt.
That looks incredibly scary. But here's the thing you need to know: none of it is what we'd traditionally consider debt.
- Leases: About $350 billion is future leases (think: rent payments) for data centers that aren't up and running yet. If there's no building to rent yet, there's no obligation to throw on the balance sheet.
β
- Purchase agreements: Another $450 billion is for purchasing things like energy, chips, and compute for providers. Importantly, so long as those providers have the stuff they're supposed to, Meta cannot cancel those orders. But those orders aren't being placed today, so there's no cash outflow right now.
2 - The size of the number only matters relative to cash flows
Here's another part -- and this is where it really relates to our example of buying a house from above: while many of the purchase agreements are set to go into effect over the next few years, the leases will cover up to 20 years.
In other words, while that $350 billion figure looks scary, it's a lot like the all-in cost for a new house: it's not the key figure to consider.
If we spread that $350 billion over 20 years and assume it isn't being discounted (meaning $350 billion in 2046 will be worth much less than it is today), it suddenly looks less scary. That's especially true if we put these payments in perspective.
If we just divide $350 billion by 20 years for a crude proxy, that's about $18 billion per year in lease payments. Here's what Meta's operating cash flow has looked like over the past few years (presented on a trailing twelve-month basis):
While most of that cash is now going towards new data centers today, Meta could cover its lease payments six times over just from its existing business lines. That makes the future debt load seem much less scary.
3 - What's the worst case scenario?
It's worth asking: What's the worst that could happen, and how likely is it?
In one sense, this is unknowable: Black Swans are -- by definition -- impossible to predict. But it would be irresponsible not to consider the worst-case scenario on paper.
In Meta's case, that would be losing $40 billion if its joint-venture mega-data centers in Louisiana and Texas are worth nothing when its leases expire. That would most likely happen if demand for AI completely dries up and the buildings can't be repurposed for anything else going forward.
This is the only part of the off-balance-sheet debt that actually acts like debt -- and it's only about 5% of that really scary $900 billion number.
Given what we know about demand for compute today, those data centers being worth zero in the future seem unlikely. At the same time, even if that event did occur, the aforementioned cash flows referenced above show it would be manageable.
The key conclusion
In the end, this does not mean that Meta investors should ignore off-balance sheet debt. It also doesn't mean that we think the stock is immune from major falls should the demand for AI weaken considerably.
What it does mean, however, is that should Meta end up being a poor investment moving forward, it will be because the company has far less optionality than we once thought. And the market realizing that would lead to inferior results. It would not be "off-balance-sheet debt" that doomed the stock.
Most of the time, we encourage you to see the forest from the trees, keep your eyes trained on the long-term horizon, and take the view from 30,000 feet.
There are times, however, when it pays to examine the details. That's especially true when fear creeps into our thinking. When we do that, we can examine just how scary it is. Over the long run, that keeps us from making knee-jerk reactions that get us out of winning investments.
Wishing you investing success,
|
|
Brian Feroldi, Brian Stoffel, & Brian Withers
Long-Term Mindset
|
β