Master LMEs in Five Short Lessons —
with 9fin’s LME Crash Course

Liability management exercises (LMEs) are every bit as complex as they are prevalent in the US distressed markets. It can be tricky to master the basics, learn the lexicon, and get up to speed with everything going on, especially on top of your existing workload.

Not to worry, though: this online course will ensure you’re well-equipped with the knowledge you need in just five emails.

And it’s curated by Jane Komsky, 9fin's head of LMEs and former distressed lawyer at Kirkland & Ellis and BNP Paribas — and shares tons of free 9fin content along the way.

The course covers:

• The LME basics

• Drop-downs

• Uptiers

• Double dips and hybrids

• 2025 Trends

The course is rolling, so you can learn at your own pace and even save emails to revisit later. Sign up HERE!

Meta and xAI: Project Finance and Lease Accounting 101

$6.5 BILLION DOLLARS. That is roughly how much MORE Meta is paying to raise debt in the private credit market for their new AI data center vs. what they could have paid financing the data center in the traditional corporate credit market. 

So two questions:

1. WHY THE HECK is this financing being portrayed as a win and the playbook for other hyper-scalers like xAI to finance data centers going forward?
   

2. What are the mechanics behind the financing / accounting wizardry that is the rationale behind this structure?

First, let’s set the stage. Meta is building a $30 billion AI data center called Project Hyperion in Louisiana. It’s a massive site (roughly the size of 70 football fields) that will use so much power the local utility, Entergy Louisiana, needs to construct three additional gas-fired power plants to keep up with the demand. 

Instead of tapping the corporate debt markets — which they did, by the way, two weeks later with debt pricing at an interest rate roughly 100bps cheaper — they financed this with project level debt, normally reserved for infrastructure developers like power producers and oil & gas companies, not Big Tech. 

With an interest rate that is ~100bps higher on $27bn of debt with a 24 year term, that is approximately $27bn x 1% x 24 = $6.5bn in additional interest on this debt vs. Meta’s publicly issued corporate bonds. 

So the logical question becomes: why not just raise money via the corporate debt markets?

The answer? Likely because they’re just getting started in terms of AI spend. Per Morgan Stanley’s Head of Global Research, Andrew Sheets, “Most of this AI-related CapEx is still ahead of us, It’s only just starting to ramp up, so this theme will be with us for a while.” Meta knows they are going to need to raise a lot of CAPITAL to finance this future build.  

Therefore, Meta structured this particular data center financing to avoid recording the debt on its balance sheet. It’s an approach that helps keep Meta’s leverage low →  corporate credit rating high → allows them to keep their corporate interest rates low AND preserves balance sheet capacity to raise additional debt in the future. Technically, Meta isn’t the one borrowing this money…they set up a special purpose vehicle (more on that later) that’s ultimately liable for the debt instead.

The second, and arguably more interesting, question is: how does this work? The answer offers us a perfect case study in two foundational finance concepts every investor should understand: project finance and lease accounting.

1. Project Finance 101

When most people think about corporate debt, they picture a company like Meta issuing bonds. All interest and debt repayment is owed by the company and backed by its full balance sheet and credit rating.

Project finance works differently. Here, the debt is NOT RAISED BY THE COMPANY but rather by a “special purpose vehicle” (SPV). Think of it as some outside entity that is NOT Meta. They are the ones borrowing the debt.

Historically this type of financing was done for “boring” industrial type companies. In fact (Kristen here), when I worked in Project Finance at Morgan Stanley in 2010, the majority of the deals I worked on were for companies in the power and utilities or oil & gas space. I worked on power plant financing, wind farm financing, oil field financing etc. However, today companies are taking that same playbook and applying it to big tech.

For Meta’s "Project Hyperion”, project finance lenders don’t look at Meta’s cash flows but rather at the project’s cash flows.

To paint a picture of what a project finance banker is doing, imagine they sit down at their computer and model out the project revenues — in this case, 24 years of lease payments made by Meta to this data center. Assume that’s fixed, locked in, confirmed. You then know the costs to running the data center, the capex, and with decent accuracy can predict the cash flows for say 24 years. Because those revenues are stable and predictable AND lenders get senior claims to the assets of the project, project debt can often support higher leverage and lower yields than general corporate debt.

Remember, lenders care greatly about predictability and stability because they don’t need to pick winners so much as they just need to avoid losers. 

With that said, let’s illustrate how underwriters think about sizing project bonds at a high level (ignoring construction-phase financing):

1. Start with projected cash flows.
   These are the net operating cash flows the project is expected to produce each year once it’s operational, so taking Revenues - Operating Costs - CapEx etc.
   

2. Choose a Debt Service Coverage Ratio, “DSCR”.
   The DSCR measures how much cushion the project has to cover debt payments. A target DSCR depends on the desired credit rating and perceived risk. For example, if you’re targeting a BBB credit, you would work with rating agencies and choose say a 1.2x DSCR. If your target is an A credit rating, that DSCR bumps up to say 1.4.
   
   The DSCR acts as the constraint on leverage. A higher DSCR target means less debt and a larger equity cushion; a lower DSCR target means more debt and less equity.
   

3. Determine the Cash Available for Debt Service (CADS) the project can handle by taking Cash Flow and Dividing by DSCR.
   For example, if projected annual cash flow in year 1 is 100 and the target DSCR is 1.4, you divide 100 by 1.4 to get 71. That means the project can safely support 71 of annual debt service in that one year. You do this for all projected years.
   

4. Discount those debt payments back to today to estimate total debt capacity.
   You then discount the CADS from step 3 using an appropriate interest rate (based on the risks, credit rating, comps etc.) to calculate how much total debt the project can support at the targeted rating. 

Please note, this is wildly oversimplified, but this was the Excel exercise I was asked to perform when I was interviewing with the Project Finance group at Morgan Stanley in 2008!

Whatever portion of the total project cost is not covered by that debt must be funded with equity from the equity sponsors — in this case Meta and Blue Owl. In Meta’s case, that SPV was majority owned NOT by Meta but rather by Blue Owl, a firm that historically has been seen as a private credit investor. Meta owns a 20% stake and agreed to lease the compute capacity once complete.

The key takeaway though, is that Meta didn’t borrow the $30 billion itself. The SPV did. And per US GAAP accounting, since Meta is not the owner, they have only a 20% stake, they don’t have to consolidate this on their books, rather it shows up as an equity investment (another topic for another day).

2. Lease Accounting Wizardry

Here’s where the financial engineering gets clever. Remember, Meta’s not directly borrowing $30 billion. The special-purpose vehicle (SPV) that owns the Hyperion data-center campus is. Meta simply leases the compute capacity once it’s built.

At first glance, that sounds like a simple rent payment. But in accounting, not all leases are created equal. There are two main kinds:

1. Capital leases (now called finance leases): These are long-term arrangements that, economically, look like ownership. The company controls the asset and bears the risks and rewards of owning it. Therefore, accountants make you record the asset and the corresponding debt on your balance sheet – just as if you’d bought the asset outright and borrowed debt to fund it.

   1. You record the asset (PP+E), the liability (capital lease which is treated as debt), and the equity.
      

2. Operating leases: These are shorter-term or cancelable agreements, where the company is just “renting.” The obligation doesn’t sit on the balance sheet – it simply shows up as rent expense on the income statement each year…similar to how you think about leasing an apartment. Your 2-year rental in Manhattan isn’t PP+E, and isn’t considered debt.

   1. No additional PP+E, no additional liability. 

Meta clearly wanted the data center to look like the latter. So instead of committing to a long-term, non-cancelable lease — which would have to be recorded as a capital lease and put the debt on its books — Meta structured its agreements in four-year, cancelable increments to avoid $27 billion of new “debt” showing up in its financials.

So even though Meta is effectively financing and using a data-center asset for decades, the structure allows Meta to stay lower-leverage.

The Broader Trend

Meta isn’t alone here. xAI (Elon Musk’s AI startup) is reportedly exploring similar structures to fund its next-generation compute buildout. The logic is the same: isolate project risk, tap private credit or infrastructure investors hungry for yield, allowing the parent company to maintain additional borrowing capacity and keeping rating high so cost of borrowing low.