A closer look at the record-smashing ‘Hyperion’ corporate bond sale

Earlier this month a company you’ve never heard of issued the largest US corporate bond in history, raising $27.3bn in one gulp. The deal wasn’t just notable for its size and complexity, but for what for it says about the AI boom, the credit cycle and the evolution of public and private debt markets.

The company — Beignet Investor LLC — is actually a recently-formed joint venture between Mark Zuckerberg’s Meta and Blue Owl, a $145bn-in-assets private capital firm, to build a massive data centre in Louisiana dubbed “Hyperion”. However, Blue Owl clients will not be providing the bulk of the financing; this comes from various bond funds managed by Pimco. It’s all wild and fascinating, if this sort of thing is your bag.

To get a sense of the bonkers scale envisaged, this post’s image is a graphic that Zuckerberg himself showed in a presentation earlier this summer, which superimposed the proposed structure over Manhattan. Here’s the full thing:

Apparently this possibly isn’t a real demonstration of how big it will actually be in practice, but as a metaphor for Zuckerberg’s acute midlife crisis Meta’s superintelligence ambitions it’s pretty telling.

Meta is already building the first 1-gigawatt data centre called Prometheus in Ohio, which is due to come online in 2026. Hyperion will initially consume over 2 gigawatts, and supposedly up to 5 gigawatts of energy when it’s finished.

That’s enough to power about four DeLoreans. Or, if real-life comparisons are more your thing, five million American homes, according to Deloitte. That’s a lot of power. A 5 gigawatt data centre would require roughly 10 Hoover Dams to keep the lights on.

Anyway, back to the financial details. Meta owns 20 per cent of Beignet and various Blue Owl funds control the rest. Beignet will in turn own a company called Laidley, which will own Hyperion when it is finished. Meta will develop it and be a tenant when it’s finished.

However, Beignet’s financial strength comes from the fact that Meta has committed to leasing Hyperion’s data centres from June 1, 2029 for five lease periods of four years each, for a total for 20 years. Here’s what the structure looks like, courtesy of S&P (zoomable version):

Even if construction takes much longer, Meta will still pay full rent, and only pocket rent credits that will be repaid gradually. Meta will also cover cost overruns beyond 105 per cent of the budget, and there’s a “residual value guarantee” in case it tries to squirrel out of the leases in the future (NB S&P has mistakenly called it a “residential value guarantee” in their schematic).

The proceeds are being invested in Treasuries with maturities that match Hyperion’s scheduled construction period, so that Beignet knows the money will be available when needed.

Meta guaranteeing Beignet’s long term income (if not its debt) meant that S&P assigned the JV an A+ credit rating, just one notch below Meta itself. This in turn helped Beignet’s bankers at Morgan Stanley issue the record-breaking $27.294bn bond on October 16, priced at par and with a coupon of 6.58 per cent.

The bond itself is pretty funky. It is an amortising structure — so the principal is repaid gradually over time rather than having one big bullet payment at the end, as is typical for most fixed income structures — and will mature in May 2049, when the final guaranteed Meta lease expires.

Beignet represents by far the single largest individual high-grade bond ever issued (though there have been larger collective bond packages sold, such as the monster $49bn across eight maturities raised by Verizon in 2013 and the $49bn issued by AB InBev in 2016).

We like big bonds and we cannot lie

Look, if we set aside the whole demented AI capex aspects, this is actually a very cool and fascinating deal that deserves a bit of extra attention.

The original magic of the bond market was being able to efficiently pool the collective savings of the multitudes and funnel the money into grand projects — whether they be wars, canals, railways, skyscrapers or space programmes.

However, the bond market has become fairly standardised in recent years, and is in practice sometimes a bad fit with big, often risky infrastructure projects. Bond investors prefer predictable, fixed cash flows, vanilla structures and securities that will make it into the major indices they use as benchmarks. That can be hard to marry with more creative approaches.

That is why companies mostly issue generic bonds for “general corporate purposes”, and governments sometimes have to step in as sponsors or guarantors of dicier projects. Meanwhile, more exotic, bespoke structures have migrated to the private credit world, where there is — as Alphaville has highlighted many of times — enormous demand and willingness to entertain even very risky, funky stuff.

Meta and Blue Owl’s Beignet structure blurs the line between traditional corporate bonds, project finance and private credit, and shows how the circa $174tn global fixed income market can actually be pretty adaptable — if you’re willing to do a bit of finessing.

JPMorgan’s credit analysts Nate Rosenbaum and Erica Spear agree. They said in a note this week that the deal implies “a growing normalisation of hybrid structures” and argued that it represents a “landmark transaction” for both the traditional public bond market and its funkier private cousin:

From a fundamental credit perspective, we are not aware of a prior A+ rated corporate bond with this degree of structural complexity and single-asset dependence. The Beignet notes are issued by a bankruptcy-remote vehicle, fully secured by the Meta-leased data center, and rely on project-finance-style cash waterfalls, reserve accounts, and mandatory amortization. Credit strength ultimately depends on Meta’s lease and residual value guarantees, not on diversified corporate cash flows.

From a market structure perspective, this is the largest security ever issued in the HG bond market, it is sinkable (amortizing) and it was issued well-wide of any security with equivalent rating and duration in the market at the time. At $27.3bn in size, it is the largest corporate bond ever issued by a significant margin. The largest prior security was $15bn from VZ in 2013 and the largest current tranche is just $8.8bn (ABIBB). It is also secured, which is atypical but not that uncommon in the HG market (approx $758bn outstanding, 8% of the market).

More atypical is the bond’s sinkable nature, only $100bn of sinkable IG debt exists up until now. Most notably, in our view, the now largest bond in the entire market will not be going into the benchmark Bloomberg indices at month-end, on account of its 144A-for-Life status with regards to the US indices. It is also not necessarily going into the benchmark global Agg index, which does include 144A-for-Life structures on account of its status as a “private placement with limited distribution”

In case you’re blissfully unaware of US regulatory jargon, a 144A security one that is sold as a private placement to big sophisticated institutional investors — like an asset manager or insurance company — and is therefore exempt from some of the SEC’s security registration rigmarole. An issuer of an 144A-for-Life bond is, moreover, not required to provide ongoing public financial reports that are typically mandated for public companies.

In many respects the Beignet bond arguably resembles an asset-backed security, or ABS, more than a conventional corporate bond. However, the bankers on the deal presumably decided to structure and market it this way to potentially tap into the far greater universe of traditional investment-grade bond investors.

It seems to have worked. The juicy coupon and solid Meta-backed rating has clearly enticed some investors to overlook the more exotic aspects, sending the price of the bond to 109.47 cents on the dollar at pixel time.

This has pushed the yield down to 5.68 per cent and handed Pimco — Beignet’s biggest backer, having scooped up about $18bn of the bond when it was sold — a massive paper profit that Bloomberg has reported that it is already starting to crystallise.

The ability to take down big complex deals like this represents quite the fillip for big traditional bond houses, at a time when passive investing is beginning to become a major force in fixed income markets as well. As JPMorgan’s analysts wrote: “In the ongoing battle between active and passive investing, this structuring creativity is arguably an argument for active management.”

There are some obvious downsides. Most of all, the low-disclosure 144A-for-Life structure — mostly used for private placements of junk bonds to sophisticated investors — and the limited distribution has denied it entry into Bloomberg’s influential bond indices. That mean that the bond is in practice an awkward fit for a lot of the investors that Beignet presumably wanted to attract.

That said, JPMorgan’s Rosenbaum and Spear note that Bloomberg has opened up a consultation on whether it should in fact be included. They clearly think it should be.

In terms of our views, it seems somewhat incongruous that the largest corporate bond in the market would be index-ineligible for the most widely utilized set of indices, but it is equally incongruous that the largest corporate bond ever ended up being syndicated with ‘limited distribution.’ This in turn re-opens the debate around 144A-for-Life eligibility for the US Agg indices and sub-indices which currently exclude them.

There are valid arguments on both sides of this debate, but given the rapid growth of the HG market and the prevalence of retail buying via funds rather than single securities, we believe 144A-for-Life securities should be included in representative corporate bond indices (this would add approximately $1.2tr to the index).

Alphaville also suspects it ultimately will be allowed into Bloomberg’s indices. The plan was presumably always for Pimco to use its heft to buy a ton of the Beignet bond at issuance to ensure its success, and then gradually let it seep into a wider investor base. Eventually enough will be sufficiently widely distributed and traded to overcome Bloomberg’s primary objection (that this was largely a private placement to Pimco).

However . . . 

Despite Alphaville’s fascination and appreciation for the Hyperion financing package, we also have a lot of not-so-positive notes, observations and questions.

Most obviously, one person’s “creative structuring” can be another’s “stupid, opaque and overly complex financial engineering designed to obscure underlying dogshit”. After all, one of the reasons why the bond market has mostly trended towards simpler structures in recent years is that the Global Financial Crisis exposed a heinous amount of silly, exotic structuring that had happened beforehand.

The Beignet structure is obviously not enormously complex, nor obscuring dogshit. The lease guarantees are strong, and Meta is a very profitable company, despite disappointing third-quarter earnings.

But it’s very clearly a bit of clever financial engineering designed for Meta and its partners to have their cake and eat it. The cute structuring means that Beignet benefits from Meta’s creditworthiness, but Meta’s creditworthiness is magically not impacted by the financial liability that its long-term lease guarantee constitutes — at least according to S&P.

This is . . . curious, especially at a time when problematic off-balance sheet liabilities are back in the news. It’s also unusual for a corporate bond of this size to only have one single rating. JPMorgan notes that only 2.2 per cent of its high-grade universe has a solitary rating. It’s not unheard of for issuers to commission multiple rating agencies to do analysis and then make public only those they like. Alphaville has no information as to whether this has occurred, but it would make a good analyst question.

To avoid having to finance Hyperion off its own balance sheet, Meta is paying quite a bit more than it would have had the money been raised through direct bonds. For example, the yield of Meta’s 2053 bonds is just 5.44 per cent — in other words, over a full percentage point cheaper (though as a conventional bullet-payment bond it’s not exactly comparable to the amortising Beignet bond).

Surely, if Meta is as convinced by the strategic centrality of Hyperion and its other data centre projects to the coming epoch of superintelligence and superabundance, then the best thing would have been to finance it directly without involving Blue Owl, Pimco et al?

Henry Peabody, a former bond fund manager at MFS and Eaton Vance who now writes the Riverhead Research newsletter, argues that the convoluted financing for Hyperion is instead symptomatic of the current market dynamics.

What we seem to be seeing is the convergence of massive need for capital, issuers less willing to hold the residual risk (forming JV’s to finance), and dry powder. Those financing (un)secured debt should be cognizant of the risk they are taking, and push back on terms.

. . . The key aspect to us in the case of Beignet, and project finance in general, is that they are structured for smooth execution. While contractual obligations that often and in this case do, rest on the shoulders of the guarantor, and insurance policies help it gain approval from credit committees, the left tail risk is massive, and unforeseen events can and do lead to a difference of opinion on terms and recovery. It’s all fun and games until a force majeure event, etc.

Indeed, it’s hard to avoid the feeling that for all the Beignet’s merits it this is a sign that the credit cycle is now a little peaky.

While JPMorgan’s Jamie Dimon has warned about credit cockroaches lurking about, the bond market itself remains priced for perfection. The spreads of US junk bonds over Treasuries are near their lowest since 2007. And in the case of investment grade US corporate debt, we are at the tightest since 1997.

Here’s a chart showing the option-adjusted spread of Bank of America’s US investment grade corporate bond index:

And this is despite the average balance sheet health of investment-grade companies having fallen to the lowest in five years, according to UBS’s head of credit strategy Matt Mish.

Peabody is among those that reckon that the Beignet bond therefore represents the exact opposite of what investors should be doing at the moment:

. . . Late cycle credit investing should be up in quality, deeper in security, and generally lower in yield/spread. This issue, which went off without a hitch, and likely bolstered the appeal of the funds that took part, smells a lot like late cycle reach for yield and return to us. We add this to the list of signposts that investors in credit, with asymmetric risk/return profiles, should consider putting capital to risk when that dynamic is flipped on its head.

Its success will invite additional structures — what good is an investment banker if they won’t pitch this to other projects, hailing its success and showing them that the bag is held by someone else — which will bring in more reluctant lenders/debt buyers at less favourable terms. At that stage, the risk has been transferred. We’ve seen this story a million times.

Indeed.