This article doesn't mention it, but PIMCO is lead lender
"Pacific Investment Management Co. (PIMCO) is the anchor lender on the deal. The debt, which matures in 2049, is fully amortising and has been rated A+ by S&P. The bonds were priced at around 225 basis points over U.S. Treasuries."
Ugh. A year ago, I was (a) fairly confident that the AI bubble would burst, but (b) fairly confident that contagion would be limited; a bunch of startups would evaporate, and some VCs would be badly burned or fail, but the broader economy would largely shrug and carry on.
Based on this sort of thing, I'm not sure I still believe (b).
I am 99.0% percent sure that the AI bubble burst is coming. I was only 95% sure until very recently.
Does this mean that this investment spreads the bubble burst risk into people's pension funds? (those lucky enough to have such a thing) Or, not necessarily?
The Bloomberg source is more informative with respect to the financing (https://archive.ph/5WGcA)
> Blue Owl Capital Inc. and Meta will split ownership of the Hyperion data center site in Richland Parish, Louisiana, with the tech giant retaining just 20% of it, according to people with knowledge of the matter. To finance the build-out, Morgan Stanley arranged over $27 billion of debt and about $2.5 billion of equity into a special purpose vehicle
At 225 over (current?) treasuries:
> The bonds priced at about 225 basis points over Treasuries,
The sums being thrown around right now are just staggering. All the money that is being leveraged for AI these days is going to absolutely crush some unlucky sods somewhere. All these massive deals can’t all succeed.
Louisiana residents better expect higher energy costs soon as well that’s for sure. No way entergy is content eating the costs of these investments and the state basically has an allergy to taxing or in any way financially inconveniencing companies.
The way things typically go in LA the only guarantee is the residents will be on the losing end in some form or fashion (also many politicians are probably skimming off plenty for themselves and their buddies). Especially with Landry at the helm.
People wanted massive investments in the electric grid. Thanks to AI boom, that's finally happening. Except, they're also going to be massively using the electric grid.
Maybe we'll be blessed by AI companies making money but staying under usage projections. Then, those areas have more power.
Maybe I’m off the mark here but considering the state Facebook was in until AI put some wind back in its sales I would not say 20 years is a safe bet.
Remember how much time and money they burned on Metaverse? They’ve got nothing to show for it. And wasn’t only like 1-2 years ago they stopped publishing DAU’s in favor of a more favorable metric?
Someone feel free to correct me but they seemed to be just dipping their toes into a bit of a house of cards situation just a couple of years ago.
2049... 20 years. So what exactly is being funded? Just the infra? Walls, power grid, cooling, security cameras and such. Or is any part of this the servers? Is there need for more money in say 2039?
Are the design specifications, like interconnect, GPUs, CPUs, memory and storage of this new cluster public? I seem to recall xAI has made public theirs. I’m mostly asking out of curiosity and a desire to read up on the SOTA hardware specs being used. EDIT: this has some interesting details — https://www.adwaitx.com/meta-prometheus-ai-cluster/
I assume eventually all this investment should result in price drops for cloud GPU rates. Maybe somebody has setup an automated rate aggregator and collected the data? It would be interesting to see the historical data and monitor the changes, like dollars per TFLOP/hr or something standardized to track over time like other economic data or prices.
EDIT: this is along the lines and pretty interesting — https://www.unitedcompute.ai/gpu-price-tracker
I know I’m mixing two different thoughts but they are connected in my head for entrepreneurs interested in starting independent tech/AI/LLM businesses needing heavy compute infrastructure.
The way the recent deals have been structured is that the capacity and thus revenue is pre-booked, which upgrades the credit quality in the eyes of the lenders. Blue Owl is private equity, so they are likely loading the special purpose vehicle itself with the debt (in a way that it off Meta's balance sheet, which is the primary objective), and then possibly funding the capital outlay through secondary markets (if not now, then perhaps later they can bundle it up and sell it - it's Meta adjacent so they will have no problem selling on that debt).
If Blue Owl is providing capital for an equity slice, they get huge leverage on their cut baked into the deal. Pension funds that may end up buying debt in the deal eventually don't want to actually fund the equity of the project and take the risk booting it all up while sitting at the front of the capital stack with corresponding risk of getting wiped out (even if it's a Meta partnership), they simply want securitized fixed income, and make it as vanilla as possible.
So the question is more "will Private Credit (or pension funds/institutions) take debt backed by datacenter collateral with long term service agreements with Meta" and the answer is yes. There is much lower quality stuff than that in the PC space.
But the one thing that doesn’t compute is the commitment. There is a long term obligation now incurred by meta to use this infrastructure. If it’s a capital lease I assume this is now a liability on their books (and disclosures)?
If this data center in Richland Parish, Louisiana, is full of obsolete hardware in a few years, do they just wait for the next hurricane, open all the doors and windows, and phone the insurance company with the bad news?
The article is poorly written. This deal is mostly debt financing with only a little bit of equity.
In terms of corporate capital structure, shareholder returns are usually maximized by taking on at least some debt (leverage). The precise optimal proportion of debt depends on several factors, particularly credit rating.
Make the numbers look better? There must be benefits of moving these numbers from column to some other column. Or even partially hiding them. Thus allowing stock to be priced higher based on some metric...
if you have $100M and you need a $1M, you'd use your credit line and borrow $1M and pay it back from interests coming from $100M. it's not that different in corp.
> "As we learned in December, Meta has commissioned a new natural gas generator plant to be built by local utility operator Entergy. At least for the initial build out, the gas plant would employ three combined cycle combustion turbine generators with a total generative capacity of over 2.2 gigawatts."
This article doesn't mention it, but PIMCO is lead lender
"Pacific Investment Management Co. (PIMCO) is the anchor lender on the deal. The debt, which matures in 2049, is fully amortising and has been rated A+ by S&P. The bonds were priced at around 225 basis points over U.S. Treasuries."
https://pe-insights.com/blue-owl-and-meta-close-record-30bn-...
Oof.. I don't know about this one.
Oh, we're doing this again, are we?
Ugh. A year ago, I was (a) fairly confident that the AI bubble would burst, but (b) fairly confident that contagion would be limited; a bunch of startups would evaporate, and some VCs would be badly burned or fail, but the broader economy would largely shrug and carry on.
Based on this sort of thing, I'm not sure I still believe (b).
What I find fascinating is, everyone is making the same exact points:
1. Yes, there is a bubble, but it's not the same. Companies are profitable.
2. It may burst, but this may go on for another 2+ years. No one knows.
3. Even after the burst, we would have valuable AI infrastructure, just like all the excess internet capacities after the Dotcom.
Base on this, I think the bubble will burst sooner. Sentiment can shift real quick.
Pimco - the company that originated all those bonds every pension fund is sitting on.
I am 99.0% percent sure that the AI bubble burst is coming. I was only 95% sure until very recently.
Does this mean that this investment spreads the bubble burst risk into people's pension funds? (those lucky enough to have such a thing) Or, not necessarily?
6 replies →
The Bloomberg source is more informative with respect to the financing (https://archive.ph/5WGcA)
> Blue Owl Capital Inc. and Meta will split ownership of the Hyperion data center site in Richland Parish, Louisiana, with the tech giant retaining just 20% of it, according to people with knowledge of the matter. To finance the build-out, Morgan Stanley arranged over $27 billion of debt and about $2.5 billion of equity into a special purpose vehicle
At 225 over (current?) treasuries:
> The bonds priced at about 225 basis points over Treasuries,
The sums being thrown around right now are just staggering. All the money that is being leveraged for AI these days is going to absolutely crush some unlucky sods somewhere. All these massive deals can’t all succeed.
Louisiana residents better expect higher energy costs soon as well that’s for sure. No way entergy is content eating the costs of these investments and the state basically has an allergy to taxing or in any way financially inconveniencing companies.
The way things typically go in LA the only guarantee is the residents will be on the losing end in some form or fashion (also many politicians are probably skimming off plenty for themselves and their buddies). Especially with Landry at the helm.
If it doesn't pay off, and becomes just a warehouse space with outdated computers in a few years, who takes the financial hit?
(Meta? Blue Owl Capital's institutional investors? Blue Owl Capital's private investors?)
People wanted massive investments in the electric grid. Thanks to AI boom, that's finally happening. Except, they're also going to be massively using the electric grid.
Maybe we'll be blessed by AI companies making money but staying under usage projections. Then, those areas have more power.
Debt financed. Who is loaning money to these things? I feel there must be an other level of bubble there...
Seems like a stable investment with returns locked in through 2049 unless Facebook defaults.
Maybe I’m off the mark here but considering the state Facebook was in until AI put some wind back in its sales I would not say 20 years is a safe bet.
Remember how much time and money they burned on Metaverse? They’ve got nothing to show for it. And wasn’t only like 1-2 years ago they stopped publishing DAU’s in favor of a more favorable metric?
Someone feel free to correct me but they seemed to be just dipping their toes into a bit of a house of cards situation just a couple of years ago.
2049... 20 years. So what exactly is being funded? Just the infra? Walls, power grid, cooling, security cameras and such. Or is any part of this the servers? Is there need for more money in say 2039?
1 reply →
Hence the big push currently ongoing to allow 401ks to invest in private markets
Wouldnt that be pocket change?
8 replies →
It’s private money, like if you went and asked your uncle for a loan.
Are the design specifications, like interconnect, GPUs, CPUs, memory and storage of this new cluster public? I seem to recall xAI has made public theirs. I’m mostly asking out of curiosity and a desire to read up on the SOTA hardware specs being used. EDIT: this has some interesting details — https://www.adwaitx.com/meta-prometheus-ai-cluster/
I assume eventually all this investment should result in price drops for cloud GPU rates. Maybe somebody has setup an automated rate aggregator and collected the data? It would be interesting to see the historical data and monitor the changes, like dollars per TFLOP/hr or something standardized to track over time like other economic data or prices. EDIT: this is along the lines and pretty interesting — https://www.unitedcompute.ai/gpu-price-tracker
I know I’m mixing two different thoughts but they are connected in my head for entrepreneurs interested in starting independent tech/AI/LLM businesses needing heavy compute infrastructure.
It’s complicated.
Old GPUs (ex hopper, A100) prices has been dropping but the new ones will go up.. so yes it doesn’t need to crash for you to have cheaper gpus
Would love to see the business plan that convinced Blue Owl to sink that much money.
Or maybe they’re ok with the collateral on offer.
The way the recent deals have been structured is that the capacity and thus revenue is pre-booked, which upgrades the credit quality in the eyes of the lenders. Blue Owl is private equity, so they are likely loading the special purpose vehicle itself with the debt (in a way that it off Meta's balance sheet, which is the primary objective), and then possibly funding the capital outlay through secondary markets (if not now, then perhaps later they can bundle it up and sell it - it's Meta adjacent so they will have no problem selling on that debt).
If Blue Owl is providing capital for an equity slice, they get huge leverage on their cut baked into the deal. Pension funds that may end up buying debt in the deal eventually don't want to actually fund the equity of the project and take the risk booting it all up while sitting at the front of the capital stack with corresponding risk of getting wiped out (even if it's a Meta partnership), they simply want securitized fixed income, and make it as vanilla as possible.
So the question is more "will Private Credit (or pension funds/institutions) take debt backed by datacenter collateral with long term service agreements with Meta" and the answer is yes. There is much lower quality stuff than that in the PC space.
This makes a lot of sense.
But the one thing that doesn’t compute is the commitment. There is a long term obligation now incurred by meta to use this infrastructure. If it’s a capital lease I assume this is now a liability on their books (and disclosures)?
1 reply →
Too much money chasing too few investment opportunities with large potential returns. Risk really isn't being acknowledged these days.
$30B ... 2GW datacenter ... ??? ... AGI
Everyone is very silent on AGI in the past months.
The latest from AI is better targeted ads and better adult content didn't you hear?
4 replies →
If this data center in Richland Parish, Louisiana, is full of obsolete hardware in a few years, do they just wait for the next hurricane, open all the doors and windows, and phone the insurance company with the bad news?
Meta has plenty of money, why do they need private equity to fund this?
Isn't the first rule of business that you spend other people's money, whenever possible?
Another #1 is to get a good return on your money.
But tech companies horde cash because they don't have anywhere they see as a good investment.
You'd think investing in their own data centers would get a better return than cash.
Kind of makes you wonder why everyone is so eager to fund these projects for them.
4 replies →
The article is poorly written. This deal is mostly debt financing with only a little bit of equity.
In terms of corporate capital structure, shareholder returns are usually maximized by taking on at least some debt (leverage). The precise optimal proportion of debt depends on several factors, particularly credit rating.
Make the numbers look better? There must be benefits of moving these numbers from column to some other column. Or even partially hiding them. Thus allowing stock to be priced higher based on some metric...
if you have $100M and you need a $1M, you'd use your credit line and borrow $1M and pay it back from interests coming from $100M. it's not that different in corp.
That doesn't make any sense.
You'll be paying a higher rate of interest on your loan than you're receiving on your cash.
You'd be better off taking the $1m directly out of your cash pile.
5 replies →
Dump risk on others
> structured in such a way as to keep the debt off of the Social Network's balance sheets
Is there enough power in the nearby parishes to supply 2.2GW? Also - will there be enough skilled labor?
> "As we learned in December, Meta has commissioned a new natural gas generator plant to be built by local utility operator Entergy. At least for the initial build out, the gas plant would employ three combined cycle combustion turbine generators with a total generative capacity of over 2.2 gigawatts."
Nope, they're not betting on connecting to the grid. They're going to burn natural gas / shale oil.
A little more detail in the linked Bloomberg report. https://archive.ph/2025.10.18-043843/https://www.bloomberg.c...