Everyone is watching CoreWeave’s debt. 👀
I’m watching what Wall Street is building behind it.
For most of the AI boom, investors have been trained to look for scarcity.
🔥 GPUs
⚡ Power
🏗️ Data centres
🧠 Models
🌐 Networking
But after going down another rabbit hole, I think a different bottleneck is beginning to emerge.
💰 CAPITAL.
AI infrastructure is becoming so expensive that the next phase of the boom may not simply be about who can build the most compute.
It may increasingly be about:
Who can finance it?
Who can structure the risk?
And perhaps most importantly…
Who gets paid to make all this debt investable?
⸻
🚨 THE WARNING SHOT: COREWEAVE
CoreWeave has become one of the clearest examples of what happens when enormous AI demand collides with enormous capital requirements.
Reuters reported roughly US$50 billion of debt. This week, yields on its 2030 bonds jumped roughly 500 basis points to near 16%, while the cost of insuring its debt against default climbed to around 1,200 basis points.
That doesn’t automatically mean CoreWeave fails.
The company also has substantial contracted demand, including a revenue backlog exceeding US$100 billion, and NVIDIA has agreed to purchase unsold capacity through April 2032.
But that’s not the question that interests me.
The bigger question is:
What happens when the AI industry still needs extraordinary amounts of infrastructure… but credit markets start demanding a much higher price to fund it?
Because AI companies aren’t slowing down their ambitions.
They still need GPUs.
They still need data centres.
They still need power.
And somebody has to write the cheque.
⸻
🏦 AI IS BECOMING AN ASSET CLASS
This is where the rabbit hole gets interesting.
In August, NVIDIA announced partnerships with:
Apollo
BlackRock
Blackstone
Brookfield
Goldman Sachs
KKR
The goal is to establish independent compute-financing platforms capable of mobilising more than US$500 billion of third-party capital over time for AI infrastructure.
NVIDIA’s own announcement describes the broader idea as turning compute and AI infrastructure into an investable asset class for global capital. The partnerships remain subject to final agreements.
Read that again.
The world’s dominant AI chip company isn’t only thinking about selling GPUs anymore.
It is helping create the financial architecture capable of funding the customers buying them.
That feels important.
⸻
🧱 THE NEW AI CAPITAL STACK
Then I found an even better example.
Broadcom, Apollo and Blackstone have established an AI infrastructure financing platform designed to enable more than 20GW of compute capacity through 2028.
Its first transaction is an initial US$35 billion capital solution led by Apollo, in partnership with Blackstone and leading global banks, supporting more than 1GW of Anthropic compute infrastructure.
Apollo describes this as a new model for mobilising institutional capital at the scale required by AI infrastructure.
And Apollo isn’t merely arranging the transaction from the sidelines.
Its own disclosure says Apollo-managed funds and affiliates, including Athene, are involved.
That’s where my rabbit hole changed completely.
Because now the story isn’t simply:
AI company → buys GPUs → builds data centre.
It starts looking more like:
AI demand
⬇️
GPUs + data centres + power + customer contracts
⬇️
Financial firms originate and structure the financing
⬇️
Risk gets divided into different layers
⬇️
Institutional capital buys the pieces that fit its risk requirements
Suddenly AI compute isn’t just technology.
It’s becoming financial product.
⸻
🐇💡 MARKET OBSESSED WITH X → I’M WATCHING Y → WHICH COULD CREATE Z
🔥 X: WHAT THE MARKET IS OBSESSED WITH
Right now the conversation is:
“Is AI becoming a debt bubble?”
CoreWeave.
Data-centre leverage.
Hyperscaler capex.
GPU depreciation.
Higher interest rates.
All legitimate questions.
But I think there is another layer.
👀 Y: WHAT MAY BE OVERLOOKED
AI infrastructure is being financialised.
GPUs.
Data centres.
Leases.
Power agreements.
Customer contracts.
Compute payments.
All of these can potentially become pieces of financing structures that allocate different risks to different investors.
Which leads to…
💰 Z: THE UNEXPECTED CONSEQUENCE
Some of the biggest financial winners from the AI infrastructure boom may not build a chip or train a model.
They may be the firms that:
originate the financing
structure the transactions
underwrite the risk
package the assets
distribute them to institutional investors
and then…
do it again.
⸻
🏆 WHO COULD WIN?
Apollo immediately caught my attention.
Not because I’m declaring APO a buy here. I’m not.
But structurally, the company sits in an interesting position.
Private credit.
Infrastructure financing.
Structured finance.
Origination.
Capital markets.
And through Athene, access to long-duration retirement capital.
The Broadcom transaction also shows Apollo’s High-Grade Capital Solutions, Apollo Capital Solutions and ATLAS SP Partners being used together to structure the financing.
Blackstone, KKR, Brookfield, BlackRock and others are playing in different parts of the same ecosystem.
So I’m not ready to declare a winner.
But I’m absolutely watching the people building the financing machine behind AI.
⸻
⚠️ NOW FOR THE PART THAT COULD BLOW THE WHOLE THING UP 😂
There is a reason these structures need sophisticated underwriting.
AI infrastructure carries some nasty risks.
GPUs depreciate.
Technology changes ridiculously quickly.
Compute prices can fall.
Customers can be concentrated.
Financing can outlast the contracts supporting it.
And if AI demand disappoints, today’s incredibly valuable GPU cluster could become tomorrow’s very expensive depreciating asset.
That’s why my thesis is NOT:
Private credit wins because AI needs money.
Way too simple.
My thesis is:
The AI financing winner may not be the firm willing to take the most risk. It may be the firm best at deciding who takes which risk.
That’s a very different business.
⸻
👀 AND THERE MAY BE ONE MORE CONSEQUENCE
This is the bit I’m going to keep watching.
What if the price of AI debt becomes an early indicator for the AI capex cycle?
Think about the chain:
Credit risk rises.
⬇️
Financing becomes more expensive.
⬇️
Marginal data-centre projects become harder to justify.
⬇️
Infrastructure spending slows.
⬇️
Eventually that can flow through to:
semiconductor equipment
memory
networking
power infrastructure
data-centre construction
and potentially GPU orders.
That’s only a hypothesis right now.
But if it’s correct, investors obsessing over semiconductor earnings might eventually discover that some of the earliest warning signals were appearing somewhere else entirely:
THE BOND MARKET.
Because now the story isn’t simply:
AI company → buys GPUs → builds data centre.
It starts looking more like:
AI demand
⬇️
GPUs + data centres + power + customer contracts
⬇️
Financial firms originate and structure the financing
⬇️
Risk gets divided into different layers
⬇️
Institutional capital buys the pieces that fit its risk requirements
Suddenly AI compute isn’t just technology.
It’s becoming financial product.
⸻
🐇💡 MARKET OBSESSED WITH X → I’M WATCHING Y → WHICH COULD CREATE Z
🔥 X: WHAT THE MARKET IS OBSESSED WITH
Right now the conversation is:
“Is AI becoming a debt bubble?”
CoreWeave.
Data-centre leverage.
Hyperscaler capex.
GPU depreciation.
Higher interest rates.
All legitimate questions.
But I think there is another layer.
👀 Y: WHAT MAY BE OVERLOOKED
AI infrastructure is being financialised.
GPUs.
Data centres.
Leases.
Power agreements.
Customer contracts.
Compute payments.
All of these can potentially become pieces of financing structures that allocate different risks to different investors.
Which leads to…
💰 Z: THE UNEXPECTED CONSEQUENCE
Some of the biggest financial winners from the AI infrastructure boom may not build a chip or train a model.
They may be the firms that:
originate the financing
structure the transactions
underwrite the risk
package the assets
distribute them to institutional investors
and then…
do it again.
⸻
🏆 WHO COULD WIN?
Apollo immediately caught my attention.
Not because I’m declaring APO a buy here. I’m not.
But structurally, the company sits in an interesting position.
Private credit.
Infrastructure financing.
Structured finance.
Origination.
Capital markets.
And through Athene, access to long-duration retirement capital.
The Broadcom transaction also shows Apollo’s High-Grade Capital Solutions, Apollo Capital Solutions and ATLAS SP Partners being used together to structure the financing.
Blackstone, KKR, Brookfield, BlackRock and others are playing in different parts of the same ecosystem.
So I’m not ready to declare a winner.
But I’m absolutely watching the people building the financing machine behind AI.
⸻
⚠️ NOW FOR THE PART THAT COULD BLOW THE WHOLE THING UP 😂
There is a reason these structures need sophisticated underwriting.
AI infrastructure carries some nasty risks.
GPUs depreciate.
Technology changes ridiculously quickly.
Compute prices can fall.
Customers can be concentrated.
Financing can outlast the contracts supporting it.
And if AI demand disappoints, today’s incredibly valuable GPU cluster could become tomorrow’s very expensive depreciating asset.
That’s why my thesis is NOT:
Private credit wins because AI needs money.
Way too simple.
My thesis is:
The AI financing winner may not be the firm willing to take the most risk. It may be the firm best at deciding who takes which risk.
That’s a very different business.
⸻
👀 AND THERE MAY BE ONE MORE CONSEQUENCE
This is the bit I’m going to keep watching.
What if the price of AI debt becomes an early indicator for the AI capex cycle?
Think about the chain:
Credit risk rises.
⬇️
Financing becomes more expensive.
⬇️
Marginal data-centre projects become harder to justify.
⬇️
Infrastructure spending slows.
⬇️
Eventually that can flow through to:
semiconductor equipment
memory
networking
power infrastructure
data-centre construction
and potentially GPU orders.
That’s only a hypothesis right now.
But if it’s correct, investors obsessing over semiconductor earnings might eventually discover that some of the earliest warning signals were appearing somewhere else entirely:
THE BOND MARKET.
🐯 MY TAKE
The first phase of the AI boom was dominated by technological scarcity.
Not enough GPUs.
Not enough power.
Not enough data centres.
The next phase may increasingly be dominated by something much older:
THE PRICE OF MONEY.
CoreWeave’s problems don’t prove the AI boom is finished.
If anything, they show why the next stage may require much more sophisticated financing.
And while everyone argues about whether the next winner is NVIDIA, Broadcom, Microsoft or another AI lab…
I’m starting to watch the firms sitting quietly behind them…
building the machinery that finances the whole thing.
Because sometimes the most interesting investment isn’t the company selling the shovel.
It’s the company financing the mine.
💳🤖🏗️💰
Follow the damn@WallStreet_Tiger @TigerPicks @TigerObserver @TigerStars @TigerEvents money.
⸻
🗳️ YOUR TAKE
If AI infrastructure really does become a major institutional asset class, where would you rather have exposure?
A) NVIDIA / AI hardware
B) Apollo / Blackstone / financing & structuring
C) Insurers / long-duration institutional capital
D) None of them. This eventually becomes the next financial accident 😂
Curious what everyone thinks. 👇
🐯 ADZ5150
Just my research and opinion. Not financial advice. Do your own research.
Comments
Great article, would you like to share it?