The Factory Learned to Borrow
- Rich Washburn
- 1 day ago
- 10 min read
I have been staring at this Ohio thing for hours.
And I am going to say something that sounds completely ridiculous given the scale of what was just announced: I think the data center may be the least interesting part. Which is insane. Because we are talking about an approximately 8-gigawatt OpenAI compute deployment in southern Ohio.
Eight gigawatts.
That is not a typo. This is the kind of number where normally you stop there, write the article about the giant data center, throw a picture of cooling towers on top of it and go home.
Except I kept looking at it. And the more I looked at it, the stranger it got.
OpenAI signs up for long-term capacity. SB Energy builds the physical infrastructure. There is massive new power generation. Billions of dollars of grid work. NVIDIA supplies the compute.
Fine. Huge. Crazy.
Then you get to the part where NVIDIA is also investing in SB Energy and providing enormous credit support around the project. And I kind of froze there. Because... Wait. What?
Haven't We Just Been Talking About This?
A little over a week ago I wrote I Owe My GPUs to the Company Store.
That one genuinely bothered my brain. Not because I thought NVIDIA was doing anything nefarious. It was the shape of the thing. The company selling the GPUs was increasingly involved in financing the ecosystem buying the GPUs.
I remember sitting with that and thinking: Well, that's weird. Not bad weird. Interesting weird. The kind of weird where you realize the economic map you have been using may be missing a road. I ended that piece with essentially one question: At some point, somebody outside this circle still has to create enough economic value to pay for all this stuff.
That felt like the big question. And honestly, I thought I would probably sit with that one for a while.
Apparently not. Because now, barely any time later, I am looking at Ohio.
And the company-store thing suddenly does not look like an interesting little financial anomaly anymore. It looks like one piece of a much larger machine. And I am having one of those moments where the blocks start falling together faster than I am entirely comfortable with.
Pull on the Ohio Thread
At first glance, PORTS-Pike is a data-center story. A gigantic one.
OpenAI has agreed to take roughly 8 IT gigawatts of capacity at the PORTS-Pike Technology Campus. SB Energy will build and operate it. NVIDIA will be the exclusive AI compute provider. The surrounding development includes plans for enormous new power generation and billions of dollars in grid infrastructure.
Again: That alone is nuts.
We should not become numb to these numbers. Eight gigawatts of compute is not some ordinary hyperscale announcement. This is industrial geography changing. But then comes the piece I cannot stop looking at.
NVIDIA is also investing capital into SB Energy and providing credit support around portions of the project. And suddenly the question changes.
It is no longer: Who is going to build all this AI infrastructure?
It becomes: How the hell are we going to finance enough of it?
And I think Ohio may be showing us part of the answer. Maybe This Is Bigger Than the Data Center
I know how stupid that sentence sounds. We are talking about one of the largest AI infrastructure projects imaginable, and I am saying the financing structure behind it may be more important. But follow this for a minute.
A giant infrastructure project does not necessarily require somebody to have the entire project cost sitting around in cash. That is not how civilization normally builds expensive things.
We finance them. You secure future revenue. You turn that future revenue into something lenders can understand. You borrow against it. Then you build the thing now.
Power plants. Pipelines. Aircraft. Railroads. Buildings. Same basic magic trick over and over again: Make the future credible enough that somebody will give you money today.
Nothing revolutionary about that.
What is making me stare at Ohio is the thing we may now be applying that machinery to. Future machine intelligence.
That is where my brain starts doing the thing. Wait. Is This a New Credit Architecture? I want to be careful here because it would be easy to oversell this.
AI did not invent debt. It did not invent project finance. It did not invent guarantees. It did not invent long-term leases. Wall Street has been doing this stuff forever. But I am starting to wonder whether we are watching those old tools assemble themselves around something genuinely new: industrial-scale intelligence production.
That is different. Because what is the actual asset here?
It cannot just be the GPU. The GPU will age. Today's incredible accelerator becomes tomorrow's old accelerator. The model changes. The chips change. The racks change.
But the site? The power? The interconnection? The cooling? The networking? The ability to keep dropping new generations of compute into a giant energized machine?
That persists. And then add a 20-year demand commitment on top of it.
Now you have something finance starts to recognize. That is when I had the moment.
Maybe the thing we are financing is not really the chips. Maybe it is the capacity to keep producing intelligence. And that is a very different idea.
This Goes Back Further Than I Realized A while back I wrote The Gravity Is in the Ground. That came out of watching everyone suddenly rush toward the physical layer. Land. Power. Cooling. Data centers. Interconnection. All the ugly infrastructure underneath the magic.
My conclusion then was that the gravity in AI was moving downward. Software gets copied. Models get surpassed. But energized land with real power attached to it is hard to fake.
Then came From Silicon to Minds.
That was another one of those moments where I stared at the machinery long enough that the abstraction changed.
What do these buildings actually make?
At some level: Electrons go in. Intelligence comes out.
Cognition is becoming an industrial product.
I liked that observation. I thought it was interesting. I did not realize how much it mattered financially. Because now Ohio is sitting there making me ask: If cognition is the product... Can future demand for cognition become something you finance against?
And I think the answer may be: Yeah. Apparently we're starting to figure out how.
That is wild.
Then the Stack Started Showing Itself
In Tracking the Stack, I was following another trail.
Hyperscaler spending. Power shortages. Transmission. Data-center scarcity. Private credit. NVIDIA financing. Infrastructure funds.
The deeper I went, the more all these supposedly separate industries started looking like one thing.
Compute infrastructure. Electrical infrastructure. Financial infrastructure.
At some point I wrote: They are becoming one system.
I meant it. But even then I don't think I fully understood what that sentence implied.
Then in Leopold Was Right. The Terrain Was Bigger Than the Map, the thought got even larger.
Follow the compute downward and eventually you get to power. But follow the power downward and eventually you get to capital. Capital is not beside the AI stack. Capital is part of the AI stack.
Okay. Great. That felt like a satisfying enlargement of the map.
And now I am looking at Ohio and realizing: Shit. Maybe capital was not the bottom either. Because underneath capital is credit. Capital tells you who has money. Credit tells you how much of the future can be pulled into the present.
That is a completely different level of leverage. Credit Makes the Future Spendable
This is the part that I think has crystallized for me.
Power made AI physical. Capital made AI scalable. Credit makes the future spendable today. I keep coming back to that sentence. Because if this infrastructure can become sufficiently bankable, then the size of the AI buildout is no longer constrained only by the balance sheets of the companies building it.
Now you can reach into the global credit system. Project-finance debt. Bonds. Private credit. Institutional capital. Insurance money. Pension capital. All the giant pools of money that do not want to bet on whether the next chatbot wins a benchmark but are perfectly happy to own or finance durable infrastructure with contractual cash flows. And suddenly the scale changes again.
We have spent two years asking: Where are the trillions going to come from? Maybe that was the wrong question.
The financial system does not need trillions sitting in a vault somewhere waiting for an AI company to ask nicely. It needs something it can lend against.
That is much more interesting. Because now AI does not just need investment. It needs to become bankable. And maybe that is exactly what we are watching happen.
The Company Store Looks Different
Now this is where the NVIDIA piece snapped into place for me.
When NVIDIA starts financially supporting the infrastructure that ultimately consumes NVIDIA hardware, the obvious response is to yell "circular financing."
Maybe. There is definitely a loop. But I think that framing may actually be too shallow.
What if NVIDIA is also doing something much more useful to the system? What if it is taking technology risk that banks do not understand and wrapping part of that risk in NVIDIA's balance sheet and expertise?
A lender may have no clue what an accelerator installed in 2029 will be worth in 2034. NVIDIA has a considerably better idea. So NVIDIA can help support residual value, lease obligations, capacity economics or other risks that would otherwise make the project harder to finance.
Suddenly NVIDIA is not merely financing customers. It is helping translate AI infrastructure into a language the credit market understands.
That is a very different animal.
The company store may be turning into something closer to a credit factory. And again: I did not see that coming when I wrote the article.
That is the part of all this that I love. Every time I think I have reached the bottom of the stack, somebody removes another floorboard.
The Really Strange Part
If this works, I do not think the endpoint is simply "more data centers."
I think we may eventually end up with a recognizable financial asset called something like AI infrastructure credit or compute-backed credit.
Not literally a loan secured by a pile of GPUs.
Something broader. A claim built around: power, energized land, cooling, networking, compute infrastructure, long-term capacity demand, hardware refresh cycles, residual value, and somebody credible standing behind the payments.
Once those things become repeatable enough, finance does what finance always does. It standardizes them. Rates them. Pools them. Trades them. Securitizes them. Creates secondary markets. Probably eventually invents derivatives around them because of course it will. And then something that sounds completely insane today becomes boring financial plumbing.
There may eventually be people sitting at Bloomberg terminals trading exposure to long-duration machine-intelligence capacity and nobody will think that sentence is interesting anymore. I find that absolutely bonkers.
But Somebody Still Has to Buy the Gold
There is one thing I do not want to lose in the excitement. None of this means the economics automatically work.
Credit is not magic. You can move risk. You can price risk. You can spread risk. You can move money through time. You cannot make bad economics good forever.
At some point somebody still has to buy the output. Businesses have to become more productive. Scientists have to discover things. Engineers have to design things. Consumers have to pay for services. Governments have to derive enough value from the intelligence.
The economy outside the AI infrastructure loop ultimately has to absorb enough useful cognition to pay the bill.
That was the question in Company Store. It is still the question. Except now the bill is getting very, very large. Which is why Ohio is so fascinating. We may be watching both things happen simultaneously: The infrastructure buildout is getting bigger than almost anybody imagined. And the financial system is evolving fast enough to keep feeding it.
I genuinely do not know where that ends. That is not false humility. I am looking at this the same way I was looking at NVIDIA a week ago: Wow. Okay. What is this thing becoming? Pull the Chain
So that is where I am tonight. Staring at an 8-gigawatt data-center announcement and somehow ending up thinking about credit markets.
Which feels absurd.
Except every time I pull on the chain, another piece comes out of the ground. The ground led to power. Power led to compute. Compute led to cognition. Cognition led to capital. Capital led to credit. And credit may be the mechanism that allows all the other pieces to move at a scale that would otherwise be impossible.
Maybe that is the real story in Pike County. Not that OpenAI is building an enormous AI factory. Although, again: holy shit, OpenAI is building an enormous AI factory.
Maybe the bigger story is that we are starting to understand how an economy finances one. And if that model becomes repeatable, then what happened in Ohio is not simply one giant infrastructure project. It is a template.
That is the part I cannot stop thinking about.
For months I have been following these breadcrumbs mostly because each one seemed strange enough to deserve a closer look. Then another one appeared. And another. And another.
Now I can see enough of them at once that they are starting to look less like breadcrumbs. They are starting to look like a road.
I don't know exactly where it goes yet. But I am increasingly convinced we are watching the financial architecture of the intelligence economy being built in real time. And I have to admit: I did not think it would get this big this fast.
Rich Washburn is a technologist, strategist, and Founder & Chief AI Architect of ARIA AI Labs, working at the intersection of AI, infrastructure, communications, and capital. He also serves as Managing Partner and Chief AI Officer at Eliakim Capital.
Sources
MarketWatch — Nvidia teams with Wall Street firms to help finance $500 billion for AI infrastructure
Earlier pieces in this thread
Comments