Monday, July 27th 2026

NVIDIA and OpenAI Reportedly Discuss $250 Billion Deal for 10-Gigawatt Ohio Data Center

Jul 27th, 2026 10:59 Discuss (52 Comments)

NVIDIA is in talks to provide roughly $250 billion in financing guarantees for OpenAI's planned data center campus in southern Ohio, according to the Wall Street Journal, cited by Reuters. The project is being developed by SB Energy, a SoftBank subsidiary and is a 10-gigawatt facility sited on a decommissioned uranium-enrichment site about 50 miles south of Columbus, Ohio. The total cost, including chips, could exceed $500 billion making it the largest data center project announced to date. The financing structure would have NVIDIA backstop a set of lending instruments to reassure investors, which is necessary partly because OpenAI hasn't turned a profit and can't qualify for investment-grade credit on its own. The $250 billion commitment covers the lease and construction financing but not the chips. For those it seems that NVIDIA is also in talks to help fund OpenAI's chip purchases, a figure the WSJ puts at up to $350 billion. Remember that last year, OpenAI and NVIDIA teamed up on a huge project to set up 10 gigawatts of NVIDIA systems, the latter planning (at that time) to invest up to $100 billion into OpenAI.

For OpenAI the deal would be its first directly leased data center, reducing dependence on Microsoft, Amazon, and Oracle for cloud infrastructure. The company has since raised its projected computing spend to around $750 billion through 2030. For NVIDIA, backing the project effectively locks in chip demand for years. Power for the site is controlled by the U.S. government and funded separately through a recent trade deal with Japan, which agreed to invest $33 billion in a natural gas plant on the federal land in exchange for lower tariffs. Power allocation is controlled by US Commerce Secretary Howard Lutnick, OpenAI, Anthropic, Microsoft, and Google have all been in contact with him about the site. The first phase is expected to be completed in 2028 at around 800 megawatts. However, the terms of the deal are not final, and it could be scrapped.

Source: Reuters

Add your own comment

52 Comments on NVIDIA and OpenAI Reportedly Discuss $250 Billion Deal for 10-Gigawatt Ohio Data Center

#26

lexluthermiester

AquinusProfitable != revenue. Anthropic recorded what, over 10b in revenue in Q2 alone? That's not a small amount of money and it's been growing rather quickly for many quarters now. For scale, they're expected to rake in 60b in Q3 in ARR alone. That's a 600%/6x increase quarter to quarter. Show me that kind of growth in any sector of the economy right now other then semis.

And none of it profit. They're still in massive debt. OpenAI is severely in debt.

It's just not viable long term. In my opinion, it will collapse, it's just a matter of time.

ChaitanyaSomehow investors in these companies are the dumb lot who keep throwing money down the drain. None of these companies come anywhere close to generating their target revenue let alone making profits.
finance.yahoo.com/technology/ai/articles/openai-pace-miss-own-ad-142128219.html

This. There's just no avoiding the facts.

#27

Aquinus

Resident Wat-man

lexluthermiesterAnd none of it profit. They're still in massive debt. OpenAI is severely in debt.

It's just not viable long term. In my opinion, it will collapse, it's just a matter of time.

This. There's just no avoiding the facts.

We'll see, but there aren't really any signs that this is going to cool off any time soon, as much as we'd love GPU and memory prices to return to some semblance of normal and to be fair, ad revenue isn't where Anthropic and OpenAI make their money.

#28

watzupken

Assuming they have the money to build the data centers, the money is not enough to supply that 10 GW of power. Everything just don't add up, but the lies continue. They are trying very hard to keep investors excited and try and for the CEO and board members to fill their pockets as much as possible before things goes south. Once it blows up, these rich CEOs and board members are fine. Just not everyone else.

#29

Darmok N Jalad

Again, where's the power coming from? A spare 10GW is not just sitting around, waiting to be tapped. It takes years to build generation capacity. Nuclear is a possible option, but it's so heavily regulated that it is very expensive to build and a long time getting it online. No doubt it will be distributed across the county, but that just means more players needing to get capacity built.

#30

AusWolf

When the US government is involved, that's an obviously large scale project. But what is it going to be used for besides sinking money?

#31

wNotyarD

Darmok N JaladAgain, where's the power coming from? A spare 10GW is not just sitting around, waiting to be tapped. It takes years to build generation capacity. Nuclear is a possible option [...]

The biggest nuclears in service don't surpass 8GW. Coal peaks 6.7GW and oil at 8.7GW.
The only plants running higher than 10GW (at peak production, which requires VERY favorable conditions) are some solar parks at China and hydro.

#32

AusWolf

Darmok N JaladAgain, where's the power coming from?

From the article:

Power for the site is controlled by the U.S. government and funded separately through a recent trade deal with Japan, which agreed to invest $33 billion in a natural gas plant on the federal land in exchange for lower tariffs. Power allocation is controlled by US Commerce Secretary Howard Lutnick, OpenAI, Anthropic, Microsoft, and Google have all been in contact with him about the site.

#33

Chomiq

The financing structure would have NVIDIA backstop a set of lending instruments to reassure investors, which is necessary partly because OpenAI hasn't turned a profit and can't qualify for investment-grade credit on its own. The $250 billion commitment covers the lease and construction financing but not the chips. For those it seems that NVIDIA is also in talks to help fund OpenAI's chip purchases, a figure the WSJ puts at up to $350 billion.

Circular economy at its finest.

#34

ixi

Another attempt to keep prices for stock and AI stuff high, lol. Good luck.

#35

Dimitriman

The northern hemmisphere is burning to the ground on 40C heat waves but sure let's build more coal power stations so that chatgpt can make my next Piña Colada recipe.

#36

Mack4285

How come that I, a normal citizen, see nothing good come from AI or its infrastructure? It's bad for the environment and your electricity bill, bad for your job prospects, etc. What is the end goal here that benfits humans and the human society? Probably absolutely none. For sure no universal income and properous lives on a healthy earth. So yeah, what's going on here, really?

#37

Dimitriman

Mack4285How come that I, a normal citizen, see nothing good come from AI or its infrastructure? It's bad for the environment and your electricity bill, bad for your job prospects, etc. What is the end goal here that benfits humans and the human society? Probably absolutely none. For sure no universal income and properous lives on a healthy earth. So yeah, what's going on here, really?

The goal is to increase shareholder value and more leather jackets.

#38

Prima.Vera

10GW of Power ??! Are they nuts?? This will require a powerplant with 10 medium size Nuclear reactors, dedicated just for it! Are they going to build one?? Or where are they going to get all that power from??

Mr. PerfectWe all knew OpenAI weren't turning a profit, but up until now everyone has seemed happy to throw money at them for the promise of future profit.

They are just like Amazon a couple of years ago. Guess what? No profit, means no taxes. All their so called profit is automatically, tax free, re-invested automatically, so the government takes squat. Meanwhile the stocks are sky rocketing and smart people are getting rich.
Callous Capitalism 101.

#39

Vayra86

Space LynxI think they will struggle to convince any local city councils to build anything this big. The cat's out of the bag, and it's pissed.

The cat better be ready able and willing to take actual torches and pitchforks to hand to nuke those datacenters into the ground then, or the cat may get strangled after all.

AusWolfFrom the article:

Let's have faith in Lutnick, the man's proven to be a very capable dude so far, eh. Nothing is backfiring at all from those policies, the US is makin buck and winning the race globally and in trade, as we can all clearly see.

(/s)

Mack4285How come that I, a normal citizen, see nothing good come from AI or its infrastructure? It's bad for the environment and your electricity bill, bad for your job prospects, etc. What is the end goal here that benfits humans and the human society? Probably absolutely none. For sure no universal income and properous lives on a healthy earth. So yeah, what's going on here, really?

Its called a dystopian future not quite unlike what you can see in Cyberpunk 2077. Corpo control over information flows etc. Its all there. You're looking at a dark future inspired by Ayn Rand.

Think of that the next time you open the socials or use services offered by Big Tech. We are keeping this fairy tale afloat by doing so. The only thing keeping it from happening is geopolitics (luckily, but even that is blessing & curse).

#40

JustBenching

Aquinusboth are producing pretty great products that are improving by leaps and bounds practically every quarter. I think a lot of you are looking at this through the lens of a personal consumer. Both companies highly subsidize their subscription plans and is not where they make their money. Enterprises paying API pricing either via tools like Claude Code or Codex is one part, but integrations via AWS Bedrock is another by embedding this tech in things like SaaS solutions is another. This is making these companies a heck of a lot more than the lowly consumer is. So yeah, I'm paying $100/mo. for personal Codex usage, (underwhelming) wooo, but businesses like the one I work for are paying millions to the likes of AWS and Anthropic for these same tools.

Seriously, if you look at the TAM (total addressable market) for any of these businesses, it's measured in the trillions of dollars. That's a huge market and they're gearing up with the intent to control most of it.

Someone finally gets it. Im too bored and lazy to repeat this every other thread so thank you for this. Not that it's going to make a dent. People think AI is for posting cat generated images on instagram or something, good luck changing that.

#41

user556

JustBenchingSomeone finally gets it...

Except there's nothing written there to get. The products are gimmicks, there hasn't been progress, and cost is way way way too high even if there had been progress. Big money from corporate customers is fleeting. Many are already seeing it as money down the toilet.

#42

JustBenching

user556Except there's nothing written there to get. The products are gimmicks, there hasn't been progress, and cost is way way way too high even if there had been progress. Big money from corporate customers is fleeting. Many are already seeing it as money down the toilet.

Ok, what do you think people and companies that pay for AI use it for? Serious question

Clearly you are not working on any field that can use it, and that's fine, but in fields that you can, not using it is shooting yourself. Not in the foot, in the head.

#43

lexluthermiester

user556Many are already seeing it as money down the toilet.

This! Big business is picking up on what many of us tech geeks have been saying all along..

JustBenchingClearly you are not working on any field that can use it, and that's fine, but in fields that you can, not using it is shooting yourself. Not in the foot, in the head.

Oh please.. I have yet to see anything that AI can do better and more accurately than a human. Faster isn't always better...

#44

Random_User

ixiAnother attempt to keep prices for stock and AI stuff high, lol. Good luck.

Mack4285How come that I, a normal citizen, see nothing good come from AI or its infrastructure? It's bad for the environment and your electricity bill, bad for your job prospects, etc. What is the end goal here that benfits humans and the human society? Probably absolutely none. For sure no universal income and properous lives on a healthy earth. So yeah, what's going on here, really?

Yeah! This is an old game! Actually this is the same scam as crypto was. Same scarcity, same struggle, when the cost of token rises with the complexity growth, needing more compute power, using... well, GPUs. To be precise, nVidia ones.

Is this just a coincidence, that the AI scam began, just when nV released it's last batches of LHR cards (which accidentally turned into the regular hash cards, by leaked drivers) during the deadly convulsions of Etherium? Suddenly everyone has got an "urgent need" to use an A.I., as the CEO of the very same company that sold these cards, told so.. The very moment its target NFT/E coin chicken that laid golden eggs, eventually became very dead..

#45

Aquinus

Resident Wat-man

lexluthermiesterOh please.. I have yet to see anything that AI can do better and more accurately than a human. Faster isn't always better...

You clearly don't build software, because I've seen the frontier models take things that would have taken weeks be done in hours and things that used to take hour be done in minutes.

If you don't use Claude Code or Codex regularly for work, then I think you're going to completely miss the point.

JustBenchingOk, what do you think people and companies that pay for AI use it for? Serious question

Clearly you are not working on any field that can use it, and that's fine, but in fields that you can, not using it is shooting yourself. Not in the foot, in the head.

^ This. Businesses wouldn't be dumping millions into this tech if they didn't think it was going to improve productivity and they are demanding metrics to show that it has actually made returns. I've pulled metrics for my teams and 50-90% productivity gain based on features completed and defects resolved is what I'm seeing after Claude Code via Bedrock was introduced.

#46

lexluthermiester

AquinusYou clearly don't build software, because I've seen the frontier models take things that would have taken weeks be done in hours and things that used to take hour be done in minutes.

If you don't use Claude Code or Codex regularly for work, then I think you're going to completely miss the point.

Oh please. Are trying to tell us you're willing to trust AI generated code without debugging, checking it for errors and optimizing? That is still a hands on, human involved procedure. Anyone who says otherwise should not be trusted with mission critical coding, or any coding. To suggest such directly implies careless, lazy incompetence. Evidence you say? Look no further than all of the problems microsoft has been having with updates. Not to mention all the slop coming out of certain Linux repositories.

Your point wasn't a good one.

AquinusBusinesses wouldn't be dumping millions into this tech if they didn't think it was going to improve productivity and they are demanding metrics to show that it has actually made returns.

Corporate/Business execs have proven time and again just how cluelessly incompetent they are. Still not a good point.

#47

Aquinus

Resident Wat-man

lexluthermiesterOh please. Are trying to tell us you're willing to trust AI generated code without debugging, checking it for errors and optimizing? That is still a hands on, human involved procedure. Anyone who says otherwise should not be trusted with mission critical coding, or any coding. To suggest such directly implies careless, lazy incompetence. Evidence you say? Look no further than all of the problems microsoft has been having with updates. Not to mention all the slop coming out of certain Linux repositories.

Your point wasn't a good one.

Read the rest of my edited post. No shit you don't just vibe your way to success with AI. I manage a distributed team of engineers and I have metrics that show that it helps. What do you have? An armchair warrior's opinion? Please. I'm literally on the front lines seeing what this stuff is doing. Also our defect backlog is smaller than it's ever been. AI has even been making it easier to understand the most complicated bugs. Just the other day it discovered a memory leak in a dependency we pull in. I gave it the symptoms and it found the cause in Docusign's Java esign library and proposed a solution in less than an hour. Also Microsoft has been crap well before AI came around. Garbage in, garbage out.

Your post is clear evidence that you don't live in this world.

Tell me again, @lexluthermiester, what do you do for a living? You a software engineer? You manage a team of software engineers? Are you even in the industry at all?

#48

lexluthermiester

AquinusYour post is clear evidence that you don't live in this world.

Yeah, your definition of evidence is lacking to say the least. Amusing though.

AquinusTell me again, @lexluthermiester, what do you do for a living? You a software engineer? You manage a team of software engineers? Are you even in the industry at all?

Seriously? Kinda proves you don't pay attention here. Thank You for proving my point.

However, we're getting off-topic..

#49

JustBenching

AquinusYou clearly don't build software, because I've seen the frontier models take things that would have taken weeks be done in hours and things that used to take hour be done in minutes.

For decompilation - stuff that would takes years take days / weeks.

lexluthermiesterOh please.. I have yet to see anything that AI can do better and more accurately than a human. Faster isn't always better...

Are you sure you are not letting your anti AI bias get in the way of seeing it though? For a hobbyist example, decompilations of games took years to do. Now in a few weeks they fully decompile a game and port it to 10 different platforms. I get access to games on my switch that I could never have without AI. IE. check dusklight project and it's port to horizon.

Obviously no one is suggesting you should go 100% vibe codding. That's ridiculous. Of course it needs human oversight

#50

lexluthermiester

JustBenchingAre you sure you are not letting your anti AI bias get in the way of seeing it though?

I won't deny I have a bias, but it's not irrational. I do see benefits to AI, but not the widespread usage that it's being used for currently. It's being widely abused and that abuse must stop.

JustBenchingObviously no one is suggesting you should go 100% vibe codding. That's ridiculous. Of course it needs human oversight

The problem is that it's happening far too much.

Add your own comment

NVIDIA and OpenAI Reportedly Discuss $250 Billion Deal for 10-Gigawatt Ohio Data Center

52 Comments on NVIDIA and OpenAI Reportedly Discuss $250 Billion Deal for 10-Gigawatt Ohio Data Center

Latest GPU Drivers

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts

NVIDIA and OpenAI Reportedly Discuss $250 Billion Deal for 10-Gigawatt Ohio Data Center

Related News

52 Comments on NVIDIA and OpenAI Reportedly Discuss $250 Billion Deal for 10-Gigawatt Ohio Data Center

Latest GPU Drivers

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts