Just How Big is the AI Buildout - and How Risky? (usnews.com)
- Reference: 0185864120
- News link: https://slashdot.org/story/26/09/25/230252/just-how-big-is-the-ai-buildout---and-how-risky
- Source link: https://money.usnews.com/investing/news/articles/2026-09-24/financing-of-historic-ai-buildout-raises-systemic-risks-in-us-researcher-says
This is pushing up prices for workers, electricity, and even commercial real estate (as well as consumer products that use chips), [2]notes the Wall Street Journal , and reducing the construction on new houses and apartment buildings. And in addition, the paper points out, projections for this buildout "would double the electricity consumption of the entire U.S. residential sector."
The calculations come from Columbia Business School finance/real estate professor Stijn van Nieuwerburgh — and Reuters [3]explains their significance :
> Just as the rail and telecoms expansions led to notable bubbles and busts, Van Nieuwerburgh wrote that the extent of the buildout, the still-untested revenue streams, and the intricate financing structure emerging around AI mean it could be primed for a fall. "This is freaking complicated," he said in a briefing with reporters of the arrangements emerging between AI firms, major tech hyperscalers, banks, private credit lenders, real estate firms, and a host of other players involved in building what he conservatively estimated at 183 gigawatts worth of new data-center capacity over the next seven years, compared with about 57 gigawatts currently installed.... The investment underway already has outstripped what the major players can fund from their own cash flows. The shift to outside financing has increased leverage, redistributed risks across the economy, and made the venture dependent on revenue streams that have yet to be proven, Van Nieuwerburgh noted in the paper, which will be presented on Friday...
>
> "These developments do not imply that financial distress is imminent. Strong growth in AI applications, high utilization, and continued improvements in model capability could support the projected infrastructure and generate stable cash flows," he wrote. "But the combination of uncertain demand, rapid technological change, execution bottlenecks, and high leverage creates meaningful downside risk if expectations are revised." As an example, he wrote that the AI industry will need to be earning about $3.7 trillion in annual revenue by 2032 to achieve the expected return on the investment, and "given current estimates of annual combined revenues of OpenAI and Anthropic of around $100 billion, revenues would need to grow at roughly 80% per year."
The paper suggests policies that "improve measurement and transparency" for financing.
[1] https://www.brookings.edu/wp-content/uploads/2026/09/4c_Van-Nieuwerburgh.pdf
[2] https://www.wsj.com/economy/the-ai-build-out-is-becoming-the-biggest-economic-bet-in-u-s-history-c60716dd
[3] https://money.usnews.com/investing/news/articles/2026-09-24/financing-of-historic-ai-buildout-raises-systemic-risks-in-us-researcher-says
Re: (Score:1)
How does the AI risk compare with the risk of allowing regimes like the IRGC to have nuclear weapons?
You'd think common sense would make you want to go after the easy and obvious things first.
Re: (Score:1)
death cults don't care if they die, as long as they achieve their goals.
You're projecting your rationality onto extreme terrorists.
Re: (Score:1)
Question time. Is this a bot slop coded by a right wing autist or does Israel really care enough to astroturf slashdot?
Re: (Score:2)
I think the bot would have better grammar.
Re: (Score:2)
Dead internet theory in practice here.
Re: (Score:3)
> Voters will get distracted by some little moral panic like trans or heavy metal music
Human nature. If not those, its toxic masculinity, systemic racism, AI datacentres, imminent facism, climate doom, or evil corporations. We love a bogey-man, the "folk devil".
The one thing everybody seems to be worried about, regardless of politics, is perhaps social media algorithms? And China. Two things.
Oracle's New Mexico "Project Jupiter" in trouble (Score:3)
From: [1]Oracle Wants Protection If Its AI Data Center Is Delayed. Here's What That Changes for Investors. [fool.com]
> Bloomberg reported that Oracle has sent a legal notice to the developer of its $X billion "Project Jupiter" AI data center campus scheduled to come online in 2028. It cites "force majeure," a contract provision that offers protection if events outside of a company's control disrupt a deal -- in this case, the delay of a critical natural gas pipeline.
> Oracle is hoping to delay rent payments if the data center is not delivered by the deadline. The developer, a subsidiary of Blue Owl Capital, must first agree that a qualifying event occurred, and Oracle would still be on the hook for certain fees and other payments.
> While Oracle insists the project is on schedule, investors are growing increasingly worried. The $18 billion in project financing for Poject Jupitor have fallen into "stressed" territory. That means the banks that loaned the money are willing to let them go for considerably less than they are worth in name, fearing the loans may go bad. The Financial Times reported that as of last Friday, they had fallen to around 89 to 91 cents on the dollar.
> That's in part due to the stiff local opposition the 1,400-acre campus is facing. Residents are concerned with the environmental impact as well as noise pollution -- a common theme from projects around the country. A recent poll by the Albuquerque Journal found 65% opposed to their construction in the state.
> The company has had trouble with regulators as well. The plan to bring natural gas to the site was blocked by the state's land office, so was a revised plan.
[1] https://www.fool.com/investing/2026/09/24/oracle-wants-protection-if-its-ai-data-center-is-delayed-heres-what-that-changes-for-investors/
Re: (Score:1)
Yes, but think how great the fire sale is going to be for everyone else.
Re: (Score:2)
It would have to be truly amazing to make up for not being to buy RAM for 3+ years.
It’s a seller’s market . . . for now (Score:1)
At the moment, GPU-heavy AI data centers are paying off their silicon within a relatively quick 3 to 4 year time frame.
Compare this to the more typical 4 to 8 years before getting a decent ROI on chips from CPU-heavy “traditional” cloud data centers.
Re: (Score:2)
> GPU-heavy AI data centers are paying off their silicon within a relatively quick 3 to 4 year time frame.
Oh, please. The whole "AI rivalyushon" has only been with us for "3 to 4 years" (since December 2022, when chatgpt launched), and the datacenter gold rush started in late 2024, so we're not even 2 years into it.
Nobody has been in operation for 3 years yet to "pay off the silicon" even by your yardstick, it is all crazy projections based on daydreaming.
The resource misallocation failure that is the "AI" bubble is the most stupid thing that the "huge manatee" could have done and because of the moronic "look
80% and nonusers still have to foot the bill? (Score:2)
You can't reasonably double revenue infinitely. At some point gravity takes effect.
Us non AI users have to pay increased electrical, increased water, and increased computer prices among many other induced problems. These fuckers are why you are seeing less physical switch games as an example
And despite the increased costs for non users, which is never calculated like how inflation is, THE NEED MORE MONEY?
If more people said, Kill Sam Altman, I think we would see more social progress on this devil.
Re: (Score:2)
I'm with you on the computer prices, but not installing rooftop solar was a choice you made.
Unless you're Gen-Z in which case, get your ass over to Reddit where it belongs.
Re: (Score:2)
I've not even fixed my leaking roof yet, you insensitive Claude!
Re: (Score:2)
LOL, that's briliant. I seriously wish I had mod points to mod parent and grand-parent funny. Nevertheless, it gave me a half serios idea for my 100% solar over-production. I consume about half of what I generate and have 20kWh battery backup for overnight. Instead of selling back to the grid for a fraction of their minimum purchase rate (about $0.08/kWh and I get a fraction of that for feeding-in whether it's peak or off-peak time). So, if there are any data centers in-place or under construction with effe
AI first step: Give details about AI design. (Score:2)
The first step in dealing with AI issues is to give full details about how AI is developed.
A HUGE amount has been written about AI, but apparently there are no books or articles about how AI is developed.
If people knew more, maybe there would be contributions to understanding of how to make AI safe.
Re: AI first step: Give details about AI design. (Score:3)
Much of the research is published in peer reviewed journals. There are textbooks and even online tutorials that will walk you through building a toy llm 'from scratch' in a few hours on a reasonably powerful desktop.
Without an undergraduate degree in math, computer science, or similar those will be very difficult to read, but not impossible. You'll have to start from the foundational concepts and work your way up , don't expect a 5-page overview that tells you how to build a Frontier Model. There are als
IPO (Score:1)
IPO, anyone?
Gaming contribution story (Score:2)
When I was in my late teens there was a neighborhood with computer stores.
They would have sheets of specials on components. I would go around and collect all the sheets, find somewhere to sit and compare prices.
Hardware review websites were big at the time (Tom's, etc.) and I knew where all the components ranked. So I made an educated decision before speccing out a system.
Saving a few dollars here and there I would optimize.
Moore's law was big and framerates were low, benchmarks between components would
Bankers are vultures (Score:4, Insightful)
It's kind of amazing how quickly investment bankers noticed a limited resource (video cards, etc) and then injected themselves into the middle of it to get a cut. Insult them if you must, they are amazing at what they do.