AI's $3T infrastructure binge continues despite lack of clear profits
- Reference: 1768398084
- News link: https://www.theregister.co.uk/2026/01/14/ai_investment/
- Source link:
Moody's 2026 Outlook report on the global datacenter market, seen by The Register , forecasts business as usual, with demand for server farm capacity continuing to rise in support of AI, cloud computing, and internet services.
Tech leaders fill $1T AI bubble, insist it doesn't exist [1]READ MORE
The financial services biz estimates at least $3 trillion in investment is required to keep pace with the projected level of capacity expansion between now and the end of the decade. This covers the cost of buildings, the IT infrastructure, and the power required to keep the lights on.
However, the report flags worries over power grid constraints and construction bottlenecks, warning that demonstrating actual revenue generation will become "increasingly important in the AI ecosystem" to silence growing chatter about an "AI bubble."
Question marks over sustainability are hardly new. McKinsey & Company [2]warned in May 2025 that huge sums are being invested in AI on demand forecasts amounting to little more than educated guesswork.
[3]
MIT researchers claimed 95 percent of enterprise organizations [4]had seen no return from their AI efforts so far, and Moody's itself says circular deals involving companies like OpenAI and Microsoft are [5]spooking investors .
[6]
[7]
Capital spending by six hyperscalers in the US – Microsoft, Amazon, Alphabet, Oracle, Meta, and CoreWeave – approached $400 billion in 2025 and is on track to hit $500 billion in 2026, then $600 billion in 2027. A chart in Moody's report shows total global investment peaking in 2029, before declining in 2030.
New datacenter development faces growing headwinds, primarily access to power in most markets, with utilities and power generators hard pressed to meet unexpected electricity demand surges.
[8]
Moody's also notes that [9]local opposition to new builds has increased in some markets, often due to public concerns about their consumption of [10]electricity and [11]water , plus the [12]impact on utility bills .
However, regions with "supportive laws" will continue to attract development, with some governments tweaking regulatory frameworks to encourage AI datacenters. The UK announced " [13]AI Growth Zones " with streamlined planning processes last year.
[14]OpenAI's ChatGPT is so popular that almost no one will pay for it
[15]Meta to sell $30B in bonds to build AI datacenters
[16]Oracle raises AI spending estimate, spooks investors
[17]HSBC spies $207B crater in OpenAI's expansion goals
As the risks mount, developers are coming under growing pressure from clients to shorten construction schedules so that hyperscale tenants can add new capacity as soon as possible.
But this collides with issues including high demand for skilled labor, building materials, and essential equipment, driving up construction costs.
Some tenants are now willing to shoulder delivery risks they previously avoided, including exempting power and utilities availability from completion requirements, the report claims.
[18]
On those [19]circular deals , Moody's notes that OpenAI has signed sizeable contracts for gigawatts of new datacenter capacity and other assets while attempting to grow its revenue base.
Most financing for these assets relies on long-term leases with hyperscalers like Microsoft or Oracle, but "OpenAI's growing presence in the AI ecosystem poses a growing credit risk depending on its success," the report states.
OpenAI [20]suffered a net loss of $11.5 billion or more during the quarter to September 30, 2025, as The Register reported. ®
Get our [21]Tech Resources
[1] https://www.theregister.com/2025/12/05/ai_is_not_a_bubble/
[2] https://www.theregister.com/2025/05/01/ai_dc_investment_gamble/
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aWlxnmUpTMwko5BdQgwySwAAAlU&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[4] https://www.theregister.com/2025/08/18/generative_ai_zero_return_95_percent/
[5] https://www.moodys.com/web/en/us/insights/podcasts/the-big-picture/outlooks-2026-how-ai-and-digital-disruption-is-affecting-credit.html
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aWlxnmUpTMwko5BdQgwySwAAAlU&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[7] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aWlxnmUpTMwko5BdQgwySwAAAlU&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[8] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aWlxnmUpTMwko5BdQgwySwAAAlU&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[9] https://www.theregister.com/2025/06/05/datacenters_have_a_public_image/
[10] https://www.theregister.com/2024/07/25/ireland_datacenter_power_consumption/
[11] https://www.theregister.com/2025/01/04/how_datacenters_use_water/
[12] https://www.theregister.com/2024/11/22/ai_hike_energy_bills/
[13] https://www.theregister.com/2025/01/13/uk_government_ai_plans/
[14] https://www.theregister.com/2025/10/15/openais_chatgpt_popular_few_pay/
[15] https://www.theregister.com/2025/10/31/meta_launches_30_billion_bond/
[16] https://www.theregister.com/2025/12/11/oracle_q2_fy_2026/
[17] https://www.theregister.com/2025/11/26/openai_funding_gap_hsbc/
[18] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aWlxnmUpTMwko5BdQgwySwAAAlU&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[19] https://www.theregister.com/2025/11/04/the_circular_economy_of_ai/
[20] https://www.theregister.com/2025/10/29/microsoft_earnings_q1_26_openai_loss/
[21] https://whitepapers.theregister.com/
But will it happen before the nuclear apocalypse?
I'm the meantime, popcorn futures looks like a good investment
Lunacy
This whole thing is so ridiculous that you could make a new Dr. Strangelove film based on it.
Like many, I can’t wait for it to pop but as has been suggested above, I think another world war coupled with total global ecological collapse will get here before any of the AI protagonists stop smoking whatever the fuck they are on and wake up
Re: Lunacy
Gentlemen, we can’t think in here, this is the AI boardroom
I'll take the under on 2029 being the peak of AI data center investment/construction.
2025 is more likely. No profit = no investment, and the Y2K bubble demonstrated that capex investment can literally turn on a dime.
> Y2K bubble demonstrated
Huh?
Do you mean the Dot-com Bubble?
Otherwise, suggesting around here that the work done to prevent Y2k problems was a bubble might cause some backlash...
Core uncertainty.
To me, the core probablem is that we can do some cool stuff with AI, but we’re not yet to the point where we can just grind or death-march or brute-force compute our way to generally useful, reliable AI.
There are still fundamental breakthroughs needed on the software architecture side and on the hardware side. There are some cool concepts being developed in the hardware space. There might be similar things happening on the software side, but people are less chatty about the nuts and bolts of the bleeding edge there.
Right now, AI is a concept that will revolutionize the world, eventually,
Just like great battery tech (maybe cracked by Donut, time will tell), practical fusion energy, etc. The last 5% to 10% of the work takes 99+% of the time.
You can never predict when fundemental breakthroughs will happen. Took decades for blue LEDs. Lots of money came and went, and it turned out that the one guy who refused to give up eventually got it after spending a significant portion of his life chasing it. That guy should never be allowed to buy his own meals and drinks.
But AI is the same. I'm sure it will happen. Maybe next week. Maybe in 50 years. Nobody really knows.
But they're going to sell it to us / force it down our throats as best they can anyway.
Re: Core uncertainty.
Maybe a useful AI will be developed in our lifetimes, maybe it won't, but I don't think it will look like anything like what's being worked on at the moment.
And I think there are several more AI winters between now and when that happens, if it ever does. "X eventually happened" is no good reason to believe that a totally unrelated Y will also happen.
My forecast is for mass bankruptcy, thousands of lawsuits against OpenAI, Microsoft, and Anthropic for failing to deliver the promised ROI benefits, and pleading to "bail us out - we're too big to fail!"
And I expect it to begin before the end of 2026. It's time to prove that it's a net positive investment or the bubble bursts within six months.
Just think of all the coming shareholder litigation!
Good time to get into corporate law, perhaps.
I look forward to these AI companies imploding and with any luck ceasing to exist in their current form.