News: 1766604186

  ARM Give a man a fire and he's warm for a day, but set fire to him and he's warm for the rest of his life (Terry Pratchett, Jingo)

AI faces closing time at the cash buffet

(2025/12/24)


opinion It is the season of overindulgence, and no one has overindulged like the tech industry: this year, it has burned through roughly $1.5 trillion in AI, a level of spending usually reserved for wartime.

Between 2001 and 2014, the wars in Iraq and Afghanistan [1]cost the US an estimated $1.5 trillion to $1.7 trillion in direct spending. Global AI spending, according to Gartner, is forecast to reach nearly [2]$1.5 trillion this year, putting today's AI boom in the same cash-burning league as two major wars.

What's missing now - like in the last few wars - is a reason for spending this money at all. Some people are starting to ask uncomfortable questions like "When you invest all the money in the bank on a perishable technology, how many years will it take to see a return on that investment?"

[3]

Yep. It's looking like closing time at the Golden Corral's endless buffet and the waiter just brought out the last platter of cash rangoon.

[4]

[5]

So now with a belly full of corporate speak, banking executives who wouldn't know a server if it fell on their wingtips are vomiting "AI" and "agents" all over their press releases, so they can keep the gravy train running.

Consider this recent [6]bon mot from Larry Feinsmith, head of Global Tech Strategy, Innovation & Partnerships at JPMorgan Chase:

[7]

"In the era of AI and agents, the benefits and value will be enormous, but so is the complexity."

You don't say! This is from a guy who went to the Wharton School and now makes more money than most of us so that he can influence billion-dollar companies' business decisions. His advice: "buy more AI!"

Why? There will be enormous value … someday.

[8]

When? It’s complicated.

The real question is: how many versions of that quote did he go through before he settled on that particular platitude? Or did Copilot give him an auto assist?

Without AI's prompting, many of the bankers like him would have had a terrible year and not just when it comes to writing press releases.

Notoriously dour economist with Apollo Global Management Torsten Slok [9]said in October that there is essentially no growth in corporate capital spending "outside of AI," while other economists have said that AI investment is the only thing [10]keeping the US out of a recession.

Why are companies so eager to funnel cash into something that up to now has only produced business value with fancier and fancier chatbots?

[11]Microsoft is building datacenter superclusters that span continents

[12]AI has pumped hyperscale capex, capacity – but how long can it last?

[13]AI datacenter boom could end badly, Goldman Sachs warns

[14]Datacenters feel the Bern as Senator Sanders pushes bit barn building pause

When customers need a quick AI victory, that use case is the first one that tumbles from the mouths of executives at massive infrastructure providers like Dell and Nvidia. Each company has bullishly predicted an AI revolution on par with the advent of electricity, only they can't seem to find a use for this trillion-dollar technological terror except "chatbot" and coming in 2025 2026 the "AI agent!"

Of course, when he's not publicly backing a National Guard deployment in San Francisco, Salesforce CEO Marc Benioff claims his AI agents are already working alongside customers. If so, he's developed an outlier in a field that sees [15]agents fail 70 percent of the time , according to a Carnegie Mellon University study.

This makes the AI agent boondoggle seem more akin to the CEO putting his favorite cousin to work beside you. If they win, you lose. If they lose, you lose.

Forrester said that [16]AI needs to put on a hardhat and get to work if it wants to keep winning deals, or it risks watching that spend get pushed back, with 25 percent of surveyed enterprises delaying AI spend into 2027. Customers, it seems, were beginning to notice that the investments in AI had not resulted in gains to the initials that truly matter in business: EBITDA, earnings before interest, taxes, depreciation, and amortization.

Well, you can’t spell EBITDA without AI, but the trouble is that it's in the ITDA, not in the E. They don’t teach that at Wharton. ®

Get our [17]Tech Resources



[1] https://online.norwich.edu/online/about/resource-library/cost-us-wars-then-and-now

[2] https://www.gartner.com/en/newsroom/press-releases/2025-09-17-gartner-says-worldwide-ai-spending-will-total-1-point-5-trillion-in-2025

[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aUxwhajWe42KKeGUy_8WmQAAAYE&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0

[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aUxwhajWe42KKeGUy_8WmQAAAYE&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0

[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aUxwhajWe42KKeGUy_8WmQAAAYE&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0

[6] https://newsroom.servicenow.com/press-releases/details/2025/ServiceNow-to-acquire-Armis-to-expand-cyber-exposure-and-security-across-the-full-attack-surface-in-IT-OT-and-medical-devices-for-companies-governments-and-critical-infrastructure-worldwide/default.aspx

[7] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aUxwhajWe42KKeGUy_8WmQAAAYE&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0

[8] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aUxwhajWe42KKeGUy_8WmQAAAYE&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0

[9] https://www.apolloacademy.com/ai-boom-debt-vs-equity/

[10] https://finance.yahoo.com/news/ai-is-keeping-the-us-economy-out-of-a-recession-100012367.html

[11] https://www.theregister.com/2025/11/13/microsoft_fairwater_dataceter_superclusters/

[12] https://www.theregister.com/2025/12/22/ai_hyperscale_capex_research/

[13] https://www.theregister.com/2025/12/12/ai_datacenter_investments_goldman/

[14] https://www.theregister.com/2025/12/17/datacenter_ai_bernie_sanders/

[15] https://www.theregister.com/2025/06/29/ai_agents_fail_a_lot/

[16] https://www.forrester.com/blogs/predictions-2026-ai-moves-from-hype-to-hard-hat-work/

[17] https://whitepapers.theregister.com/



Excused Boots

Well as per the question posed by the byline;

yes they absolutely will continue to piss money up the wall, Because nobody wants to be the one to actually laugh and point out that the Emperor has no clothes.

Pointless

Rich 2

I have yet to see any beneficial use of LLMs (let’s not call it “AI” because we all know it isn’t). So far we have…

- chat bots which infuriate any poor sod that calls customer service

- apps that “remove” womens’ clothes in photos

- apps that help fraudsters be better fraudsters

- apps that regurgitate plagiarised prose

- apps that regurgitate plagiarised art

- colossal use of electricity and water

- eating up the world’s supply of RAM

And that’s just the list my shocking memory can recall

Any examples of beneficial “AI” (as far as I have seen) are not LLM based - things like protein folding, drug development, and medical scan analysts - instead they tend to be very targeted, bespoke algorithms that are impressive at what they do, but are not generally applicable to other areas. Unfortunately, this is not the “AI” that is usually talked about (and still not AI anyway!) and certainly not the type that has trillions of your Earth currency thrown at it

Re: Pointless

Bluck Mutter

"Any examples of beneficial “AI” (as far as I have seen) are not LLM based - things like protein folding, drug development, and medical scan analysts - instead they tend to be very targeted, bespoke algorithms that are impressive at what they do, but are not generally applicable to other areas"

When you spunk that amount of money on a dead end, "narrowband" targeted uses of AI won't pay the bills.

The only hope they (AI providers) have is to get "broadband" use inside companies (with the ROI being firing lots of people to pay for it) or getting "broadband" usage in the consumer space.

In the former case, firing people to be able to afford the AI deployment means when it turns to shit, you customers will suffer (more) and the company will suffer reputational damage.

The latter is "pushing shit up hill" cause consumers are used to apps on the WWW being free or low cost. They won't on mass pay the per month fee that the AI companies would need to become profitable. OpenAI say they have 800 million monthly end users but most are paying next to nothing. Ask them to pay $500 per month to actually cover costs/make a profit and the number will plumment.

People say any collapse won't be like the dotcom bust cause the AI big players (aside from the likes of OpenAI) all have money..this is true but the difference here (which will cause issues with the collapse) is established end user companies (not dotcom the startups) could nibble at the edges during the dotcom era.

They could keep all their backend "stuff" running as is and just build an internet facing portal. Thus no disruption... just a realtime API between the existing transactional back end and the web enabled front end. If it all turned to crap, turn off the web front end and all existing business processes/customer facing processes which never went away, continue as before.

Banks are an example of this even today...they still run mainframes for the mission critical transactional engine and you engage with them via the web/phone app which have an API Into the back end.

With AI, they are having to re-engineering your core business processes (like deploying SAP or Oracle EBS...we know how well that goes) so when it turns to shit, you (as a company) can't simply turn it off. And hope the people you fired will come back.

Bluck

Re: Pointless

elsergiovolador

App that *puts* clothes on women could be useful.

Gold rush

Boris the Cockroach

is how I view a.i.

A lucky few to stake a claim at the start will make a bit of money.

However most of the profits will be made by the people selling shovels, carts, and other suppliers.

The vast majority will lose their shirts

Re: Gold rush

iron

In the dotcom era the shovels were sold by Sun, IBM & Cisco.

So how's that going?

I had to re-read the heading.

Yorick Hunt

Thrice!

My brain kept seeing it as "AI Faeces..."

7:30, Channel 5: The Bionic Dog (Action/Adventure)
The Bionic Dog drinks too much and kicks over the National
Redwood Forest.

7:30, Channel 8: The Bionic Dog (Action/Adventure)
The Bionic Dog gets a hormonal short-circuit and violates the
Mann Act with an interstate Greyhound bus.