Enterprise AI adoption stalls as inferencing costs confound cloud customers

Broader AI adoption by enterprise customers is being hindered by the complexity of trying to forecast inferencing costs amid a fear being saddled with excessive bills for cloud services. Or so says market watcher Canalys, which today published stats that show businesses spent $90.9 billion globally on infrastructure and …

[Pascal Monett]

"it becomes unsustainable from a cost perspective"

But, but . . we've been told for nearly two decades now that The Cloud<supTM</sup> is the best-performing and most cost-effective computing environment there is ! That businesses should all move massively to cloudy platforms because their costs will become negligeable !

Was that all, <gasp>, a lie ?!?

[Paul Crawford]

Re: "it becomes unsustainable from a cost perspective"

"it becomes unsustainable from a customer cost perspective" - fixed it for you.

But it then is sustainable from a snake-oil cloud vendor's perspective.

[Anonymous Coward]

Slowly slowly slowly ... the penny is dropping !!!

The 'AI' Bubble is begining to burst ... the performance and reliability of 'AI' is nowhere near the what was 'pitched'.

It is NOT getting better and we are getting proof that it does not work EVEN if you throw HUGE amounts of money at it.

The hope was that by time the 'Customer' has committed to the 'AI' idea and spent the money needed to implement it, the time to actually build & commission the end product would allow 'AI' to improve and catch up with the hype.

Instead, the hype is still just that 'Hype' and the reality is there for all to see.

Save your money and sanity and kick 'AI' into the dustbin of History ... again ... again (sure there is a pattern here somewhere !!!???) !!!

It still cannot be delivered and is still a 'Sci-Fi' wet dream !!!

What next ...

Flying Cars or Portable Nuclear Reactors or Fusion Reactors ...

maybe we should go the whole hog and start on 'Anti-Gravity Generators' !!!???

There is ALWAYS some greedy 'Investor' who will spend BIG to try to get in on the ground floor of the next Greatest Idea !!!

:)

[abend0c4]

Re: Slowly slowly slowly ... the penny is dropping !!!

There's some evidence it's actually getting worse as it ingests its own garbage.

I think part of the problem is that at first sight the results are really quite plausible and it's easy to imagine it being a panacea. It's only when you become better acquainted that you start to suspect there may be bodies under the floorboards, by which time there's no escape.

Taking a rational perspective, one would assume that investors would be cautious when there were predictions of "500 to 1,000 percent errors of ... cost estimates" or that the datacentres would consume more power than Japan at a time when electricity demand is already increasing owing to the transition from fossil fuels.

However, we've passed the point of rationality. We're now in the broligarch stage where mountains are being hollowed out as personal fiefdoms by very rich men who despise humanity and want to replace it with machines in their own image. They seem to have captured politics, so let's hope the rest of us have the sense to reject this nonsense when it's offered.

[Throatwarbler Mangrove]

Training is a one-time cost?

You mean, you never want your AI to learn anything new? I guess that's what happens when you privilege religious "education" over actual learning.

[nobody who matters]

>......"There is ALWAYS some greedy 'Investor' who will spend BIG to try to get in on the ground floor of the next Greatest Idea !!!".....>

There is ALWAYS some greedy fool who will spend BIG to try to get in on the ground floor or the next plausible con!!!

FTFY

[Anonymous Coward]

Alchemy and snake oil

There was a time alchemy was thing and snake oil sellers were called the new revolution.. There is a lot of similarity to all the jump ups and downs with the latest AI developments. A monster on efficiency system that is swallowing vast resources and energy and can not even mimic a single human brain simple activity powerd with few watts tells a long story. A lot of people will be hurt when this goes kaboom.

[Gnisho]

Re: Alchemy and snake oil

Yeah, the thing is currently the top performing models are *narrowly* useful, and one needs to be intimately familiar with the knowledge domains required for their own work while understanding the current limitations of current LLMs in order to effectively farm out busywork and extract useful time savings from the process. This favors having highly skilled workers over just giving chatbot access to a couple of interns. And yet! It's being hawked as an easy way to boost worker productivity, shave headcount, and pay lower wages to less skilled workers. Maybe some of the could be true in a few more years, but sure isn't now, and model capabilities growth continues to slow even as model sizes and training data repositories balloon. Yet still, "You need to use AI to enhance your toothbrush! Put it on your breakfast cereal! Punch up shareholder meetings! Renegotiate contracts! Handle those incoming lawsuits!" and watch as model hallucinations very visibly crop up at just the wrong moment and the judge in the case asks you about this cite from Pooky Bear. Sorry, suckers.

[Biff Kickass]

Stop screwing this up!

Just lease the voice to someone who is established in the business - just because you have this idea and it's good doesn't mean YOU have the ability to make it real so just lease the voice already so we can be happy!