AI infrastructure investment may be $8T shot in the dark A report from consultancy McKinsey & Company highlights the widespread unease over AI, pointing to the bewildering sums being invested into infrastructure to support it, while warning that forecasts of future demand are based on little more than guesswork. The boom in AI investment has gained momentum over the past year or so …
McKinsey claims that people don't understand AI, that people are unsure about its benefits and:
"McKinsey advises that companies need to assess AI computing needs early, anticipate potential shifts in demand, and design scalable investment strategies that can adapt as AI models and use cases evolve."
This reminds me of what the 'Independent Financial Advisors'* told me when politely asking me to trust them with my pension funds:
'It all depends on what you think the stock market is going to drover the next 30 years'.
If I knew, I wouldn't want an IFA**.
But regarding AI. The article points out that maybe specific AI's can be quite good, but General AI may be really difficult. Anyone reading the Register for the last 5 years could have told McKinsey that.
D'Oh icon, 'coz stating the bleeding' obvious. innit.
* An IFA is basically someone who gets paid to advise you how to lose, sorry, invest, your pension fund and other monies to ensure they, sorry, you, get the highest return.
** I don't. No investor management firm has ever beaten just investing in the FTSE 100 tracker find over the long term. And, as I intent to live for 'the long term
Apologies for replying to my own post, but I have just returned to this site to find: https://www.theregister.com/2025/05/01/x_accounts_europe_drop/
In which is the paragraph:
We asked Musk's AI chatbot, Grok, why Europeans were leaving X, and it responded, "Europeans are leaving the social media platform X primarily due to concerns over misinformation, hate speech, and a perceived decline in content moderation since Elon Musk's acquisition in 2022."
So AI's can get stuff right, after all.
I have not left Xitter, as I was never on it, and never issued a Xit (pronounced, I understand, as 'zit'). Have I missed much?
In the AI rush, those who sell shovels will win.
If the UK won't invest in home grown GPU that could compete with nVidia, then it's very much pointless. We will never be able to compete.
Best we can do is using foreign APIs and compute power we manage to snatch from someone else's jaws.
With that kind of thinking we will never get anything nice.
AMD does compete with nVidia quite well, but Intel is more focused on improving room heaters disguised as CPUs.
You also have to take into account the impact of late stage capitalism where all those three corporations might be owned by the same funds that can put pressure to ensure these entities don't compete too aggressively with each other (burning R&D money instead of paying dividends etc.).
There is quit some work done developing chips that implement AI "neurons" and "synapses" directly in a few CMOS transistors using low energy action potentials. Making for cheaper, more powerful and more efficient chips. Just another thing to keep in mind while building yet another multi $B DC.
Whatever the hardware developments, we already know that whatever you buy today will be worthless in 3 years. And there is no realistic plan to recoup the costs in three years.
I think I could get by with my share of the $8T AI investment being about $1.00. Assuming that I am representative of the rest of the world population (sure!), that would be a total investment of $8 Billion (not $T). It amazes me that they plan to invest $1000 per human on the planet for AI. How about you keep my one dollar, send me the remaining $999, and I will continue to do my own thinking? As an aside, I'd be willing to bet that 99% of the humans would take that deal and run with it.
The LLM sellers are busy generating propaganda to convince executives that they can replace large numbers of workers with their LLM-based products. If they succeed then they will have created a captive customer base that will end up having to pay more than their previous labour costs for LLM product licences, updates, consultancy that will still not be capable of effectively replacing those workers.
Its not unlike the MLM business model that made the USA what it is today - companies that do make the leap of faith will then spend the next couple of years telling all and sundry what a great success it has been and how they'd be fools not to follow suit.. because they'll need to bring their competition down to their new level (an example of "misery needs company").
As I pointed out in a different thread, the built-in obsolescence of the hardware, time taken to get functioning, lack of consumer desire to pay extra as the global economy is trashed by Trump, and lack of any killer app means that data centres for AI are money pits from Day 1. There is no business case for them, even without the taxiffs that will ramp up the cost. The less you spend, the less you lose.
Unless you have a pressing need to show fealty to King Donald by flushing billions down the lavvy to make him happy, it is insane to spend anything on the AI bubble/scam.
Keep your cash in your wallet and wait until the AI bubble bursts/deflates and Trump is gone. If every investment option has more red flags than parade day in Beijing, choose none of them and wait it out. There will always be a good deal out there at some point. Wait until it arrives. Sit back and watch the idiots throw their money away.
… for introducing me to a new experience. McKinsey being right…
While bloody thing is a bubble. Just like sitcom book and the fibre that was invested in at the time.
The outcome was a whole bunch of dark fibre that was then bought at cents on the dollar… though in the case of AI the models will have less value and the kit even less in 3 years time…
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
