Google parent company spending like a drunken sailor as capex triples over 2 years

Alphabet, Google's parent company, expects capital expenditures to hit $93 billion in 2025, largely to meet demand from cloud customers, according to its recent financial report. On the back of results that revealed revenue for the third quarter was $102.34 billion, representing 16 percent year-on-year growth, Alphabet …

[ecofeco]

Worrying about overshooting demand?

They already have. But hype train must keep rolling!

[Dinanziame]

Re: Worrying about overshooting demand?

There is very little risk that these data centers will eventually not find a use, whether it is for external customers or internal workloads.

[Like a badger]

Re: Worrying about overshooting demand?

But to undertake what economically valuable task, or meet what social need?

[IGotOut]

Re: Worrying about overshooting demand?

"There is very little risk that these data centers will eventually not find a use, whether it is for external customers or internal workloads."

Incorrect. If they are built for AI, they are pretty much useless for other tasks. Sure, you can not use the incredibly power hungry GPUs, but that's going to be a huge waste of capex.

[Dinanziame]

Re: Worrying about overshooting demand?

If they are built for AI, they are pretty much useless for other tasks

Admittedly I don't know much about the subject, but I severely doubt that their usefulness is so limited. As an example, bitcoin mining companies certainly have hardware optimized for their own specialized use case, but they are apparently able to rent their rigs for AI workloads when mining is not cost effective.

[Excused Boots]

You know, when the bubble pops, it really is going to be quite spectacular, isn't it?

[Like a badger]

We can hope so. Apparently Nvidia's market cap hit $5 trillion recently, and CEO Huang has told everybody "there is no AI bubble".

To put $5 trillion in context, that's the estimated net wealth of Switzerland. Now, faced with the difficult choice of owning Switzerland, or owning Nvidia, I know which I'd choose.

https://www.forbes.com/sites/lewisnunn/2025/08/14/the-worlds-50-richest-countries-2025-according-to-financial-experts/

I did enjoy the little keyboard portrait that Forbes offered us of the Britain.

[Anonymous Coward]

Which of these counties has universal healthcare ??

How concentrated is the wealth to the 0.1%

US and China also both have the largest debt piles.

[VoiceOfTruth]

We will all be paying for this

- More adverts in general.

- More adverts which cost more with the costs factored in to whatever is being sold..

- More software products with the AI label slapped on it, and sold by subscription so pay month in month out whether you use it or not.

- More hardware with the AI label slapped on it. AI lightbulbs, with an AI app on your AI phone.

- AI firewalls to protect against AI-discovered security holes.

- AI dog collars, which detect if your dog wants to go out for a walk.

- AI fraud detection to prevent fraudsters using AI.

- AI products used by BIG COMPANY, with negligible benefits to us, but suddenly cost more.

- AI motor insurance, connected to the AI in your car, putting your premiums up as necessary (but never down).

- AI documentary evidence, of dubious provenance, being debated in court as to how reliable they are (or not).

- AI call handling which miraculously tell the truth: your call is not important to us, there's not more punters in the queue.

- AI predicted weather for tomorrow. For a small monthly fee.

[DoctorPaul]

Re: We will all be paying for this

After years of dental woes I've just had to have some dentures fitted, so I bought some cleaning tablets that you drop into a glass of water.

On the top of the pack is a QR code to "download the app". WTF?!!!

[Bluck Mutter]

It's not like dotcom and it is like dotcom

So the parallels in regard to the AI bubble and the dotcom bubble are valid (in terms of money to be lost).

But what's not valid is AI is not like the growth of internet enabled commerce, supply chains etc.

Internet enabled commerce, supply chains etc didn't exist so being able to hook your PC then later your phone into the internet and as a consumer avail yourself of the services that internet facing business were providing was revolutionary.

Now Internet enabled commerce is a stable (and maybe stale) paradigm.

AI on the other hand (while useful in narrow targeted areas) provides no benefit to consumers... the services that are being touted aren't any better than non-AI services and the ones that provide a novelty factor (like video's of people with six fingers on their hands) will loss patronage as the novelty wears off and/or they start to be charged to make cat videos.

The other factor is the dotcom bust didn't invalidate the dark fibre that was laid (it all was used) nor the generally purpose equipment in the data centres (that all was used as well).

But the AI build out will create white elephants (after the bust) that can't be repurposed but someone still needs to pay the loans (noting that the big players can handle this from existing revenues but it is still a big drag on the bottom line).

And the other difference is the dotcom boom/bust was mainly about core networking so that was all about scaling up existing networking supply chains (i.e. business weren't betting the bank on an unproven technology).

Secondly to this, while at the software level there was a "winner take all" mentality (which did play out), this was all being done by startups thus the levels of investment were not AI scale (i.e. you simple can't invest $US 1 trillion in software development). Now we have large profitable companies still with the "winner take all" mentality and all developing their own AI backend in the hope of winning. But this is all on the back of hardware with little differentiation in the software offerings (i.e. how many chatbots does the world need, how many image generation portals do we need etc).

Unlike the dotcom boom, there can be no winners cause the underlying tech, in the broad sense, has no value. Phoenix's will not raise from the ashes.

Bluck

[Tron]

Short-lived assets.

More focus needed here. Not only is AI a money pit with bugger all ROI, but the shareholders in these companies are buying silicon that depreciates at a phenomenal rate.

But if you want to flush your cash down the lavvy by the sackful, don't let us stop you.

[Anonymous Coward]

nuclear bunkers

possible use of AI datacentres if it does all go tits up. The Russians and Americans are now all testing nukes, DPRK been doing it for a while, there's more phsyco's around today willing to take the risk. They have solar power and backup generators for heating and cooling, just need to reto-fit radiation shielding and somewhere for food supplies and a few septic tanks , hayho jobs a goodun. Oh and getting to one close by might be an issue in the US but UK is ok. There won't be many billionaire's there , because most , if not all , already have one or a few metres underground. I'm sure there's more needed to build a proper one but in the event of an emergency ...

[Dwarf]

Can you change a bit then ?

Great, so you are making a fortune off of cloud computing and AI. Could you back off a bit on the pointless adverts then ?

After all, you don't need the cash and everyone always falsely claims that the adverts pay for everything. You seem to have proven that to be false.