Google tells employees it must double capacity every 6 months to meet AI demand
arstechnica.com
arstechnica.com
As opposed to external demand, where vastly more compute is needed just to keep up with users torching through Gemini tokens.
Here is the relevant part of the article:
"It’s unclear how much of this “demand” Google mentioned represents organic user interest in AI capabilities versus the company integrating AI features into existing services like Search, Gmail, and Workspace."
I don't really care either way what the source of the demand is -- but it seems like an uncontroversial take.
Semiconductor density, speed, and power efficiency grow much slower than doubling every six months. Creating custom silicon for this won't help - plenty of companies, including Nvidia, are already optimizing their hardware for these tasks and they are very good at it. Production capacity can't scale nearly that fast for myriad reasons, including access to materials, production capacity of inputs, the time and complexity of building new production facilities, the lack of experts available for all of this, and the long time frames for training new people.
This to me is the biggest sign that this is a bubble. Even if demand shrinks, it could still be huge. Even if many use cases are impractical, we will still find some where it's valuable. But the market is basing its valuations on forecasts of tremendous growth that simply can't be supported physically.
Even if you can make the chips, the power supply to run them is very unlikely to grow at that kind of pace.
Most of the demand seems to be driven to by a bubble strategy to sell dollar bills for $0 rather than $1.
So now I am wondering what will be available once the AI investment implodes. I am thinking about computer hardware, available cloud infrastructure employees for hire on the cheap, and more.
I am also looking at the consequences for the incumbents reliant on both AI and cloud that either cannot pay their bills or who fail because their service providers no longer exist. I can’t help but see lots of opportunity on the horizon.
"If the price of inference drops through the floor all the AI wrapper companies become instantly more valuable. Cursor's agents suck ...but their position would get much better with cheap inference." (edited)
"Buy the application layer near winners. When computing costs shrink, usage expands."
"massive Oracle Crypto Farm"
So the answer seems to be: 1. look at what is being priced out now, as this will be cheaper later. 2. Assume the big players will shrink and the smaller players will grow.
If crypto fills the gap I think we could also see a devaluation in coins?
If the data centres stop using so much electricity, will supply of electricity increase and so the costs will decrease?
Personally I'm looking forward to cheaper consumer GPU cards.
They're not trying to build double the data centers every 6 months. But Ars likes the "economic collapse" narrative because I guess their journalist spends too much time on social media.
If a recession comes it likely won't come from AI anyway, and anyhow nothing will happen to these huge tech companies.
Memory/RAM. See also:
They will find another bullshit to use them. Just like crypto to AI transition.
I assume the excess electrical and HVAC gear would be stripped out and sold before converting the tip-up building into another yet another Amazon distribution center.
It’s not particularly useful infrastructure, unlike all of the dark fiber laid during the dotcom boom.
https://blogs.microsoft.com/blog/2025/11/12/infinite-scale-t...
We are pushing the envelope in serving this compute with cost-efficient, reliable power. The Atlanta site was selected with resilient utility power in mind and is capable of achieving 4×9 availability at 3×9 cost. By securing highly available grid power, we can also forgo traditional resiliency approaches for the GPU fleet (such as on-site generation, UPS systems and dual-corded distribution), driving cost savings for customers and faster time-to-market for Microsoft.
- The forced "AI" summary is wrong.
- If you click on "Google AI", it gives a new summary that contradicts the initial one.
- If you check Wikipedia or the top real search result, they contradict both the above.
Should the board intervene and fire Pichai? Does the board know something we don't, e.g., are there massive surveillance contracts with the NSA and the "AI" demand is internal?
IMHO, there are few "big thing" for mass consumer markets and we haven't had one recently.
In chronological order, I'd put PC (90's), internet and online shopping (2000's), smart phones and 'social' media (2010's) and that's it.
I consider something reached "mass consumer market" when a 75 years old grandma considers it normal.
Yes there are many new stuff (online services around phones, smart watches, assistant like Alexa) but nothing that's used by almost everyone.
The easiest way to tell why is that every high schooler or university student is using LLMs (whether it is a good thing or not is irrelevant). These people will go to the job market in a few years and carry the habit.
Notably, not through AI ;)
The bubble thing. Here's more concerning evidence we're heavily focused on a wild and unknown. High risk is high.
This is not a bubble, ChatGPT is the number one app on Android. Many businesses are replacing most of their employees.
No, just to improved surveillance. This new species cannot exist and reproduce on its own.