19 Comments
User's avatar
Chris Tottman's avatar

Why did I sell all my data centers 😡 Probably the millions of dollars that was offered 🤣

Joel Salinas's avatar

Hah probably a good idea

Dr Sam Illingworth's avatar

Thanks, Tim, this is a fantastic post that really highlights the need and importance of data centre hubs for the whole of AI infrastructure. However, do you think there's an alternative that might result in less of a need for these centres, for example, with the advent of quantum computing?

Joel Salinas's avatar

Good point! I could see size decreasing as is common in tech, smaller footprint

ToxSec's avatar

“Microsoft plans to invest $80 billion to double its global footprint by 2027. Amazon’s AWS is deploying $55 billion this year, while Meta expects to spend $70–72 billion building AI-ready campuses in North America and Europe. Google has more than 30 new projects under construction worldwide.”

really shows the growth we are talking about here at scale. also, on a personal level, the projects we are working on are basically all ai. ai or die lol.

nice read thanks.

Michael Bridges's avatar

I read another Substack post about how data centers shift their costs onto the local economy.

It took me a while to find it.

I hope it adds value.

https://open.substack.com/pub/hardresetmedia/p/who-pays-for-ais-power-hunger?r=44d6c&utm_medium=ios&shareImageVariant=overlay

Joel Salinas's avatar

That's well written! Thank you for sharing, Michael! And happy New Year!

Michael Bridges's avatar

Happy to help. We need to learn more.

John Brewton's avatar

Power, energy, and compute quietly decide what scales and what breaks.

Dennis Berry's avatar

Yes the ones who control the data and power will control everything moving forward

Joel Salinas's avatar

That’s how the modern world is powered

Raghav Mehra's avatar

Datacenters are the powerhouse of AI systems. A useful read into understanding the basics of it, and also amazing hero image!

Joel Salinas's avatar

They are, and thanks!!

Ashwin Francis's avatar

This is a really interesting read for leaders. Leaders who integrate AI into their operations have to pay for inference. Inference can be directly impacted by energy use, compute utilization, etc.

Joel Salinas's avatar

Yes! It’s important to understand the infrastructure that lets you use ai

Raj's avatar

What will be fascinating to see is if more and more companies provide cloud context/memory as a service via the cloud or if we will continue to see the large providers owning that user relationship.

Given we can all now setup our own MCP servers fairly easily, I think more and more people will setup local desktops that they can tunnel in to, to handle task automation and context retrieval. I know of at least 3 or 4 people in more broader group of friends who have bought a Mac Mini for this reason alone.

Joel Salinas's avatar

That’s a great point, Raj! I self host my n8n automations, for example. That’s all going to increase I imagine

Raj's avatar

Are you self-hosting on a local machine? What’s the setup? I think this will get more common among the crowd who goes one layer deeper than automating agents through UIs.

Joel Salinas's avatar

Self-hosting, just didnt want to pay monthly for an n8n subscription haha