Why memory and storage have become strategic assets

AI is moving beyond the data centre and into the real world, from semiconductor factories to the cars we drive and the devices we carry.

That's one of many things I learned recording Episode 1 of Why Memory and Storage Matter, my new podcast series with Micron Technology Singapore. It's out now (YouTube).

Beyond the data centre

As AI gets more powerful, memory and storage have become more crucial than ever. Not just in the data centre, which I write extensively about, but also in factories, in the devices we carry, and in the cars we drive.

In this episode, I sat down with Koen De Backer, who leads Smart Manufacturing and AI at Micron, and tech content creator Daren Yoong, to talk about where AI is getting real.

A few things I hadn't considered before talking to them. A single wafer takes 1,500 steps and 90 days to make. Memory for cars has to operate from -40°C to more than 120°C. And industrial AI often relies on proprietary data: a single wafer fab can generate around 78TB of compressed data a day.

Strategic assets

Just a few years ago, I was wondering whether the 64GB of RAM I installed in my last desktop was overkill. Today, I find myself carefully rationing 128GB of RAM between coding, test infrastructure, AI models, and keeping test databases in memory.

I wish I had more.

What is evident is how AI is making memory increasingly specialised. Just as GPUs spawned a generation of specialist chips like TPUs, NPUs and ASICs, AI is now driving demand for specialty memory built for specific workloads.

Memory used to be the component nobody thought much about. Today, both memory and storage have become strategic assets.

Which leaves one question worth sitting with: if you could upgrade only one thing in your setup today, would memory now be the first thing you reach for?