AI Is Moving Forward by Dragging Us Back to the Basics

As AI pushes forward, are we actually circling back to the basics? Electricity → memory → chips → compute.

To run the most advanced technology of the future, we are becoming dependent again on the most basic pieces of technology.

After watching the AI ecosystem up close for a few months, one pattern feels clear: technology behaves like a pendulum.

We swing toward abstraction, software, and the cloud. Then a new wave of innovation drops us back in front of the physical world’s constraints.

Data centers make the point obvious

Having GPUs is not enough. You need electricity, grid capacity, transformers, cooling, memory, and transmission infrastructure.

We are building the future — and then discovering that future still needs yesterday’s infrastructure to stay on.

The same story in memory

Samsung, SK hynix, and Micron still hold a large share of the DRAM market. China’s CXMT is moving fast. Counterpoint data put CXMT at roughly 9% of global DRAM shipments at the start of 2026. AI demand is only adding pressure.

So the interesting questions are no longer only “who has the most powerful model?” They are also:

• Who can do more with less electricity?
• Who can do more with less memory?
• Who can do more with less compute?
• Who is actually efficient?

Why systems tech like Rust matters here

This is where systems technology such as Rust gets interesting to me. Memory safety plus performance is making it a serious alternative in systems programming — exactly the layer that starts to matter when efficiency becomes the bottleneck.

The chain

AI → more compute → more electricity → more infrastructure → demand for efficiency → hardware and low-level software matter again.

The higher we climb, the more visible the bottom layers become.

A small but meaningful signal from Bangladesh

BUET’s Pragati v1 is an open-source 32-bit RISC-V processor design that was taped out on a 130nm process. It is not an advanced AI chip. That is not the point.

The point is building chip-design capability and an ecosystem. You do not become a semiconductor power overnight. Learning to design, contributing to open ecosystems, and slowly reducing dependency is how bigger things start.

The questions get basic again

1. Where will the electricity come from?
2. Where will the compute come from?
3. Who will make the memory?
4. Who will design the chips?
5. Who will build the infrastructure?

AI is taking us into advanced territory. At the same time, it is reminding us how much raw power and physical infrastructure still decide what is possible.

We are moving forward in technology — and to keep moving forward, we keep returning to the basics.