Andy Halliday

We talked yesterday briefly about there being constraints like structural impediments in the construction industry to building these things out.

But I wanted to talk about this new Microsoft Super Factory.

They're going to double their data center footprint over the next couple of years.

And the newest super factory is a whole campus of two-story structures in Atlanta.

And it's going to have hundreds of thousands of NVIDIA GPUs.

and dedicated high-speed interconnects, and a special set of connections out to the world of the web in order to make this inference-generating system available with very low latency.

Here's the stats on it.

It's a million square feet, which is a lot of square footage for our data center, that covers 85 acres.

Now, an acre is big.

Most homes sit on a quarter of an acre.

And you can pack a lot of server blades in a quarter acre.

But 85 acres, of course, it has lots of other

things that have to be supplied to it, including parking lots for everybody who's gonna be working in that data center.

But anyway, this is just an idea of the scale of the data centers that are being built out.

It's not just Microsoft who's building their own set of data centers.

There's many other players out there doing that.

So it's one of the major construction projects in the world.

Now I wanted to talk about what those data centers, let's just take a second, what those data centers actually cost to operate.

Most of these data centers are not used for training data.

our training models, they're used for inference, right?