When a big tech wants to build some huge datacenters, where they plan to put hundreds of thousands of ultra-expensive GPUs, they want to run those GPUs as close to 100% of the time as possible. Every hour the GPUs don't run costs them money. From the point of view of Microsoft, having an SMR next to a datacenter makes perfect sense. Solar and wind can do the job, if coupled with batteries and/or natural gas. But than you need a grid operator. If all you need is electricity for a datacenter, and you don't care about being connected to the rest of the grid, then you want as simple a solution as possible. And an SMR promises to be just that, a turn-key solution to get continuous and constant electricity.

▲

bruce511 2 days ago | parent [-]

I think "promises" is the key word there. Data centre's want power, as you say, but they want it now, not 15 years from now.

So yes, when SMR's are "off the shelf" (aka from "order" to producing) , including permitting, construction etc, within a couple years then they are appealing.

I don't think we're quite there yet.

	▲	credit_guy 2 days ago \| parent [-]
		Sure. But this is why what Microsoft did here was just a hedge. It did not cost them much (if anything at all) to become a member of the World Nuclear Association. If the SMRs become reality a few years down the road, and if the demand for datacenters increases significantly because of the increased use of LLMs, then they stand to benefit a lot. If either of this does not pan out, then what's the risk for Microsoft?