There's an implied "other things being equal" in there.
Given the commits follow some normal distribution of quality and risk, it's less risky on average to apply one of those average commits than it is to apply a bunch of them.
But if you try to deliver one giant risky commit, you're doing CD in name only.
I would likewise like to know the basis behind these claims. There have been countless instances when I've been reviewing a PR and found changes completely unrelated to the task at hand (user story or bug). One commit often has a scattering of changes that get lumped into one git add * git commit. In a disciplined world full of mindful developers, I could see this being a thing. But I have little faith this could be sustained in a majority of dev shops.
I'm not trying to be negative Nancy but realistic Ron.
Given the commits follow some normal distribution of quality and risk, it's less risky on average to apply one of those average commits than it is to apply a bunch of them.
But if you try to deliver one giant risky commit, you're doing CD in name only.