It's not just Hawking, you hear variations on this refrain from a lot of people. And given that they're mostly very smart, well educated, well informed people (Elon Musk is another, for example), it probably shouldn't be dismissed out of hand.
Anyway, the basic idea seems to be this: If we create "real" artificial intelligence, at some point, it will be able to improve itself, which improves it's ability to improve itself, which means it can improve it's ability to improve itself even more, and so on... a runaway cascade leading to "superhuman intelligence". That is to say, leading to something that more intelligent than we area.
So what happens if there is an entity on this planet which is literally more intelligent than us (humans)? Would it be a threat to us? Well, it certainly seems reasonable to speculate that it could be so. OTOH, we have no particular reason, right now, to think that it will be so.
So it seems that Hawking, Musk, etc. are just coming down on the more cautious / fearful side of things. Since we don't know if a superhuman AI will be dangerous or not, and given that it could be unstoppable if it were to become malicious (remember, it's smarter than we are!), it's a reasonable thing to take under consideration.
Eliezer Yudkowsky has also written quite a bit on this subject, including come up with the famous "AI Box" experiment. I think anybody interested in this topic should read some of his material.