I think that a terrorist attack with a nuclear weapon would be less likely to trigger a world ending counter attack since there is no state to target.
Therefore, I think it's possible that while the chances of a nuclear attack have gone up, the chances of nuclear Armageddon have gone down.
However, I am definitely not an expert so I may be mistaken.
I think it's actually the opposite. The US stockpile consists of mostly (entirely?) thermonuclear weapons. They're smaller for a given yield, and are therefore much easier to deliver. Most of them are also variable-yield, so you can still use a hydrogen bomb to make a smaller boom if that fits your strategic goals.
Edit: it looks like this is the case for the other superpowers as well. The only nuclear state for which this may not be the case is Pakistan, although it's possible that they have a boosted weapon of some sort.
Some of the scenarios I outlined—rogue non-state actors, regional wars—may lead to isolated nuclear attacks but not full exchanges. I’m not sure if the hydrogen bomb distinction is relevant here.
(Yes, afaik what you say is true, but I think this is sort of a red herring.)
When considering the risks of nuclear war, there are (I would say) three principle risks:
1. An intentional first strike by an actor who is non-rational. (E.g., a believer in an apocalyptic ideology who does not respond to traditional deterrence.)
2. An intentional first strike by an actor who believes they can reduce or eliminate their target's second-strike capability (or for whom a nuclear first strike is otherwise a rational response).
3. An accidental launch due to C&C failures. (E.g., https://en.wikipedia.org/wiki/1983_Soviet_nuclear_false_alar...)
I think one can easily argue that two of the three are more likely since the end of the Cold War!
Risk #1 is rather obvious: the rise of international terrorism combined with an increase in the number of (domestically) weak state actors who have nuclear capabilities--post-Soviet Russia, Pakistan--makes the risk of nuclear attack (via stolen weapons) by non-state actors credible.
Risk #2 is the one people normally think of, and, on a large scale--between traditional nuclear powers--it is probably lower than during the Cold War. I would consider two points here, however: first, this might never have been the principle risk during the Cold War (insofar as, with rational actors and full information, deterrence is effective), and second, there are multiple regional powers (again, the usual suspects: North Korea, Pakistan and India) for whom a first strike might be somewhat more tempting.
Risk #3 is also arguably worse. While the total size of deployed unclear arsenals has declined, the number of nuclear states has increased, increasing the odds on a C&C failure for any given nuclear state.
Interestingly, the significant point among all three observations boils down to "nuclear arms proliferation", which, well, duh--that increases the opportunity for mistaken or miscalculated use of nuclear weapons.
More interestingly, the US considering a withdrawal from the Open Skies Treaty goes to another point: information access. As I noted above, deterrence seems to work with rational actors who have perfect information--by reducing access to information, we increase risks.