You're quite right to question this.

Unless they've significantly changed the product since last time I talked to them it's just a chip that does blind signing. It can prevent a key from being leaked, hopefully. But it can't really prevent a key from being misused. Which in most circumstances is pretty much just as bad.

Now, if you could actually run code on the chip and get a remotely verifiable attestation that the code was run correctly, it could be more useful. But it can't. It just blindly signs things provided to it.