If it were limited to that, I wouldn't buy one period. Fortunately they can do a lot more. I'd pay probably $225 tops, so picked <$300to see thousands of particles flying around after an explosion?
I do not want this thing on a graphics card though. If they came out tomorrow, I don't want to have to lose the GPU I just purchased recently. If they come out down the road, don't want to put off a GPU purchase waiting for physics to be incorporated. Don't want it splitting GPU horsepower between pixels and physics. If it's just tacked on the graphics card, don't want to get rid of the physics unit just because its time to upgrade the GPU.
I'm looking at the first gen physics chips being supplied along the lines of the Sound Blaster 16. It'll have its own slot and quantifiable capabilities for developers to target. It should be have/don't have, not "well, this game does take advantage of physics processing, but with that particular physics card it'll still be lacking. You want to upgrade to the XPhysXtremeXX-X model." It'll last a good long while as a standard, and be moveable from your current rig to the next one. One base set of performance specs for everyone to adopt during this transatory period, then when the buyers are more savvy and take physics cards to be as standard for gaming as sound cards, can start exploring ways to upgrade.