Hardware implementations are common, and likely to become more common as security needs increase, rather than less common. One place they're used is in the digital signature of messages, such as messages going to/coming from a Smart Card used for banking. Those are tiny, slow, limited processors, and a hardware implementation has the potential to improve their performance. Americans aren't used to the slowness that Chip and PIN will bring to the checkout experience, so anything that speeds up that process is going to be of great value.
Security is also becoming more critical for other embedded devices. Smart homes use secure communications to tell door locks to unlatch, and alarms to disarm. And in any battery operated device, such as an alarm sensor, power efficiency is often the top requirement. If a hardware implementation saves a few milliwatts, that could extend the life of a battery in an alarm sensor by a month, which is a huge benefit to the homeowner.
With all the talk of cyber-terror attacks on Industrial Control Systems, the individual controlling hardware devices may soon need to validate signatures on each instruction they receive. They need extremely fast response times to perform real time controls of manufacturing equipment, and must be able to quickly process an emergency shutdown request. Encryption and authentication is always a bottleneck, so a high speed hardware implementation could help maintain near-real-time response times. Look for those algorithms to be baked into the communications chips.
At the other end of the computing spectrum, giant web servers use HSMs to perform the encryption and decryption of thousands of https: sessions per second. Dedicated hardware implementations make such devices far more efficient than general purpose CPUs.
In the middle, you have PCs and laptops, where encryption on the CPU is fast enough for the end users, and the performance benefits of hardware implementations is negligible. But they're also the most subject to malware like viruses, and embedding the security features in a chip could prevent malware from accessing sensitive data like private keys. (If TPM wasn't all about enforcing DRM and other forms of rights restrictions, it might have been useful for actual security functions.)