Years ago when I worked for a local projector company, we were introducing a relatively compact projector lighted by a Halogen lamp. Cooling such bulb in a big, wide open projector wasn't a problem, but we barely had a few cubic inches of open space around the 400 Watt "heating element" in our projector. Our engineers had to dig up old and nearly lost information about cooling high-powered vacuum tubes in constrained places.
We don't have to worry about cooling glass devices, and big processors, big regulators and other big power components have needed heatsinks and fans for quite a while, so cooling isn't really a new science. But the science of cooling is changing. We're seeing more and more tiny components needing advanced power dissipation techniques.
With our projector bulbs, just sticking a fan next to the bulb wasn't good enough. These new tiny power components, like the MCP1726 regulator, have a similar issue. You can't just stick a heat sink on them and call it good. You need to engineer the cooling system with thermal planes, thermal vias and other layout considerations. Some, like the CMLDM7484, a dual MOSFET from Central Semiconductor, in a 1.7mm x 1.7mm package, ask for aluminum or ceramic core PCBs to survive its maximum power dissipation. Using the PCB for cooling can be a lot more complex than cooling with heatsinks and fans. Anyone remember how to cool a vacuum tube base area and PCB surrounding it?
Kelvons have feet, but photons can fly.