At its foundation, Thunderbolt is exactly what Intel promised in 2009 during its Light Peak demonstration: an interface that can transmit a combination of different data over one cable. However, while initially the technology was to combine USB, Firewire and HDMI, Thunderbolt in the end is based on two other interfaces: DisplayPort and PCI-Express.
DisplayPort is the most recent connection for monitors, which is included on more and more new PCs, laptops and graphics cards. PCI-Express is basically an interface that's used in the PC for controlling various chips and expansion cards. The standard is universal. GPUs, USB controllers, network chips, sound cards, Firewire controllers - you name it, they all exist as PCI-Express-based products. While it's not the first attempt to make PCI-Express available outside of the computer, Thunderbolt is the first standardised version to become available. That means that all the products we just mentioned can be developed in external versions.
Another promise was also realised, and that was the speed of 10 Gigabit per second. That's both ways, so in total a Thunderbolt cable can send 20 Gb/s, clearly more than the 5 Gb/s of USB 3.0.
Thunderbolt is made with copper instead of optical cables. Intel had two reasons for this. The first one is that it's cheaper and the second one that it allows the new interface to power devices. The maximum cable length also became a lot shorter, with three metres instead of the promised 100 metres.
They did keep their last promise, the ability to control multiple devices simultaneously. Not via hubs like we see them with USB, but with a technique called daisy chaining. You can connect a maximum of six devices per port.
The Thunderbolt controller in a system receives a DisplayPort signal from the GPU and PCI-Express from the chipset.