Think of the following as a stream of consciousness that may be of some benefit to you gaining perspective on the problems you're about to solve:
 
Thunderbolt will be an advantage for connecting external HDD's and external SSD's. Be prepared to spend a couple of hundred dollars on a multi-bay enclosure if you're planning to use them a JBOD and spend more if you want RAID capability. It's a natural for short-distance connections to video drives and other stuff that needs high multiple terabytes of capacity. 
 
Thunderbolt also offers a latency advantage over USB implementations if you're using DAW-based effects while recording.  For Apollo users, the 'real-time' effects in the UAD DSP of the Apollo makes this less necessary, but there are always things you have in your DAW that cost too much on the UAD platform. Thunderbolt's latency advantage derives from its PCIe roots and the driver/hardware interactions that are very fast via that hardware path.
 
Thunderbolt does *not* offer any practical advantage in channel count for audio strictly due to its superior bandwidth.  There is already plenty of bandwidth even in the lowly USB 2.0.  For example, 128 channels of 24-bit audio at 96KHz sampling rate is still less than 300Mbits/sec.  USB 2.0 has trouble with high throughput due to its half-duplex architecture and the polling scheme.  USB 3.0/3.1 gen1, gen2 does much better in that respect with its dual-simplex (full-duplex) architecture but retains some of the driver baggage of USB 2.0.
 
I'm a very big advocate of Dante protocol.  The Dante PCIe card works at latency values similar to Thunderbolt using gigabit Ethernet.  That 128-channel example that I mentioned above will work on Dante gigabit Ethernet. In addition to the channel count, it has the advantage of larger distances and excellent topology flexibility at very low cost.  Thunderbolt, on the other hand, has short copper links, or expensive and short optical links. I don't know if any Tbolt-3 optical cables exist, but if you need Tbolt-2 to go 100-feet, be prepared to spend around $400.
 
If the Dante product set could get enough market share to drive down the chipset price, it would prove to be exactly what the audio world needs.  The needed growth rate and uptake for Dante is not happening fast enough, and the ascent of AVB is diluting the market, while the 'V' part of AVB doesn't seem to have anything happening in it despite its deleterious effects on Dante.