Not necessarily because they didn't have PCIe (or ISA/NuBus) slots.
Has very little to do with presence, or not, of PCIe/AGP/PCI/NuBUS/etc. and far more about getting so caught up in make a super, duper forward looking box that jump significantly past where most targeted users are in adoption.
By the way, the Cube's motherboard didn't have multiple NUBUS slots on it because the motherboard itself was a Nubus card.
http://simson.net/ref/NeXT/byte_article_page_3.htm
The Mac Cube dies even more quickly because it was even more of a "statement" box than actually trying to solve user problems/needs. Expansion wouldn't have saved the Mac Cube.
[ to a much lessor extent the tossing of the Luxo Jr. iMac design. ]
Why? 3 ports to run the monitors, 1 port to run a 4xHD enclosure, 2 spare (plus USB 3) for portable drives... and that's without daisy-chaining anything.
Because the enclosure is probably $350-400 (presuming it isn't a chain dead-ender. so additional DisplayPort infrastructure inside of box that doesn't do any DP otherwise. ). Throw in the cable and it is even more likely around $400.
If the new Mac is priced similar to the old Mac Pro then at a higher price point. It is only if Apple shifts the new one down around $300-400 that it is not more money for the equivalent system.
Sure could cut costs if that drive box is a dead-ender. But if vast majority of TB products sold are dead enders..... Long term prospective for TB are pretty dismal. Condition users so that they buy increasing larger numbers of those kind of TB devices and this "slap some more controller in and soak up PCI-e lanes" approach goes into a positive feedback loop. That is going to cause problems because the competing alternatives do not have such loops.
Those solutions aren't tossing the integrated functionality of the chipset away. As the chipsets incorporate more functionality that loop to tend throw even more away just to rebuy it again and put it in an external box.