As samfisher83 pointed out, $34M was in escrow, so they were probably running on closer to $15M. When setting up a volume production line for a complicated product, it's really easy to burn through tens of millions of dollars. Kickstarter scares the crap out of me for this reason. Typically, I would recommend that a startup begin with a small domestic run of the simplest product possible, targeted at an early adopter market. There's so much a startup has to put into place for hardware: a contract manufacturer (and a sane contract, which takes time to establish), suppliers, supply chain, testing (development and production), certification, packaging, sales channels, distribution, support, reverse logistics, etc. Companies such as PCH, Dragon, etc. can help accelerate some of this, but there's still a ton of work to do. Trying to simultaneously tackle all of that and a complicated engineering design requires a ton of resources--more than a typical startup will have.

Setting up a reliable manufacturing line and test procedure typically takes about 6 months from the time your engineering team begins handoff to the contract manufacturer. Moreover, as much as we'd like it to be, manufacturing is not agile. All of that work is very change-intolerant. So, if you begin the handoff with a half-baked design (which most startups do, in my experience) that you have to change as you proceed through the setup process, you can easily blow away that nominal, 6-month timeline with tooling changes, contract renegotiation, supplier verification, etc. as your engineers scramble to work with the factory to make the design manufacturable and cost-effective.

Jumping straight into a full-blown manufacturing effort is not a cash-effective way for a startup to prove a complicated design. You can force it to work, but it's really expensive. Apple, for example, parallel-paths a lot of stuff. They can afford to discard a million-dollar tool and move on if it doesn't work out. They also send huge engineering teams to Asia for months at a time. Instead, I'd suggest that startups begin with more manual domestic builds in the low hundreds of units until the engineering team is confident in the manufacturability of the design and the manufacturing requirements are understood. These will, of course, have much higher per-unit costs, but you can iterate way, way faster because you're not constantly flying overseas to battle the inertia of a manufacturing line.

Unfortunately, Kickstarter--while still qualifying as an early adopter market--is driving first-run numbers way higher than they should be for the average startup, which is causing them to jump too quickly into scale manufacturing efforts. Because of this, I think we are seeing some spectacular failures such as Lily.

tl;dr: Manufacturing is expensive. Kickstarter is driving hardware startups into manufacturing too soon.