One of the aspects of this job that I find the most stressful is budgeting. I have discussed why budget issues are not very straight-forward in a lab with multiple support streams and projecting years ahead is tied up into that vortex. The current funding climate only exacerbates the issue because of the volatility of funding and the fact that most budgets get cut from the start.
This leaves many PIs asking a simple question for a complex situation: When do I bet on myself?
Fact: Most grant budgets (at least IME for NSF) provide less money than the total cost of the project. A lab can make up the difference in a variety of ways, some of which include finding supplemental streams of cash (e.g. grad fellowships) and others amount to borrowing from Peter to pay Pauline.
In the latter situation, one finds oneself over-spending now to take advantage of opportunities, with the hope that there will be money down the road to make up for the shortfall being created. It's a significant risk with real consequences if there's no reinforcements coming. At the same time, it is exceptionally difficult to make the types of jumps one's research program needs to make in order to stay at the front edge, while playing it safe with the budget. There's new technology, new techniques, new findings, all of which lead you to explore your system in ways you could never have budgeted in when submitting the proposal.
With funding rates in the single digits, the decisions that involve a significant outlay of money for a potentially important piece of the research puzzle are the one that keep PIs up at night. So far, my lab has managed to not have any spectacular failings on big investments and has been able to find those financial safety nets when they were needed. But it doesn't make me feel any better about it when faced with a new costly opportunity. It's a delicate balance with a significant risk of getting burned.