To put an actual number on it, given some historical precedence, I think $500 million is a good starting point. Back in 2001, eWEEK reported that the estimated cost of W.32 Nimda worm cleanup would hit $500 million. That was 13 years ago; given inflation, the cost of Heartbleed could be much higher, though the truth is that computing costs on the whole today are cheaper than they were in 2001 and significantly more automated.
Whatever the final total figure, Heartbleed is a security incident that is like no other in recent memory. Its potential impact is widespread, and it might take weeks, months or even years until the final true cost is ever tallied.