A brief search showed that the US mint used 694,462.4 gigajoules in 2011 (~0.2Twh). The US is ~25% of world GDP, so lets times this by 4 for a naive cost of minting across the whole world.
This gives a cost of ~1Twh annually for minting the entire world's supply of money, with a total Gross world product of ~$80 trillion.
Bitcoin is using ~1.3Twh for a market cap of ~$35 billion.
Overall this means that bitcoins are ~3000x less efficient than physical money.
After writing this I've realised I should perhaps limit to the cash based economy.
The amount of US dollars in circulation is ~$1.3 trillion.
This gives a slightly better ratio of 222x less efficient.
Bitcoin is 222 times less efficient than a physical system of metal coins and paper/fabric/plastic.
And as sibling comments say, cash uses energy after it's produced.
That it is person to person makes it like cash, but moving a dollar with bitcoin isn't quite as convenient as moving a dollar with cash, whereas moving $5000 with bitcoin is vaguely comparable to moving $5000 using a bank transfer (the value has to be present at the bank/in bitcoin, etc).
Where there is confusion is because economists accept that money is an abstract thing, backed by people's willingness to accept it, whereas Bitcoin enthusiasts see money as a tangible thing that must be backed by a concrete thing. Thus, when economists say that the supply of money changes with fractional reserve banking, they are referring to dollars in the abstract, not physical dollar bills. When Bitcoin enthusiasts say that the supply is limited, they are referring to the bitcoins themselves, not the abstract availability of bitcoins.
No actual US currency is created this way, we just have a strong social convention of treating “the bank owes me $1 on demand” as equivalent to “I have $1”.
So if I make a “deposit” (which is, on point of fact, a loan) of $1, and the bank retains $0.10 in reserve and loans out $0.90 to someone else, we say that I have the equivalent of $1 and the borrower has $0.90, so it seems that $0.90 has been created. But there is really only $1.00 of currency, of which the bank has $0.10 and the borrower has $0.90. I don't have a currency, I have a right to demand (with certain conditions, depending on the kind of deposit) currency from the bank.
Most dollar-denominated trade is actually trade in future claims of dollars rather than actual dollars. Nothing  stops a parallel thing from happening with Bitcoin; obviously, such trade would be distinct from exchanges of Bitcoin recorded in the Bitcoin blockchain, just as trade in future claims of dollars are readily distinguishable from exchanges of physical greenbacks.
 Except the current immaturity of the Bitcoin ecosystem compared to the banking systems of any developed economy, but that's presumably something that would change were Bitcoin to achieve broad, durable acceptance.