Right now, Bitcoin, Ethereum, and other cryptocurrencies provide rewards to people for securing their respective blockchains. These rewards come in two parts: the block reward (for finding the block) plus transaction fees. The long-term vision for Bitcoin (my interpretation, anyway) was that the total value of transaction fees would be so great that mining would be worthwhile without the need for the block reward. Thus, block rewards in Bitcoin are continually shrinking.
In Ethereum, the block reward is 3 ETH; transaction fees come from whatever transactions are mined in the block. The number of transactions can grow if miners choose to let it. Right now, fees are high and add ~10% to the amount miners make per block. Historically, this percentage has been much lower. People were still mining. Thus, there is a subsidy of sorts, in the form of monetary inflation which has paid for miners to secure the network. Individual transactions (pre-Cryptokitties) could be a few cents while the miners were still making a couple/few thousand dollars per block.
In the long run, to support cheap transactions, Ethereum could go the route of Costco/Walmart: large numbers of cheap transactions would help pay for mining. If fees go up, this would support more miners entering the ecosystem. As a side note, the daily profit, denominated in USD, caps miners at around a 150% ROC/PA (it can drop if prices drop, but only briefly exceeds this number). As a side note, I've done the math, but haven't posted it yet (I've got historical data here): the likely cause for this is that if ROC is high, more miners will start; the Nash equilibrium for mineable currencies shows that mining will always underperform simply buying a cryptocurrency if the currency appreciates an average of 5% PA. Thus, anyone trying to maximize profit who expects Ethereum to rise by more than 5% per year should buy ether; anyone who thinks it'll be under 5% should mine (I haven't calculated the lower bound, which would be negative).