Think of it as an alternative to investing in a house to let out to tenants, but without the inconvenience, risk and cost of buying/selling and owning it.
Another factor is that with decreasing mining reward transaction fees are becoming more and more important for financing the entire system. But by its very nature using Bitcoin as a long term storage will lead to a small number of transaction and the relatively few transactions moving money into or out of long term storage will have to pay for the network resulting in high transaction fees or alternatively the hash rate will have to go down which could itself be problematic for the system.
One would have to run the numbers - how many people are storing how much money for how long - to see how problematic or unproblematic it would be to use Bitcoin primarily for long term storage. My gut feeling however is that without other usages like speculation or actually buying and selling stuff, it would be a rather expensive way to store money long term.
EDIT: Back-of-the-envelope estimate. Bitcoin market capitalization is currently 1000B $, electricity consumption is an estimated 138 TWh/y, 13.8B $/y at 0.10 $/kWh. So if it was all long term storage financed by transaction fees and with a stable price and hash rate, it would cost about 1.5 %/y to store your money in Bitcoin, assuming hardware, infrastructure, labor and so on adds another 10 % to the costs.
This isn't necessary to maintain value (think of a rare painting), but it is important to provide liquidity when you come to sell, so that you don't have to wait weeks for someone to buy your bitcoin. Bitcoin has plenty of liquidity - no shortage of buyer and sellers. Coinbase alone trades ~$1 billion of bitcoin per day.
You're correct that the hash-rate is related to block subsidy + fees, but it's only one of the variables. Cost of energy for miners is an equally import factor (continuously decreasing as miners tap into wasted energy around the earth, and without geographical constraints), along with efficiency of the mining HW.
Bitcoin averages around 2700 transactions per block which currently provides miners with total reward of 6.25 + 0.3 (tx fees) BTC = ~$334,050. If we suddenly went to fees-only today, that would be ~$123 per transaction. Considering that this fee is independent of transaction amount, it is actually very cheap for large values of money (e.g. >$1M) considering it covers the cost of moving the bearer asset around the world within ~30 minutes, and also covers the cost of protecting your value against debasement and theft for however long you had it stored for (often years).
I'm not sure how you came up with a % cost per year - the cost of storage is a one-off tx fee at purchase and sale (just like physical gold, but bitcoin transactions are generally cheaper and independent of the transaction size)
In the end the details don't really matter, currently running Bitcoin seems to cost about 1B $ per month and the users have to pay for that one way or another. Whether there are 1M users with 1M $ each paying 1k $ per month or 1B users with 1k $ and paying 1 $ in fees per month or no one paying any fees and the newly mined coins just diluting the value, that are all details.
It's extremely important to remember that when the fees make up the vast majority of the miner's income, there will be far more people using bitcoin. Ultimately, if bitcoin fulfils it's promise as the best store of value mankind has ever seen, everyone in the world will want some. In this situation, the demand for the 7 tn/sec will be enormous. As humans we're just not used to seeing hard limits on supply of a liquid asset, and it's easy to overlook it's effects. Just as the hard limit on the bitcoin supply issuance is fundamental to it holding its value against essentially anything else (even the gold supply doubles every 30-50 years) and will lead to enormous growth in demand against a falling supply, the fixed supply of transactions will lead to similar increases in the price of transactions due to fixed transaction supply and increasing overall demand.
Once hundred's of millions to billions of people are fighting over transaction space that can only service 150 million transactions a year, the supply/demand ratio will be plenty to support a high price. By that point people would _ideally_ be using it every day/week, and so the potential demand would be enormous, and the transaction price will increase until only their larger transactions are economical.
The real question is: how much hash-rate is really needed?
