One of the oft-made claims about Bitcoin and its blockchain transaction ledger is that they make transactions really cheap, so you can pay someone anywhere in the world for free, or close to it. But when you look closer, is that really true? Not by a long shot.

Bitcoin transactions are stored in a large shared database called the blockchain. Since new transactions are added to the end and never removed, the blockchain keeps growing, currently close to 70 gigabytes, growing at several megabytes per hour.

The blocks in the blockchain are created by miners, which publish a new block every time they find an answer to a difficult cryptographic puzzle. When they publish a block, they can and usually do also include recent transactions from other people. Miners have two incentives: they get a fixed reward of bitcoins for each block, and they can get a commission from the transactions they publish. In practice, the miners get about 99% of their bitcoins from the fixed reward, and 1% for the commissions. The fixed number is currently 25 bitcoins/block, and drops by half every 210,000 blocks. There are currently about 410,000 blocks, with a new block about every 10 minutes, so the fixed number will drop to to 12.5 when there are 420,000 blocks, which will be sometime in July.

Why do miners mine? Because people are willing to pay for the newly mined bitcoins. By all reports, mining is at best marginally profitable. At current prices, each block is worth about $11,000 so that’s as good an estimate as any of the cost of mining. Blocks typically include about 1200 transactions, so the cost per transaction is about $10. (None of this should be controversial; there’s similar numbers at sites like blockchain.info.)

A ledger system that costs ten bucks per transaction is really, really, really expensive. It only seems cheap to the users because 99% of the cost is subsidized by the mining reward and actually, even 10 cents per transaction for a ledger isn’t very cheap.

So what will happen in July when the mining reward drops? Nobody knows, but we can make some educated guesses. The price of bitcoins won’t change—every day there’s about 500,000 bitcoins traded of which only 3600 come from mining. When that 3600 drops to 1800, the effect on the supply and hence the price will be insignificant.

What is likely to happen is that some miners will drop out, since mining that is marginally profitable at 25 btc is hopeless at 12.5 btc. It’s not at all clear how long this adjustment will take. The amount of compute power thrown at bitcoin mining has continued to increase even as the price of bitcoins has gone nowhere. You can tell how much computing the miners are doing from the “difficulty”, a parameter that adjusts the mining puzzle so that on average someone finds a solution every 10 minutes. The difficulty flattened out during 2015, but last November it started shooting up, probably because there’s a new generation of mining hardware.

All else being equal, if the mining reward drops in half, each miner would need to find solutions twice as often to make the same amount, which would mean the difficulty would need to drop by half, too. The current difficulty is about 180 billion, and was 90 billion in January. So the reward drop in effect sets the efficiency of mining hardware back four months, which is not a lot.

Also, a lot of mining is now in China, where bitcoins are often seen as a portable asset like diamonds or gold coins. In that environment, mining below cost makes sense since it turns a non-portable asset into a portable one, and a 50% haircut when you buy and sell diamonds is not unusual.

Hence the mining is likely to go on as fast as ever. The value of the bitcoins mined will drop in half so the obvious subsidy per transaction will drop from $10 to $5, but we have no idea how much hidden subsidy there is from mining at a loss. This reminds us that Bitcoin is a really, really, expensive way to run a ledger, and due to the hidden subsidies from mining at a loss, we have no idea how expensive it is.

We do have a way to estimate how expensive a blockchain is, since there are blockchain-like ledgers not tied to pseudo-currencies, such as the ones developed by Wall St firm R3. They’re pretty closed mouthed about the details of what they’re doing, but the press releases and newspaper reports make it clear that they’re interested in high-value transactions (credit default swaps, repos, and the like) and they’re at least as interested in smart transactions where logic in the ledger makes various things happen as in simply recording what’s already happened. So again, useful though this may be for what it does, there is no reason to think that it is a cheap way to record anything.