A bitcoin miner is in essence a kind of inside-out electrical meter: It measures, with remarkable precision, how many electrons (bundled as SHA256 operations) you burn to support the network finding a block, rather than how many you bought, assuming you even bought them at all.
The question is not how many you burned, which is what the mining algo answers with such precision, but how much you paid for them in the first place, which it can offer no insight into.
Some possible shortcuts to paying retail prices for electron mining fuel:
1. You're a large business power customer and pay vastly discounted rates on a contract basis.
2. You're stealing some or all of your power, either by pirate wiring or by paying some official to look the other way, or a little of both.
3. Your mining operation is a pork barrel project with funding secured from your local government's (either clueless or remarkably pragmatic) business development office. Sometimes there may be a legitimate exit plan for this, like repurposing the new datacenter facility you built, after wringing every mining penny from it possible, to host cloud services or as the core of some government-sponsored technology park.
4. You may have been mining quite along time and now have enough early coin held back from the $100 days to slowly sell off into the retail market at 80x profit, or borrow against. It may even be that this reserve of held back coin is partly responsible for the thin liquidity in the market, and consequent bubbled-up price.
5. You publicly, or covertly own or have an interest in one or more of the mining pools and public exchanges, and retain a share some of the Tx and conversion fees which helps subsidize the mining costs.
6. Same as above but for mining rig manufacturers and fabs. Mine on newly manufactured chips for a few days or a week before shipping (for 'testing' purposes ofc) and you get to stay ahead of the difficulty curve as well as get free subsidy from your own customers.