Two people asked almost the same question on Reddit this month. Both were nineteen. Both had somewhere between forty and fifty thousand to spend. Both lived somewhere with very cheap electricity. Both asked: what do I buy, what will it earn, and how loud will it be? Here is the honest, ordered answer.
- ✓Cheap electricity is a real advantage — but on its own it is not an operation.
- ✓Work out your power budget in kilowatts before you look at a single machine.
- ✓Efficiency (J/TH) decides how much of your gross you keep. It matters more than hash rate.
- ✓Do not spend your whole budget on hardware. Electrical work, spares and a cushion are part of the build.
- ✓Start with a handful of machines, learn what breaks, then scale. Everyone who skips this learns it anyway, more expensively.
Start with power, not machines
The instinct is to pick a machine first. That is backwards. The correct first question is: how many kilowatts can I actually supply, continuously, for years?
Three numbers define your site, and you need all three before shopping:
- 1Your rate. What you pay per kilowatt-hour, all in, including any standing charge and tax. Not the advertised rate — the number on your bill divided by the units used.
- 2Your capacity. How many kilowatts your supply can carry continuously. Your electrician can tell you. This is a hard ceiling.
- 3Your reliability. How often the power actually fails where you are. In a lot of places with very cheap electricity, this is the number that ends up mattering most.
Cheap power and reliable power are different things, and they often do not come together. A site at three cents per kWh that is down eight hours a week produces less than a site at five cents that is almost never down. Work out your real availability before you build around a low rate.
Efficiency is the whole game
Every miner has a number called J/TH — joules per terahash. It is simply how much electricity the machine spends to do one unit of work. Lower is better.
Here is why it matters more than anything else on the spec sheet. From J/TH alone you can calculate the exact electricity price at which a machine stops making sense. That is called the break-even rate, and it is the most useful number in mining.
| Miner | Hash rate | Power | J/TH | Break-even rate | Price |
|---|---|---|---|---|---|
 Antminer S19k Pro - 110 TH/s |
110 TH/s | 2,530 W | 23.00 | $0.0728 | $297 |
 Antminer S21 - 188 TH/s |
188 TH/s | 3,290 W | 17.50 | $0.0957 | $1,111 |
 Antminer S21 Pro - 245 TH/s |
245 TH/s | 3,675 W | 15.00 | $0.1116 | $2,059 |
 Whatsminer M79S - 1.35 PH/s |
1,350 TH/s | 20,000 W | 14.81 | $0.1130 | $14,559 |
 Antminer S23 - 318 TH/s |
318 TH/s | 3,498 W | 11.00 | $0.1522 | $6,899 |
 Antminer S23 Hyd - 580 TH/s |
580 TH/s | 5,510 W | 9.50 | $0.1762 | $12,299 |
What a real budget looks like
Here is the mistake almost every first-time buyer makes: spending the entire budget on machines. A working site is machines plus everything that lets them run.
There is also a middle path that a lot of people miss. You do not have to choose between "all at home" and "all hosted". Running a small number of machines at home to learn, while the rest of the fleet sits in a hosted site on a fixed rate, gives you the experience without betting the whole budget on your own electrical work.
Choosing your first machines
Use your break-even rate from Chart 1 and apply these three rules.
- ✓Rule 1: leave headroom. Do not buy a machine whose break-even rate is barely above your electricity price. Difficulty rises over time, which pushes every break-even rate down. Buy margin, not a knife edge.
- ✓Rule 2: buy fewer, better machines. Two efficient machines usually beat five old ones at the same price. They use less power, take less space, break less often, and stay useful longer.
- ✓Rule 3: buy the same model. One model means one set of spares, one set of settings, one thing to learn. Mixed fleets are how small operations get complicated.
How loud, how hot, how much space
The three practical realities, in numbers, before you commit.
| Reality | What to expect | What to do about it |
|---|---|---|
| Noise | A stock air-cooled machine is roughly 75–85 dB — between a vacuum cleaner and a lawnmower, running constantly. | Separate building, or hydro-cooled models, or hosting. |
| Heat | Almost all the electricity becomes heat. A 3,290 W machine is a 3,290 W heater. | Plan airflow in and out before installation, not after. |
| Dust | Fans pull in everything in the air. Dust is the most common cause of failure. | Filters, and a cleaning schedule you actually keep. |
| Space | Machines need airflow gaps and access for maintenance, not just their own footprint. | Allow roughly double the physical size of the machines. |
| Internet | Modest bandwidth, but it must be constant. Dropouts cost production. | A wired connection, and a backup route if the site is remote. |
The order to do things in
If step 2 comes back with a low number, or step 5 shows your uptime is poor, that is not the end of the plan. It just means the machines belong somewhere with industrial power. The OneMiners hosting network runs 20 sites at an average of $0.0480 per kWh on seven-year fixed rates, with 0% management fees and a 7-year hardware warranty.
Sensible first-fleet machines
Compare real hash rate, wattage and price across the catalogue, or check what a fixed industrial rate would do to your break-even maths.
Compare hosting sitesBrowse Bitcoin minersQuestions people keep asking
Where these numbers come from
- Miner hash rate, wattage and price: live OneMiners catalogue, 27 August 2026.
- Break-even calculations use a Bitcoin price of $79,264 and network hash rate 887.8 EH/s (mempool.space, 27 August 2026), subsidy only, before pool fees.
- Budget split is a planning guideline based on typical small-fleet builds, not a survey.
- Hosting network rates and terms: oneminers.com/pages/hosting-centers.
This article is for information and education only. It is not financial advice. Mining results change with the Bitcoin price, network difficulty, your electricity rate and your hardware. Numbers shown are examples from public data on the date of writing, not a promise of any result. Always do your own research.
