Electricity is 75 to 85 per cent of the ongoing cost of running an ASIC, which means the energy question is not an ethical footnote to Bitcoin mining — it is the business. On 8 September 2026, with hashprice at $38.72 per PH/s per day, an Antminer S23 Hydro grosses $22.46 a day and burns $6.35 of electricity at the OneMiners network average of $0.0480/kWh. Move that same machine to a $0.12 domestic tariff and the power bill becomes $15.87, which is more than the machine earns. Nothing about the hardware changed. The entire outcome was decided by where the electricity came from.
- The stakes: the energy source is the business model
- Stranded, curtailed and surplus: the three cheap categories
- Why miners are the grid's ideal flexible load
- What flexible load looks like in practice: the Texas case
- The five energy sources a mining site can be built on, ranked
- Where the OneMiners network gets its power
- The 2026 complication: AI is bidding for the same megawatts
- Red flags in any hosting provider's energy story
- The energy verdict
- Frequently asked questions
- Final thoughts
First, the stakes: the energy source is the business model
Mining companies are often described as technology businesses. They are energy businesses that happen to own computers. Consider the same machine at four different rates.
| Electricity rate | Cost per kWh | Daily power cost | Daily net margin | Annual margin per machine |
|---|---|---|---|---|
| Nigeria (lowest) | $0.0364 | $4.81 | $17.64 | $6,440 |
| USA / Georgia | $0.0455 | $6.02 | $16.44 | $6,001 |
| Network average | $0.0480 | $6.35 | $16.11 | $5,880 |
| Home power | $0.1200 | $15.87 | $6.59 | $2,405 |
A spread of $4,035 a year on a machine costing $12,299 is not an operational detail, it is the whole investment case. Which raises the only question that matters in this industry: how does anyone get electricity at $0.0364 a kilowatt hour, when a household pays three or four times that?
What a mining machine actually consumes
Before going further it is worth grounding the discussion in physical quantities rather than abstractions. These are the machines in the current catalogue, and this is the electricity each one genuinely needs.
| Machine | Power draw | Daily consumption | Daily cost at $0.0364 | Daily cost at $0.12 |
|---|---|---|---|---|
 Antminer S23 Hyd 3U 1.16 PH/s |
11,020 W | 264 kWh | $9.63 | $31.74 |
 Antminer S23e Hyd 2U 865 TH/s |
8,650 W | 208 kWh | $7.56 | $24.91 |
 Antminer S23 Hydro 580 TH/s |
5,510 W | 132 kWh | $4.81 | $15.87 |
 Antminer S23 (Air) 318 TH/s |
3,498 W | 84 kWh | $3.06 | $10.07 |
 Antminer S21 188 TH/s |
3,290 W | 79 kWh | $2.87 | $9.48 |
Stranded, curtailed and surplus: the three cheap categories
What links all three is that the power has no better buyer. That is the entire source of the discount, and it is also why the arrangement is genuinely mutual rather than extractive: the generator converts output it could not otherwise monetise into revenue, and the miner gets a rate no ordinary industrial consumer could negotiate.
Why miners are the grid's ideal flexible load
Grid operators value one thing above almost everything else in a large consumer: the ability to stop consuming on request. Most industrial loads cannot. A smelter cannot let its metal solidify, a hospital cannot switch off, a chemical plant may need days to restart safely.
A Bitcoin mining facility can shed its entire load within minutes and restore it just as quickly, with no damage, no spoiled product and no safety consequence. Individual machines power down and back up almost instantly. Nothing is ruined by pausing; the only cost is the revenue that was not earned during the pause, and that cost is transparent, immediate and easy to compensate.
| Load type | Time to shed load | Cost of interruption | Value to grid operator |
|---|---|---|---|
| Bitcoin mining | Seconds to minutes | Forgone revenue only | Very high |
| Aluminium smelting | Hours, with damage risk | Ruined production run | Low |
| Data centre (conventional) | Not interruptible | Service outage | Very low |
| Residential heating | Limited, brief | Comfort and safety | Low |
| Industrial manufacturing | Hours to days | Lost output, restart costs | Low |
What flexible load looks like in practice: the Texas case
The most documented example is ERCOT, the Texas grid operator, which formally classifies facilities with expected peak demand of 75 megawatts or greater as large flexible loads, on the working assumption that they will reduce consumption when wholesale prices rise above roughly $100 per megawatt hour. The US Energy Information Administration has tracked electricity consumption from American cryptocurrency mining operations specifically because the category became large enough to matter to grid planning.
The scale involved is not small. ERCOT served a preliminary peak of about 91,308 MW in late July 2026, having passed 87,403 MW the previous day and broken a record of 85,508 MW that had stood since August 2023. In a system under that kind of pressure, a category of consumer that voluntarily disappears during the worst hours is not a nuisance to be tolerated. It is infrastructure. Texas legislation passed in 2025 moved further in this direction, requiring certain large power users connecting from 2026 onward to reduce consumption during severe grid emergencies, alongside a programme to pay large sites for doing so.
The lesson generalises well beyond Texas. Any grid with significant renewable penetration develops both surplus periods and scarcity periods, and both are made easier by a large consumer that can be switched off at will. That is the role mining plays, and it is why the relationship between miners and grid operators has moved from suspicion toward procurement.
The five energy sources a mining site can be built on, ranked
1. Long-term fixed contracts on surplus hydro and renewables
Verdict: the strongest position available, and the one OneMiners is built on. A multi-year fixed price on genuinely surplus generation gives you the two things that matter simultaneously: a very low rate and certainty about it. The OneMiners Ethiopia site runs on hydro and renewable generation at $0.0399/kWh; Norway and Finland use Arctic and cold-climate conditions where the surrounding air does much of the cooling work for free. Contracts are fixed for seven years, which spans the 2028 halving and most of the physical life of the hardware.
2. Stranded generation in energy-rich regions
Verdict: the lowest rates on earth, if the site is built properly. The OneMiners Nigeria facility sits at $0.0364/kWh, the cheapest active rate in the network, with a further 250 MW of expansion underway. Rates at that level exist only where generation genuinely has no alternative buyer. The trade-off is that everything else — cooling engineering, connectivity, technical staffing, physical security — has to be built rather than assumed.
3. Curtailment-linked and demand-response arrangements
Verdict: excellent economics, but the site must be designed for interruption. Buying power that would otherwise be curtailed produces exceptional rates and can generate availability and activation payments on top. The cost is uptime: a facility optimised around curtailment will be asked to stop, sometimes at short notice. For a hosting customer this matters enormously, because it is your machine that stops. It is a strong model when the uptime commitment is explicit and compensated, and a poor one when it is buried in the small print.
4. Grid industrial tariffs
Verdict: workable, predictable, and rarely world-class. A standard industrial connection in a well-regulated market gives reliability and simplicity at a rate that is usually somewhere in the five to eight cent range. The OneMiners USA regional sites sit at $0.0455/kWh with no installation or hidden fees, which is a strong figure for a developed-market grid connection and is why they carry a large share of the network's capacity.
5. Domestic and small commercial supply
Verdict: not an industrial energy strategy. At $0.12/kWh an S23 Hydro loses $6.59 a day , and the older generations are far worse. Domestic supply is also unfiltered, thermally uncontrolled and electrically noisy, which shortens hardware life on top of the rate problem. This row exists on the list to be excluded.
Where the OneMiners network gets its power
OneMiners operates roughly 2,163 MW of contracted capacity across 20 sites, deliberately spread across energy types and jurisdictions rather than concentrated in one market. The selection below shows what that looks like in practice.
| Site | Capacity | Rate per kWh | Energy characteristic |
|---|---|---|---|
| Nigeria | 33 MW | $0.0364 | Cheapest active rate in the network; +250 MW expansion underway |
| Ethiopia | 40 MW | $0.0399 | Hydro and renewable generation |
| UAE (Dubai + Abu Dhabi) | 34 MW | $0.0420 | Premium build with advanced cooling |
| Finland | 22 MW | $0.0448 | Cold-climate free cooling |
| Norway (Arctic) | 36 MW | $0.0448 | Arctic natural cooling |
| USA regional | 336 MW | $0.0455 | New York, Georgia, South Carolina, Houston, Kansas, Texas; no install or hidden fees |
| China (dedicated) | 288 MW | $0.0462 | Industrial-scale dedicated capacity |
| Canada | 25 MW | $0.0476 | Stable North American jurisdiction |
Three design decisions are visible in that table. The sites are placed where energy is genuinely surplus rather than where it is convenient. Cooling is treated as an energy strategy rather than an afterthought — Norway and Finland use ambient conditions, the UAE sites use engineered cooling, and much of the fleet is liquid-cooled. And the rates are fixed for seven years, which converts a volatile input into a known one for longer than most of the hardware will physically last.
The 2026 complication: AI is bidding for the same megawatts
The significant change in this decade is that Bitcoin miners are no longer the only buyer looking for large blocks of cheap power with a fast path to energisation. AI and high-performance computing operators now compete for the same sites, often with deeper pockets, and a number of listed mining companies have converted capacity to serve them.
For anyone hosting hardware this has two consequences worth understanding. Competition for megawatts puts upward pressure on rates in the most contested markets, which raises the value of capacity that was contracted years ago at a fixed price. And it removes flexibility from the grid, because conventional AI workloads cannot be interrupted the way mining can — a point grid operators have begun making publicly. A hosting provider holding long-dated fixed contracts across several jurisdictions is in a materially stronger position in that environment than one buying power at prevailing rates each year.
Red flags in any hosting provider's energy story
- A rate quoted with no contract length. A cheap rate for an unspecified period is a marketing number. Ask how many years it is fixed for.
- "Market-linked" or index-tracking pricing. This transfers all energy risk to you, and it does so most severely in precisely the conditions where margins are already compressed.
- Curtailment exposure with no uptime commitment. If the site earns from demand response, find out who bears the downtime. The OneMiners commitment is 95%+ with compensation.
- Rates that exclude transmission, capacity or grid fees. The all-in figure is the only one worth comparing. OneMiners quotes all-inclusive rates with no installation or hidden fees.
- A single site in a single jurisdiction. Concentration is a genuine risk: one regulatory change, one drought or one grid event affects everything at once.
- Vague renewable claims with no site detail. "Powered by renewables" with no named generation source, region or contract is not information.
The energy verdict
Frequently asked questions
Final thoughts
The public argument about Bitcoin mining and energy is usually conducted as though miners compete with households for the same electricity. They do not, and the economics would not work if they tried. Mining is viable precisely where power is abundant, badly located, poorly timed or about to be discarded, and it is viable because it is the one large consumer that can vanish from the grid on request and reappear when the pressure passes.
For anyone actually deploying hardware, that turns into a very practical checklist. What is the rate, how many years is it fixed for, what is the energy source behind it, what happens when the grid asks the site to stop, and who compensates you when it does. The OneMiners answers are $0.0480/kWh on average and $0.0364 at the best site, fixed for seven years, across 20 sites and roughly 2,163 MW, with a 95%+ floor and compensation attached. Revenue is weather. Fees are a negotiation. Electricity is physics — and it is the only line on the sheet you actually get to choose.
Data sources
- ERCOT large flexible load classification at 75 MW and the assumption of curtailment above roughly $100 per megawatt hour — US Energy Information Administration analysis of ERCOT interconnection data.
- US cryptocurrency mining electricity consumption tracking — US Energy Information Administration.
- ERCOT demand records of approximately 91,308 MW in July 2026 against a previous record of 85,508 MW from August 2023 — ERCOT preliminary demand data as reported in 2026 energy coverage.
- Texas Senate Bill 6 (2025) requirements for large power users connecting from 2026 onward — 2026 reporting on Texas grid legislation.
- Bitcoin price $78,500, difficulty 127.45T and the implied 912.3 EH/s — CoinGecko, Blockchair and mempool.space, 8 September 2026.
- Hashprice methodology — Hashrate Index and Luxor Technologies.
- OneMiners site capacities and seven-year fixed prepaid energy rates — OneMiners published facility data.
