Estimating the Economics of Bitcoin Mining Using Natural Gas Destined for the Flare Stack

US oil and gas producers, under both political and economic pressure to reduce natural gas flaring, are increasingly turning to onsite computing solutions to utilize excess natural gas and reduce flaring. Cryptocurrency mining has emerged as perhaps the most popular method to generate value using gas otherwise destined for the flare stack.  Companies such as Crusoe Energy and EZ Blockchain are partnering with oil and gas producers to locate mining infrastructure on location near natural gas gathering infrastructure. This post will explore the economics of Bitcoin mining with flare gas as a feedstock using publicly available data. It will demonstrate that Bitcoin mining powered by flare gas can generate meaningful revenue, particularly with bitcoin hovering around $50,000. Whether or not mining with flare gas represents a meaningful step forward in reducing the oil and gas industry’s environmental impact is a far more complicated debate for another time.

Overview

The post will examine the primary costs related to onsite (i.e., at or near oil and gas wellheads) Bitcoin mining using excess produced natural gas: capital expenditures (CAPEX) for power generation (natural gas turbines), power generation feedstock costs (flare gas isn’t quite free), mining infrastructure costs, and mining economics.  There are several cost vectors I won’t touch on, most notably licensing agreements between oil and gas companies and their infrastructure partners, maintenance, data center overhead costs, peripheral equipment and associated infrastructure. These costs have a meaningful impact on overall economics, so consider this post a back of the envelope estimate or jumping off point for a much more nuanced analysis.

Establishing Power and Feedstock Requirements

Let’s assume we want to deploy 25 MW (megawatts) of power generation capacity, as Crusoe energy had in the Bakken as of June 2021. Small-scale gas turbines appropriate for oil and gas sites cost approximately $800/kWe (kilowatt electric), per Farhan and Purwanto (this study examines the cost of utilizing excess natural gas produced at a field in Indonesia, but $800/kWe appears to be well within the prevailing range for such units). Therefore, the total cost for 25 MW of turbine capacity is $20,000,000 (25,000*$800).

Per Farhan and Purwanto, the heat rate of such units is 10,400 Btu/kWh (British Thermal Units per Kilowatt Hours) and the cost of processing excess gas is $0.35/mcf (thousand cubic feet). At that heat rate, one MWh (megawatt hour) requires 10.4 mcf, and the entire 25 MW system would consume 6,240 mcf each day at a cost of $2,184/d.

The true cost of flared gas is one major uncertainty here. Produced gas must be processed and stripped of liquids to render it a suitable feedstock. Raw gas processing costs can vary greatly depending upon the quality of the feedstock. Excess raw dry gas can require minimal processing and carry a very low cost. However, the NGI article linked to earlier notes that EZ Blockchain abandoned North Dakota due to the high cost of separating NGLs (natural gas liquids) from the gas stream. In their 2016 report on upstream oil drilling and production costs, the Energy Information Administration (EIA) put the cost of gathering and processing wet gas at $0.65 to $1.30 per mcf with additional costs for NGL fractionation and transportation running several dollars per barrel for each. Dry gas gathering and transport to regional market is estimated by EIA to cost $0.35/mcf. Even at a cost of $1.30/mcf, the economics of bitcoin mining in the oilfield remain robust under many scenarios. However, it is reasonable to assume that if a company lacks gathering and transport capacity for dry gas, then they almost certainly lack the infrastructure to strip NGLs from a wet gas stream and send those products to market. Installing the capacity to do so would represent a significant, possibly insurmountable capital cost.

Going forward this post will assume a $0.35/mcf feedstock cost, a figure identified by both the Farhan and Purwanto study as well as EIA (albeit for slightly different metrics), but beware of uncertainty here.

How much mining capacity?

I’ve decided this project will run AvalonMiner A1246 ASIC bitcoin mining rigs. Perhaps I should use Antminer S19 Pro’s…🤷‍♂️. The A1246 is capable of 90 TH/s (terahashes per second) while consuming approximately 3,420 watts at the wall plug, or 3.4 kW. The 25 MW of capacity could theoretically support over 7,000 A1246 units running simultaneously. But allowing for turbine downtime as well as additional cooling and other infrastructure, we’ll assume the system supports 5,000 units (a guess but also a nice round number).

Retail prices for the A1246 range around $7,000-$10,000 each. Let’s assume the low-end of that range at $7,500 and that supply shortages are somewhat offset by bulk pricing power. That brings the total cost of mining rigs to $37,500,000. Add in the $20 million for turbines and our price tag thus far is $57.5 million. We’ll add 20% for other overhead and supporting infrastructure, bringing total CAPEX to $69 million.

You might be wondering how this operation is ever going to make sense, as I was at this point in my calculations. But read on…

Mining Difficulty and ROI

Trying to forecast mining economics into the future is a fool’s errand – but for the sake of this post I’m going to try anyways. Bitcoin prices are volatile. The ‘difficulty’ required to mine a block is adjusted roughly every two weeks by the network. The interplay between price and difficulty affects miners differently depending upon the hashing efficiency of their hardware and their cost of electricity. Miners are constantly responding to price and difficulty, two of the key moving parts of the economic equation. Adjustments are generally by either adding or pulling capacity from the network depending upon their equipment’s efficiency – electricity is the largest operating expense. Therefore the hashrate, or total computing power of the network, is perpetually in flux. On top of this, every few years there’s a halvening in which the block reward is cut in half. Currently, each block mined produces 6.25 BTC. Sometime in 2024 that block reward will be reduced by half.

The following three charts illustrate the relationship (and implicit incentive structure) between price, difficulty, and hashrate.

https://api.blockchain.info/charts/preview/market-price.png?timespan=1year&h=600&w=1200
https://api.blockchain.info/charts/preview/difficulty.png?timespan=1year&h=600&w=1200
https://api.blockchain.info/charts/preview/hash-rate.png?timespan=1year&h=600&w=1200

Quantifying Our Mining Economics

The first step in assessing the current operational economics for our hypothetical oilfield mining operation is establishing how much BTC per day our 5000 machines will mine. Let’s assume a constant network hash rate of 150 EH/s (exahashes), a price of $50,000, and block size of 6.25 BTC per block. We know that difficulty is automatically adjusted so that a block is mined approximately every 10 minutes. We can therefore work backwards to calculate how many hashes are required to mine a block and in turn how many BTC our setup will mine in a given period of time.

If 150 EH/s are mining a block every ten minutes, then 1 block requires about 90,000 exahashes (150 EH/s * 60 seconds * 10 minutes) to be mined. At 90 TH/s, one of our A1246 rigs would require 11,574 days to mine 1 block, or 1852 days to mine 1 BTC (assuming difficulty and block rewards remain constant, which they would not). Our 5,000 rigs in aggregate are capable of 450,000 TH/s, equivalent to 0.3% of the total network hashrate. These machines would generate approximately 2.7 BTC per day (0.3% of the ~900 BTC mined each day) worth $135,000. That comes to $27 of income per day per machine. For reference, asicminervalue.com estimates that an A1246 would generate $30/d in gross income as of September 15, 2021 – so the above estimates are in the right ballpark.

Given the above assumptions and ignoring maintenance, discounting of cash flows, etc., our $69 million in CAPEX would be covered in just 511 days. Factoring in gas feedstock costs of $2,184/d drags the payback period to just 520 days. It’s easy to see why there has been such a rush to Bitcoin mining in the oilfield. Oil and gas producers have an inherent advantage in terms of cost of electricity. Because their feedstock is close to free at $0.35/mcf, they are paying the equivalent of $0.004/kWh in running electricity costs. Even with a ten-fold increase in feedstock costs, bringing the price to $0.04/kWh, oil and gas companies mining bitcoin with flare gas would pay less half of the prevailing average residential rate in the US.

Conclusion: Uncertainties and Natural Gas Markets

The above calculations demonstrate that CAPEX for this operation can be paid back in under two years. I suspect that in reality it is closer to three years – still, nothing to sneeze at. I have not accounted for licensing or profit sharing agreements, which would divvy up the spoils between two or more parties. Nor have I accounted for maintenance and other OPEX related to either the gas turbines or mining rigs themselves.

The time horizon required to generate a positive ROI is a critical variable perpetually in flux. The economics, or relative efficiency, of the machines I picked will decline with time. The A1246 is among the most performant BTC mining rigs available today. But more powerful and efficient competitors are constantly being introduced to the market. All things being equal, the network will become more efficient, difficulty will increase, and today’s machines will take longer and use more energy to mine a given amount of bitcoin. Furthermore, today’s rigs will not chug along indefinitely. In roughly five years, our largest capital expense, the miners, will have to be replaced due to failure. Last but not least, the price of BTC could crash – rendering this investment effectively worthless.

But all miners face these very same uncertainties. They are not unique to our oilfield setup. Unique to flare gas driven mining is the extremely low cost of electricity. This is a benefit against which few can compete. Over a five year period, the $69 million of CAPEX required for this site and ~$2100/d for gas feedstock runs about $40,000/d (before maintenance, interest, etc.) As our 25 MW setup is consuming ~6,000 mcf/d, producer netbacks, roughly equivalent to market prices minus processing and transportation, would have to rise to around $16/mcf to match the net revenues generated from this mining operation ([$135,000/d BTC revenue – 40,000 CAPEX & feedstock cost] / 6,000 mcf/d).

Here’s where things get interesting, however. Let’s assume daily mining revenues fall by half to $67,500/d and our wellheads have market access for their gas. Suddenly, the breakeven netback price drops to $4.58/mcf ([$67,500/d BTC revenue – 40,000 CAPEX & feedstock cost] / 6,000 mcf/d). This implies that it would still be profitable to both continue mining and payoff sunk capital costs within this very simplistic model, but as prices rise above $4.58, a producer without a mining operation would generate greater profits selling gas directly into the market than would our mining operation. With Henry Hub hovering around $5/mcf, some producer netbacks are likely approaching that target. Again, this is a very simplistic, back-of-the-envelope model, but it helps demonstrate the risks associated with oilfield cryptocurrency mining and how quickly economics can flip.

Leave a comment

Your email address will not be published. Required fields are marked *