Building MSME Resilience in Indonesia: Credit Markets,

ed
Fintech Solutions, and Risk Mitigation Strategies (2018–2023)aꝉ

iew
aSimon Poltak Hamonangan Hutabarat, Ph.D., Tax Policy Analyst, Indonesia's Directorate General of Taxes,

Ministry of Finance, Jakarta, Indonesia

ꝉPlease send all correspondences to [email protected] or [email protected]

ev
Acknowledgement
The author would like to express sincere gratitude to my wonderful supervisor, Terrence

r
Iverson, and all Members of the Dissertation Committee, i.e., Jesse Burkhardt, Anders
Fremstad, and Martin Shields, for their constructive feedback and insightful comments

er
during the Dissertation Writing Process and on an earlier version of this paper. We also
extend our appreciation to Lembaga Pengelola Dana Pendidikan (LPDP) for its generous
support in funding the conference participation and facilitating the logistical
pe
arrangements related to the development and submission of this manuscript.

Funding Statement
The authors disclosed receipt of the following financial support for the research,
ot

authorship, and/or publication of this article: This work was supported by Lembaga
Pengelola Dana Pendidikan (LPDP) Indonesia.
tn

Ethical approval and informed consent statements

There are no human participants in this article, and informed consent is not required.
rin

Data availability statement

The data is available upon request.
ep

Notes on Authors
Simon Poltak Hamonangan Hutabarat is a Senior Tax Analyst at Indonesia's Directorate
Pr

General of Taxes, Ministry of Finance. He holds a PhD in Economics from Colorado State
University in the United States. Simon's research focuses on public, monetary, regional,

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 energy, and natural resource economics. Simon has published multiple articles in

ed
journals such as the International Journal on Engineering, Science, and Technology, the
Scholar's International Journal of Business Policy & Governance, and Next Research.

iew
r ev
er
pe
ot
tn
rin
ep
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 Bitcoin Mining and Renewable Energy: Navigating

ed
Sustainability, Profitability, and Electricity Market Dynamics a,ꝉ

iew
a Simon Poltak Hamonangan Hutabarat, Ph.D., Tax Policy Analyst, Indonesia's Directorate General of
Taxes Ministry of Finance, Jakarta, Indonesia

ꝉPlease send all correspondences to [email protected] or [email protected]

ev
ABSTRACT

r
This paper examines the potential synergy between renewable energy and
Bitcoin mining, proposing a framework for aligning cryptocurrency
operations with periods of surplus electricity generation. Utilizing CAISO
er
data from 2018 to 2020, the study assesses whether Bitcoin mining can
effectively consume excess renewable energy during low-demand hours,
mitigating curtailment and optimizing profitability. By analyzing Locational
pe
Marginal Prices (LMPs) across different energy sources, the research
identifies wind and solar as viable energy inputs for mining operations,
particularly during hours with lower or negative prices. Additionally, this
study examines how Bitcoin price fluctuations affect optimal mining periods,
taking into account the economic feasibility of mining under varying
ot

electricity costs. The findings suggest that while Bitcoin mining can
potentially absorb surplus renewable power, its environmental and economic
viability depends on significantly lower Bitcoin prices than historical
tn

averages. This study contributes to the broader discourse on integrating

cryptocurrency mining into sustainable energy systems, offering implications
for energy policymakers, cryptocurrency stakeholders, and environmental
advocates. Bitcoin miners could play a pivotal role in stabilizing power grids
rin

and advancing renewable energy adoption by strategically timing their

operations to coincide with low-cost, renewable electricity sources.

Keywords:
ep

Bitcoin Mining; Renewable Energy; Locational Marginal Prices (LMPs),

CAISO; Profitability Model; Energy Market Dynamics.
JEL Classification: G17, Q21, Q28, Q42, Q47, Q48
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 ed
1. Introduction

iew
Bitcoin mining has become a major global industry, involving hundreds of

thousands to over a million participants, mostly through mining pools.1 Its decentralized,

energy-intensive process consumes more electricity than nations like Argentina and

ev
Sweden2, raising environmental concerns (Hakimi et al., 2024; Zribi et al., 2023).

However, mining can use surplus renewable energy during off-peak hours (Winton,

r
2021), reducing fossil fuel dependence and potentially promoting renewable energy

development, as argued by Jack Dorsey and Elon Musk.

er
pe
In 2021, Dorsey tweeted that "Bitcoin incentivizes renewable energy," citing a

white paper co-authored by Square and ARK Invest (Winton, 2021). While
ot

cryptocurrencies like Bitcoin and Ether bypass intermediaries (CoinMarketCap, 2018),

Bitcoin consumes 45.8 TWh annually, emitting up to 22.9 MtCO₂ (Stoll et al., 2019). Jiang
tn

et al. (2021) projected 296.59 TWh in China by 2024. Yet, mining could absorb excess

renewables as a flexible load, functioning like storage (Shynkevich, 2021; Velický, 2023).
rin

1Data from Bitbo.io (accessed on January 25, 2025) estimates that the number of Bitcoin miners ranges from hundreds
of thousands to over a million. However, precise figures remain uncertain due to the decentralized and pseudonymous
ep

nature of the Bitcoin network. See Bitbo.io, "How Many Bitcoin Miners Are There?" Available at: https://bitbo.io/how-
many-bitcoin (accessed January 25, 2025).
2 According to an analysis by IntelliNews (accessed on January 25, 2025), Bitcoin mining consumes more electricity
than several nations, including Argentina, Sweden, and the Netherlands. The report highlights that the global energy
demand for Bitcoin mining surpasses that of Ukraine, the UAE, and Argentina, reinforcing concerns about its
Pr

environmental footprint. See IntelliNews, "Who Consumes the Most Power to Mine Bitcoins?" Available at:
https://www.intellinews.com/who-consumes-the-most-power-to-mine-bitcoins-328682 (accessed January 25, 2025).

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 This paper empirically examines whether Bitcoin mining can serve as a flexible

ed
demand-side resource that complements the intermittency of renewable energy. Using

detailed hourly data from the California Independent System Operator (CAISO) between

iew
2018 and 2020, this analysis examines Locational Marginal Prices (LMPs) by hour and

generation type to identify the conditions under which mining operations can reduce

ev
costs and emissions by selectively operating during off-peak hours.

The main research questions are: (1) What are the lowest LMPs per hour across the

r
CAISO grid, and how do they differ by energy source? (2) Are LMPs systematically lower in areas

er
with high renewable energy penetration, such as solar or wind? (3) Under what market and

technical conditions can Bitcoin mining become profitable and environmentally sustainable?
pe
The paper employs quantile regression and econometric models to answer these

questions and analyze hourly LMP patterns across eight primary generation sources:
ot

wind, Solar, Water, Geothermal, Biogas/Biomass, Natural Gas, Coal, and Uranium. These
tn

analyses reveal that Wind and Solar often generate significantly lower LMPs, particularly

in the early morning hours and the lowest price quantiles (e.g., the 5th percentile).
rin

Building on this, a profit-maximizing model determines optimal mining hours

based on Bitcoin prices, ASIC efficiency, electricity costs, and fixed investments. The
ep

model shows that miners selectively mine off-peak only if prices fall below $3,000—far

below the historical average of $8,700. As Bitcoin prices rise, mining remains profitable
Pr

for longer daily periods, extending operational viability.

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 These findings suggest that intermittent Bitcoin mining can serve as a form of

ed
virtual energy storage, absorbing excess renewable energy when electricity is

inexpensive. Feasibility depends on market prices, hardware efficiency, and fixed costs.

iew
The analysis supports Jack Dorsey's claim that mining can incentivize renewables by

monetizing curtailed energy. Section 2 discusses the theoretical foundations, Section 3

ev
outlines the data and methods, Section 4 presents the results, and Section 5 concludes

with policy implications for integrating sustainable mining.

r
This article is derived in part from the author's doctoral dissertation titled "Essays

er
on Bitcoin Mining and Renewable Energy: Exploring Sustainability and Profitability"

(Hutabarat, 2023), submitted to Colorado State University.

pe

2. Theoretical Framework and Main Contribution

ot

This paper contributes to research on cryptocurrency mining, renewable energy

tn

integration, and electricity market dynamics. First, it incorporates GPU and ASIC

hardware advancements into a profitability model, extending the work of Delgado-

rin

Mohatar et al. (2019) on break-even pricing. Second, it assesses geopolitical impacts, such

as China's 2021 mining ban, which redirected operations to regions with abundant
ep

renewable resources, including Washington (Hughes, 2017; Greenberg & Bugden, 2019)

and California (Niaz et al., 2022). Third, it builds on Mills et al. (2021) and Antweiler
Pr

(2021) by examining arbitrage opportunities during negative LMPs, likened to virtual

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 storage (Vlachos et al., 2018). Fourth, it examines market dynamics using examples from

ed
Germany's day-ahead market (Clean Energy Wire, 2018), as visualized in Figure 1, and

studies by Kaffine et al. (2021) and Burkhardt et al. (2019). Fifth, echoing Badea et al.

iew
(2021), it reframes Bitcoin mining as a demand response mechanism that could stabilize

grids and reduce emissions.

ev
Figure 1: Example of the supply and demand curve in the German energy market

r
er
pe
(a)
ot
tn
rin

(b)
ep

Source: Clean Energy Wire (2018)

Also, this study contributes to the literature by integrating insights on negative

Pr

price dynamics, hardware efficiency, regulatory shifts, and renewable energy variability.

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 It positions Bitcoin mining as a potentially strategic mechanism for grid optimization and

ed
economic feasibility in a renewable energy landscape that is increasingly dominated by

renewable energy sources.

iew
ev
3. Materials and Methods

r
3.1. Data Description

er
This section analyzes a CAISO and EIA dataset, focusing on California's Locational
pe
Marginal Prices (LMPs) from 2018 to 2020. The dataset, refined to 27,078,020 observations,

includes variables such as hour, LMP, and energy source, as outlined in Table 1.

Table 1: Construction of the datasets

ot

Procedures Remaining Amount of Data Notes

Retrieving from the CAISO
Original datasets 27,079,105
website
tn

Reduction due to the 2017 Cleanup for only 2018-2020

27,078,020
exclusion datasets
Nighttime Solar observation Drop the Solar observation
23,538,005
removal from 6 PM to 6 AM
rin

Source: Author's Calculation

It is important to note that our analysis solely focuses on CAISO data and does not
ep

incorporate transmissions from other operators. The code adheres to CAISO's established

methodology to ensure consistency and compliance with industry standards. In the

Pr

subsequent section, I present summary statistics that reveal insights into the average 𝑛-

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 cheapest hours of LMPs for eight specific power sources of interest.

ed
CAISO data provides hourly LMP data for various nodes and generators within a day.

The key variables in the dataset are node or Generator ID (the generator ID corresponds

iew
to the power plant), 𝑦𝑒𝑎𝑟 (the year of observation, i.e., 2018 to 2020), 𝑚𝑜𝑛𝑡ℎ (the month

of observation, coded as 1 for January to 12 for December), 𝑑𝑎𝑦 (the day of observation,

ev
coded as 1 to 31, 1 to 30, 1 to 28, or 29), ℎ𝑜𝑢𝑟 (hourly data ranging from 1 to 24), 𝐿𝑀𝑃

(the Locational Marginal Price in $/MWh), and category or 𝑅𝑒𝑠𝑜𝑢𝑟𝑐𝑒 𝐼𝐷 (8 power sources

r
observed: Wind, Sun, Biogas/Biomass, Uranium, Natural Gas, Coal, Geothermal, and

er
Water). The LMP is a measure that represents the electricity price at different locations

and times within a given month and year. The summary statistics of the LMPs are
pe
presented in Table 2.

Approximately 20% of nodes are connected to multiple power plants, receiving

ot

electricity from diverse sources and enhancing supply reliability. For instance, a node
tn

connected to both gas-fired and solar power plants will exhibit distinct pricing for each

source, reflecting the varied generation costs of each. Overnight prices for nodes near
rin

solar plants are not anomalies; they represent compensation for gas-fired generators

when solar production is inactive. During the day, solar prices reflect their contribution
ep

to the grid. This mechanism enables accurate price differentiation, ensuring fair

compensation for each plant's output. Load-serving entities must navigate these price
Pr

dynamics when procuring power, aligning costs with specific generation sources.

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 Table 2: Summary Statistics of Each Category's LMP

ed
M (Sd) Median [Q1-Q3] Min-Max 95% CI
Category N
($/MWh) ($/MWh) ($/MWh) ($/MWh)
Biogas/Biomass 2,770,050 35.5 (44.6) 29.5 [21.6 - 40.7] -1130.0 - 2945.2 [35.4, 35.5]
Coal 224,432 34.7 (45.3) 28.8 [20.8 - 39.9] -647.9 - 2570.3 [34.5, 34.9]
Geothermal 699,033 32.2 (42.7) 27.4 [19.9 - 37.6] -738.8 - 1662.0 [32.1, 32.3]

iew
Natural Gas 7,075,675 35.5 (45.7) 29.3 [21.3 - 40.7] -4114.6 - 2770.4 [35.5, 35.6]
Sun 3,511,540 27.2 (31.7) 24.6 [17.1 - 35.4] -1213.3 - 2922.4 [27.1, 27.2]
Uranium 100,068 35.9 (44.0) 29.5 [21.2 - 41.4] -483.3 - 1372.6 [35.6, 36.2]
Water 6,098,376 33.7 (45.4) 28.6 [20.9 - 39.1] -1198.8 - 4306.2 [33.7, 33.8]
Wind 3,058,831 32.9 (43.1) 28.3 [20.3 - 39.1] -908.6 - 1424.5 [32.8, 32.9]
Note. M = Mean, SD = Standard Deviation, Q = Quartile, CI = Confidence Interval. Solar (SUN) data is excluded
when the Sun does not shine (18:00-06:00).

ev
This section highlights how multiple plants at a node affect pricing, ensuring

r
fairness by attributing LMPs to actual production. Nighttime solar prices—reflecting

er
non-solar generation—were dropped to reduce resource misidentification. Table 3

illustrates implications for load-serving entities sourcing power from multi-source nodes
pe
in CAISO.

Table 3: Power Plant and Generated Electricity Source for Nodes with Multiple Power
ot

Power Plant code name Generated Electricity Source Notes

COGNTNL_7_B1 Coal, Biogas/Biomass Depending on the source, entities
CPCSTCN_7_B1 Coal, Biogas/Biomass will pay the listed price associated
LASSEN_6_N003 Natural Gas, Biogas/Biomass with the respective generator,
tn

MISSION_2_N035 Natural Gas, Water whether it be gas-fired or coal-fired,

MTPOSO_7_N001 Coal, Biogas/Biomass reflecting the diverse pricing
SCLARA_6_N008 Natural Gas, Biogas/Biomass, Water dynamics within the market.
Source: Author's Calculation
rin

In CAISO datasets, each generation resource has a unique Resource ID to track

production at the nodal level. Locational Marginal Prices (LMPs) reflect the cost of
ep

supplying one more MWh at specific locations and times, factoring in generation,

congestion, and losses (Cretì and Fontini, 2019; Hogan, 2010). LMPs are resource-
Pr

specific, with no price spillovers between sources, such as wind or solar (California ISO,

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 2023; Joskow, 2019).

ed
3.1.1. Summary Statistics of Nodes

iew
This section summarizes the statistics for each node or category within the CAISO

datasets for the years 2018-2020, with a focus on California. One key aspect of interest is

ev
the distribution of energy sources used for power generation in the state.

Table 4 highlights the number of observed nodes within the CAISO datasets

r
during this timeframe.

Category
er
Table 4: Summary Statistics of the Nodes
Total Observed Nodes/Hours Total Unique Power Plant % of Total Nodes
pe
Natural Gas 7,074,829 282 26.71%
Sun 6,465,547 265 24.41%
Water 6,097,645 208 23.02%
Biogas/Biomass 2,769,718 105 10.46%
Wind 3,058,469 83 11.55%
ot

Geothermal 698,949 26 2.64%

Coal 224,405 6 0.85%
Uranium 100,056 4 0.38%
Source: Author's Calculation
tn

Natural Gas leads California's energy mix (26.71%), followed by Solar (24.41%)

and Water (23.02%), indicating a shift to renewables. Wind (11.55%), Biogas/Biomass

rin

(10.46%), and Geothermal (2.64%) support clean energy goals. Coal (0.85%) and Uranium

(0.38%) are marginal. The 5th percentile of 𝑛-cheapest LMP hours supports Dorsey's
ep

sustainable mining hypothesis. Regression models and Figure 2 (Bitbo.io) analyze Bitcoin
Pr

prices from 2018 to 2020 to identify cost-effective mining windows.

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 Figure 2: Bitcoin Prices 2018 - 2020

ed
iew
r ev
er
Source: downloaded from https://charts.bitbo.io/price
pe
A crucial component in Bitcoin mining is application-specific integrated circuits
ot

(ASICs), specialized hardware designed for high-efficiency mining. Data from 2017–2020

was sourced from asicminervalue.com. This study compares two representative models:
tn

the Bitmain Antminer S9 (13.5 TH) and the MicroBT Whatsminer M31S+. I examine how

Bitcoin prices and optimal mining hours affect the profitability of each device, reflecting
rin

the pivotal role of hardware in shaping energy-efficient and cost-effective mining

strategies.
ep

3.1.2. Summary of the 𝒏-th cheapest hours

Pr

10

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 Calculating average 𝑛-cheapest LMPs identifies the lowest-cost hours for each

ed
energy source. The 5th percentile analysis reveals that renewables, such as wind and

solar, often experience negative prices, highlighting cost-effective periods, as shown in

iew
Table 5. In this observation, non-renewable sources, such as natural gas, are considered

'clean' and ranked lower than coal.3 As the mining duration increases to 4 or 8 hours,

ev
electricity costs rise. Wind and solar show notable price differences between the 5th and

25th percentiles, while higher percentiles indicate more stable and slightly higher LMPs.

r
Table 5: The 𝑘-th percentile of average cheapest two, four, and eight hours by node/category (2018- 2020)

wind
Category Mean
7.71
P5
-26.44
P25er
Average Cheapest 1-Hour Prices ($/MWh)

1.00
P50
15.42
P75
21.62
P95
31.99
pe
sun 8.99 -16.85 1.47 15.10 21.61 32.66
natural gas 11.88 -14.02 4.91 16.46 22.64 33.68
uranium 14.41 -8.30 7.95 16.92 22.65 34.01
bogas/biomass 12.63 -13.48 7.35 17.31 23.32 34.36
coal 12.57 -10.41 5.85 16.23 22.40 33.32
geothermal 11.90 -15.78 6.25 16.46 22.53 32.23
ot

water 9.58 -18.13 5.92 16.97 23.10 33.73

Average Cheapest 2-Hour Prices ($/MWh)

tn

Category Mean P5 P25 P50 P75 P95

wind 9.62 -22.03 3.45 16.13 22.12 32.43
sun 10.87 -14.49 4.31 15.98 22.28 33.31
natural gas 13.74 -10.52 7.90 17.31 23.39 34.47
rin

uranium 15.65 -6.20 10.00 17.65 23.41 34.66

bogas/biomass 14.39 -10.41 9.65 18.04 24.01 35.02
coal 14.28 -8.45 8.37 17.01 23.15 34.20
geothermal 13.45 -13.95 8.32 17.28 23.08 33.03
ep

water 11.92 -15.29 8.36 17.69 23.71 34.50

Pr

3Natural Gas is burned for energy, which results in fewer emissions of nearly all types of air pollutants and carbon
dioxide (CO2) than burning coal or petroleum products to produce an equal amount of energy (eia.gov).

11

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 Table 5: The 𝑘-th percentile of average cheapest two, four, and eight hours by node/category (2018- 2020)

ed
Average Cheapest 4-Hour Prices ($/MWh)
Category Mean P5 P25 P50 P75 P95
wind 12.36 -16.43 7.38 17.33 23.24 33.97
sun 13.58 -11.34 8.17 17.35 23.53 35.04

iew
natural gas 16.30 -6.64 11.25 18.59 24.71 36.36
uranium 17.64 -3.02 12.28 18.81 25.04 36.47
bogas/biomass 16.87 -6.32 12.26 19.22 25.18 36.64
coal 16.56 -4.81 11.32 18.26 24.45 35.78
geothermal 15.61 -9.80 10.75 18.41 24.13 34.47
water 15.05 -10.24 11.25 18.86 24.80 36.01

ev
Average Cheapest 8-Hour Prices ($/MWh)
Category Mean P5 P25 P50 P75 P95

r
wind 16.42 -8.95 11.99 19.24 25.24 36.71
sun 17.76 -4.78 12.46 19.57 25.92 38.03
natural gas 19.92 -1.36 14.61 20.70 27.02 39.25
uranium
bogas/biomass
coal
20.77
20.30
19.89
0.49
-1.27
-0.48
er
15.18
15.23
14.48
20.93
21.22
20.47
27.34
27.36
26.70
39.60
39.42
38.79
pe
geothermal 18.73 -3.33 13.72 20.28 26.09 37.16
water 18.90 -3.68 14.21 20.75 26.84 38.65

Source: Author's Calculation

ot

Comparing LMP distributions across wind, solar, and natural gas using p5, p10,
tn

p90, and p95 reveals variability in pricing. Figure 3 shows data from 2018 to 2020, although

overlapping sources require an improved presentation for clearer interpretation and

rin

analysis.
ep
Pr

12

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 Figure 3: LMP Percentiles by Power Source Category

ed
iew
r ev
er
pe
Source: Author's calculation

Analyzing the p5 for average 𝑛-cheapest hours (𝑛 = 1–24) reveals that wind, solar,
ot

water, and geothermal sources yield the lowest LMPs, often with negative values from 1
tn

AM to 1 PM (Wang et al., 2024). Non-renewables, such as coal and natural gas, exhibit

higher LMPs, particularly at the 90th and 95th percentiles of the distribution.
rin

Understanding these dynamics enables Bitcoin miners to operate efficiently during

periods of low LMP, highlighting the economic advantage of renewable energy. Future
ep

research will refine these insights.

Pr

13

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 Figure 4: Bar Graph for a fraction of nodes at 𝑛 hours of the day (n/24) for 2018-2020 for selected

ed
power sources, i.e., Wind, Solar, Coal, and Natural Gas

iew
r ev
er
pe
ot
tn

Source: Author's Calculation

rin

Figure 4 reveals that negative LMPs concentrate at hour 13 (1 PM), driven by solar

energy, which accounts for 15.13% of nodes with negative prices. From 8 AM to 5 PM,
ep

wind, coal, and natural gas also show high negative LMP shares. Wind exhibits dominant

negative LMPs between 4:00 and 6:00 AM, 7:00 and 10:00 AM, 11:00 AM, and midnight.
Pr

These trends inform miners of the economic benefits of colocating near specific generators.

14

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 3.1.3. Colocation

ed
This section explores empirical models on Bitcoin mining using 2019 data,

iew
focusing on disparities in LMPs among wind, solar, and natural gas. Targeting average

cheapest hours, I sort wind plant data by LMPs, identifying 23 candidate sites. Figure 5

ranks these plants, offering insights into colocation opportunities for cost-effective

ev
mining based on wind energy generation.

Figure 5: The average negative LMPs (in $/MWh) of 1 and 2 hours Colocating near the Wind Facilities

r
in California, 2019 data

er
pe
ot
tn
rin
ep
Pr

15

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 Figure 6: Top 6 of Colocation Map in California

ed
iew
r ev
er
pe
ot
tn
rin
ep

Source: Author's Calculation using GeoPandas Python

Figure 6 and Table 6 summarize California's top five wind plants, highlighting
Pr

capacity, location, and official plant names. While Bitcoin mining activities are limited

16

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 within these wind generator facilities, it is noteworthy that in 2019, Plouton Mining4

ed
secured $1 million in funding for a proposed sustainable, solar-powered Bitcoin mining

complex in California.

iew
Table 6: Summary of the top 5 wind power plants with negative LMPs in California, 2019
Official Plant Name Capacity City/Town
San Gorgonio Farms 28 M.W. San Gorgonio Pass
Painted Hills Repower 39 M.W. Riverside

ev
Karen Avenue (San Gorgonio 3 M.W. San Gorgonio Pass
Farms)
Mountain View Power 22.2 MW Riverside
Partners II

r
Mojave 16-17-18 84.75 MW Kern
Source: Author's Calculation

er
Having examined wind plant locations and LMPs, I now turn to empirical
pe
modeling to identify variables that influence Bitcoin mining success.
ot

3.2. Methodology

This section introduces a profit-maximizing Bitcoin miner model using LMP

tn

quantiles from Figure 3 and linear trends from Figure 7 to determine optimal mining
rin

hours under perfect electricity price foresight, as follows:

𝑝𝑒𝑗(𝑛) = 𝑎𝑗 + (𝑏𝑗 ∗ 𝑛) (1)

ep

4Bitcoin Magazine (2019) released that the Western Mojave, California, is in the 92nd percentile of solar exposure.
Calvo-Pardo et al. (2022) find that several countries have the highest number of miners in their locations. Focusing on
California, they identified Bitcoin miners near Los Angeles, specifically in the Mojave Desert, where Plouton Mining
Pr

has invested in solar-powered mining, as depicted by Willms (2019).

17

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 where 𝑝𝑒𝑗 is the average 𝑛-cheapest hours of LMP for technology 𝑗 (e.g., wind, solar, and

ed
natural gas), 𝑎𝑗 is the intercept, 𝑏𝑗 is the slope of the Equation, and 𝑛 is the hour from 1 to

24.

iew
The simple profit function:

𝜋𝑗(𝑝𝑏,𝑏𝑛,𝑒𝑖𝑛,𝑛,𝑎𝑗,𝑏𝑗,𝑐𝑖) = (𝑝𝑏·𝑏𝑛·𝑛) ― (𝑝𝑒·𝑒𝑖𝑛·𝑛) ― 𝑐𝑖 (2)

ev
where 𝜋𝑗 is the Profit from technology 𝑗 (wind, solar, and natural gas) in dollars, 𝑝𝑏 is the

r
average of daily bitcoin prices for 3 years (2018-2020) in $1,000 denomination, 𝑏𝑛 is the

er
amount of Bitcoin produced (BTC/hour), 𝑒𝑖𝑛 is the power consumption of a particular 𝑖

ASIC chip in 1 hour (M.W.), and 𝑐𝑖 is the fixed cost of purchasing a particular 𝑖 ASIC chip
pe
to run it at the maximum lifetime per hour in dollars.

From Equations (1) and (2), I can have:

ot

𝜋𝑗(𝑝𝑏,𝑏𝑛,𝑒𝑖𝑛,𝑛,𝑎𝑗,𝑏𝑗,𝑐𝑖) = (𝑝𝑏·𝑏𝑛·𝑛) ― ((𝑎𝑗 + (𝑏𝑗·𝑛))·𝑒𝑖𝑛·𝑛) (3)

The break-even price (𝑝𝑏) is when the profit is zero, meaning that Bitcoin miners prefer
tn

to stop mining when the revenue equals the total costs. So, I can solve 𝑝𝑏 from Equation

(3):
rin

𝜋𝑗(𝑝𝑏,𝑏𝑛,𝑒𝑖𝑛,𝑛,𝑎𝑗,𝑏𝑗,𝑐𝑖) = (𝑝𝑏·𝑏𝑛·𝑛) ― ((𝑎𝑗 + (𝑏𝑗·𝑛))·𝑒𝑖𝑛·𝑛) ― 𝑐𝑖 = 0 (4)

ep

Thus,

1
𝑝𝑏 = 𝑏𝑛·𝑛[(𝑎𝑗 + (𝑏𝑗·𝑛))·𝑒𝑖𝑛·𝑛 ― 𝑐𝑖] (5)
Pr

The profit maximization from Equation (3) with respect to 𝑛:

18

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 ∂𝜋𝑗
= (𝑝𝑏·𝑏𝑛) ― ((𝑎𝑗.𝑒𝑖𝑛) + (2.𝑛.𝑏𝑗.𝑒𝑖𝑛)) = 0 (6)

ed
∂𝑛

Solve for (6), and I get the maximum hour (𝑛∗):

iew
(𝑝𝑏·𝑏𝑛) ― (𝑎𝑗.𝑒𝑖𝑛)
∗ (7)
𝑛 = 2.𝑒𝑖𝑛.𝑏𝑗

ev
Also, to see if Equation (3) is a convex or concave function, I have its second derivative

of Equation (3) with respect to 𝑛:

r
∂2𝜋𝑗
∂𝑛2
= ― 2.𝑏𝑗.𝑒𝑖𝑛 < 0 (8)

er
(concave for 𝑏𝑗.𝑒𝑖𝑛 > 0)

From Equation (8), I can see that the curve of the profit function is concave. Thus, it can
pe
be expected that there will be a decreasing function of the profit or that there will be

maximum or peak points for each particular point.

ot

3.2.1. Empirical Evidence

tn

This section outlines the CAISO dataset and presents the first empirical findings

addressing our initial research question. We introduce a method to assess Bitcoin miners'
rin

potential gains from selective mining based on hourly LMP variations. Using
ep

𝑅𝑒𝑠𝑜𝑢𝑟𝑐𝑒 𝐼𝐷, I focus on wind, solar, and natural gas. For each, I compute the average and

5th percentile LMPs over the 𝑛-cheapest hours (𝑛 = 1–24) from 2018 to 2020. Figure 7
Pr

visualizes that wind consistently yields the lowest average LMPs.

19

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 Figure 7: The bottom 5% and mean of LMPs for Wind, Sun, and Natural Gas Technology

ed
iew
r ev
er
pe
ot
tn

Source: Author's Calculation

rin

The line graph shows how the 𝑛-cheapest hours for key power sources align

linearly with 24-hour operations. Comparing the 5th percentile and mean highlights
ep

volatility versus stability in electricity pricing for miners.

Pr

20

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 Figure 8: The regression plot of the bottom 5% and mean of LMPs for Wind, Sun, and Natural Gas

ed
technology

iew
r ev
er
pe
ot
tn

Source: Author's Calculation

rin

Figure 8 presents regression plots linking operating hours to LMPs, revealing

energy-specific price patterns and equations that model optimal, profitable Bitcoin
ep

mining behavior, as discussed in Section 4.4.

Pr

3.2.2. Empirical Models of The Possibility of Bitcoin Mining

21

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 This section examines empirical models to evaluate the feasibility of Bitcoin

ed
mining. Using regression analysis and hypothesis testing, I uncover key variables that

influence mining success and provide insights into its practical and economic

iew
foundations.

ev
3.2.3. Hypothesis

This section examines how renewable energy generation—solar, wind, and

r
hydropower—impacts Locational Marginal Prices (LMPs) through hypothesis testing,

er
focusing on price differences and the challenges posed by their intermittency and the
pe
proximity of Bitcoin mining operations.

The hypothesis is formulated for hypothesis testing to accomplish this goal:

𝑯𝟎: There is no significant 𝑯𝒂: The average LMPs per hour are
ot

difference in the average systematically lower near

Locational Marginal Prices renewables compared to non-
(LMPs) per hour between renewable energy sources in
tn

California's renewable and California.

non-renewable energy
sources.
rin

The null hypothesis (𝐻0) posits no significant difference in hourly average LMPs
ep

between California's renewable and non-renewable energy sources. The alternative (𝐻𝑎)

suggests renewables systematically lower LMPs. Using CAISO data, this study tests these
Pr

claims, providing insights to inform policy and investment decisions, and supporting the

22

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 sustainable integration of energy in California's transitioning electricity market.

ed
iew
4. Econometrics Analysis and Results

This section examines how hourly electricity prices, also known as Locational

ev
Marginal Prices (LMPs), vary across power sources using CAISO data and econometric

modeling. Building on Panhans et al. (2017), this study examines how hourly generation

r
affects LMPs for sources such as wind, solar, and natural gas. Understanding these

er
variations is crucial for integrating renewable energy and assessing the economic
pe
feasibility of colocating Bitcoin mining near renewable energy generation sites to reduce

energy costs and enhance sustainability. The econometrics model is defined as follows:

𝐿𝑀𝑃 = ∑8𝑘=1 ∑24

𝑛=1 𝛽𝑘𝑛 × 𝑐𝑎𝑡𝑒𝑔𝑜𝑟𝑦𝑘 × ℎ𝑜𝑢𝑟𝑛 +𝜀 (9)
ot

where,
tn

𝐿𝑀𝑃 : Locational Marginal Prices

ℎ𝑜𝑢𝑟 : n ∈ {1, . . . , 24}

rin

𝑐𝑎𝑡𝑒𝑔𝑜𝑟𝑦 : Eight power sources of interest, 𝑘 = {Wind, Sun, Natural Gas, Uranium, Coal,

Biogas/Biomass, Geothermal, Water}

ep

Equation (9) represents a multiple regression model that aims to explain the

Locational Marginal Prices (LMPs) based on the interaction between different power source
Pr

categories and the hourly effect. The Equation consists of several components:

23

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 1. The terms 𝑐𝑎𝑡𝑒𝑔𝑜𝑟𝑦𝑘 and ℎ𝑜𝑢𝑟𝑛 denote the categorical variables for a power source

ed
category and hourly effect. The Equation includes eight power source categories:

Biogas/Biomass, Coal, Geothermal, Natural Gas, Sun, Uranium, Water, and Wind.

iew
Additionally, it accounts for 24 hours daily to consider the hourly variations in LMPs.

2. The interaction term, 𝑐𝑎𝑡𝑒𝑔𝑜𝑟𝑦𝑘 × ℎ𝑜𝑢𝑟𝑛, captures the combined effect of the power

ev
source category and the specific hour on the LMPs. This term enables an analysis of

how different power sources and hours affect prices.

r
3. Lastly, 𝜀 represents the error term, which accounts for unexplained variability in the

er
LMPs that the other variables in the model cannot capture.

Estimating these coefficients reveals how each power source and hour influences
pe
LMPs, highlighting economic dynamics between renewable and non-renewable

electricity generation.
ot
tn

4.1. Hourly Effect on LMP

This section analyzes 24-hour LMP variations across eight energy sources,
rin

revealing pricing dynamics shown in Figure 9.

ep
Pr

24

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 Figure 9: The coefficient plots for the eight power sources

ed
iew
r ev
Source: Author's Calculation
er
pe
Figure 9 illustrates the dynamic nature of LMPs throughout the day, highlighting

how various energy sources influence pricing. Solar energy consistently shows the lowest
ot

average LMPs, although it is absent during hours 0, 6, and 18 due to its dependency on
tn

sunlight. From hours 7 to 17, solar LMPs remain low but are generally higher than those

of wind, making wind more cost-effective. Early morning LMPs are driven down by
rin

Geothermal, Water, and Wind, but non-renewables dominate after hour 15.

In summary, our analysis confirms the consistent cost-effectiveness of solar power,

ep

which maintains the lowest average LMPs during hours 7 to 17, despite its absence at

night. Regression results highlight hourly LMP variations, showing geothermal, water,
Pr

and wind drive low prices early, while non-renewables dominate midday. This shift

25

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 emphasizes strategic Bitcoin mining during periods of low-cost electricity. Additionally,

ed
the residual plots and KDE in Figure 10 illustrate distinct price characteristics across

power sources, supporting the model's adequacy and providing distributional insights.

iew
r ev
er
pe
ot
tn
rin
ep
Pr

26

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 Figure 10: Residual plot for all power sources

ed
iew
r ev
er
pe
ot
tn
rin
ep

Source: Author's Calculation

Pr

The KDE analysis of residuals reveals distinct distributional patterns across power

27

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 sources, guiding strategic Bitcoin mining. Biomass/Biogas, Coal, and Water exhibit flat,

ed
symmetrical plots, indicating stable LMPs. Natural Gas and Geothermal show left skews,

while Wind is right-skewed, suggesting price volatility. Sun and Uranium display sharp

iew
peaks around specific values. These differences underscore the value of colocating near

renewable energy sources—wind, Solar, Geothermal, and Water—where lower LMPs

ev
support cost-effective and sustainable mining.

r
4.2. Hypothesis Testing: Comparative Electricity Prices

er
To rigorously test our hypothesis regarding the electricity prices of renewable

energy sources, I perform a 𝑡-test, a widely used statistical test to assess the significance
pe
of mean differences between groups. In this context, the 𝑡-test allows us to examine

whether the mean electricity prices of renewables are significantly lower compared to
ot

non-renewable sources. The t-test results offer valuable insights into the relationship
tn

between renewable energy generation and electricity prices, shedding light on the

economic viability and competitiveness of renewable energy sources in the energy

rin

market.

We employ a rigorous methodology that involves statistical analysis. Then, I

ep

collect data on electricity prices (LMP) across various energy source categories, including

renewable and non-renewable sources. The dataset comprises observations for each hour
Pr

of the day (𝑛) and different energy source categories (𝑘). Then, the 𝑡-test assesses whether

28

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 there is a statistically significant difference in the mean electricity prices between

ed
renewables and non-renewables. The LMP variable is the outcome variable, while the

"renewables dummy" variable functions as the grouping variable. Group 𝐹𝑜𝑠𝑠𝑖𝑙

iew
corresponds to the mean LMP when the "renewables dummy" variable equals 0. In other

words, Group 𝐹𝑜𝑠𝑠𝑖𝑙 represents the average LMP during periods with no or negligible

ev
renewable energy generation. That can include times when traditional non-renewable

sources, such as coal, natural gas, or nuclear power, dominate the electricity supply. The

r
mean LMP for Group 𝐹𝑜𝑠𝑠𝑖𝑙 is 35.5134, with a standard error of 0.0168. This group serves

er
as a reference point for comparing LMPs in the presence of renewable energy generation.

Group 𝑅𝑒𝑛𝑒𝑤𝑎𝑏𝑙𝑒𝑠 is the mean LMP when the "renewables dummy" variable
pe
equals 1. Group 𝑅𝑒𝑛𝑒𝑤𝑎𝑏𝑙𝑒𝑠 represents the average LMP when significant renewable

energy generation contributes to the electricity supply. Renewable energy sources, such
ot

as wind, solar, geothermal, and hydroelectric power, are likely to be the primary drivers
tn

in this group. The mean LMP for Group 𝑅𝑒𝑛𝑒𝑤𝑎𝑏𝑙𝑒𝑠 is 33.5790, with a standard error of

0.010. Group 𝑅𝑒𝑛𝑒𝑤𝑎𝑏𝑙𝑒𝑠 allows for examining LMPs, specifically during substantial
rin

renewable energy generation periods.

Based on the 𝑡-test results in Table 7, I observe a significant difference between the
ep

mean electricity prices of renewables and non-renewables. Specifically, renewables

exhibit prices 1.9344 units lower than those of non-renewables, with a standard error of
Pr

0.019. The high 𝑡-value of 100.1711 indicates a robust statistical significance, supporting

29

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 our alternative hypothesis that renewables have lower electricity prices than non-

ed
renewables.

These findings suggest that colocating Bitcoin mining near renewable energy

iew
sources can lower costs and support sustainable energy practices and environmental

goals.

ev
Table 7: t-test result
Group Mean Std. Error Std. Dev. 95% CI
𝐹𝑜𝑠𝑠𝑖𝑙 35.5134 0.0168 45.7125 35.4805-35.5463
𝑅𝑒𝑛𝑒𝑤𝑎𝑏𝑙𝑒𝑠 33.5790 0.0101 44.1558 33.5592-33.5988

r
Combined 34.1 0.0087 44.6045 34.1023-34.1363
Diff 1.93 0.0193 - 1.8966 - 1.973
𝑡-values 100.1711
Source: Author's Calculation
-
er - -
pe
4.3. Quantile regression

This section uses quantile regression to analyze LMP variability across wind, solar,
ot

natural gas, and coal at five quantiles (q5, q10, q50, q90, q95). The approach highlights

electricity price fluctuations by time and source, particularly during periods of high
tn

renewable energy output. It identifies optimal, cost-effective hours for Bitcoin mining,

enabling strategic operation and enhanced understanding of market dynamics.

rin

The quantile regression model is formulated as follows:

(10)
ep

𝐿𝑀𝑃𝑠,𝑞 = 𝑋𝛽𝑠,𝑞 + 𝜖𝑞

where:
Pr

𝐿𝑀𝑃𝑠,𝑞 : Predicted LMP for energy source 𝑠 at quantile 𝑞.

30

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 𝑋: Matrix of independent variables:

ed
Energy source: Dummy variables for Wind, Sun, Natural Gas, and Coal.

Time-of-day effects: Hourly demand patterns, seasonality.

iew
Market conditions: Fuel prices, transmission constraints, renewable penetration.

𝛽𝑠,𝑞: Coefficients specific to source 𝑠 and quantile 𝑞.

ev
𝜖𝑠,𝑞 : Error term with 𝜏𝑞-quantile = 0.

Estimation:

r
𝑛
𝛽𝑠,𝑞 = 𝑎𝑟𝑔 𝑚𝑖𝑛 ∑𝑖=1 𝜌𝑞 ( 𝐿𝑀𝑃𝑖 ― 𝑋𝑖 𝛽) (11)
𝛽

er
𝜌𝑞(𝑢) = 𝑢 ⋅ (𝑞 ― 𝐼(𝑢 < 0)) is the quantile loss function.
pe
Then, to look at the Cheapest LMP per Energy Source, I calculated the following:

For each hour (ℎ), compute the minimum expected LMP across quantiles for each source

(𝑠):
ot

𝐿𝑀𝑃𝑚𝑖𝑛,𝑠 = 𝑚𝑖𝑛(𝐸[𝐿𝑀𝑃𝑠,𝑞∣ℎ]) (12)

𝑞
tn

For Example: 𝑊𝑖𝑛𝑑 (𝑞5) = $25/MWh; 𝑆𝑢𝑛 (𝑞5) = $30/MWh; 𝑁𝑎𝑡𝑢𝑟𝑎𝑙 𝐺𝑎𝑠 (𝑞5) = $45

/MWh; 𝐶𝑜𝑎𝑙 (𝑞5) = $50/MWh, then 𝐿𝑀𝑃𝑚𝑖𝑛,Wind = 25,𝐿𝑀𝑃𝑚𝑖𝑛,Sun = 30

rin

The last step is to compute the average LMP per quantile and hour. For each

quantile (𝑞) and hour (ℎ), average LMP across all energy sources:
ep

1 𝑁
𝑠
𝐿𝑀𝑃ℎ,𝑞 = 𝑁𝑠∑𝑠=1 𝐸[𝐿𝑀𝑃𝑠,𝑞∣ℎ] (13)
Pr

where: 𝑁𝑠 = 4 (Wind, Sun, Natural Gas, Coal).

31

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 The subsequent analysis identifies the minimum expected LMP across quantiles

ed
for each source and provides an hourly comparison, as visualized in Figure 11.

Figure 11: Cheapest LMP by Hour for Each Energy Source (Quantile-Based)

iew
r ev
er
pe

Source: Author's Calculation

ot

The findings reinforce prior research on the integration of renewable energy and
tn

electricity pricing. While Stoll et al. (2019) emphasized Bitcoin mining's dependence on

fossil fuels, our results suggest that it can align with renewable energy sources by
rin

absorbing surplus power. Low LMPs in q05 and q10, especially for wind and solar,

support the findings of Winton et al. (2021) and Jiang et al. (2021). Sharp LMP spikes for
ep

fossil fuels in q90 and q95 confirm Hirth's (2018) and Borenstein et al.'s (2019) findings
Pr

on volatility patterns.

32

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 Wind energy exhibits the lowest and most stable LMPs between 1 AM and 5 AM,

ed
which confirms earlier findings by Deetjen et al. (2019) that wind generation tends to

peak during low-demand nighttime hours. Solar energy follows a distinct daytime

iew
pattern, with the lowest LMPs occurring between 7 AM and 3 PM, which reflects Cullen's

(2013) conclusion that solar generation significantly reduces wholesale electricity prices

ev
during periods of high sunlight. In contrast, natural gas and coal exhibit a steady price

increase throughout the day, peaking between 4 PM and 9 PM—consistent with Bushnell

r
and Novan's (2018) research, which attributes these surges to the high ramping costs of

fossil fuel plants. er

These results strengthen the view of Bitcoin mining as a flexible, location-sensitive
pe
energy consumer. It can absorb excess electricity and reduce curtailment, as proposed by

Winton (2021) and Shynkevich (2021). Furthermore, LMP volatility supports Hirth's
ot

(2018) and Bistline and Blanford's (2020) call for flexible energy systems.
tn

This study also demonstrates that Bitcoin mining can become more profitable with

lower electricity costs, providing a market-driven solution to reduce grid inefficiencies

rin

and support the integration of renewable energy, in line with the findings of Stoll et al.

(2019) and Jiang et al. (2021).

ep

4.4. Quantitative Results

Pr

33

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 4.4.1. Calibration

ed
In this section, I calibrate the theoretical model presented above and then present

three simulations that illustrate the Bitcoin mining profitability model.

iew
The calibration of the parameters is as follows:

1. Parameter 𝑝𝑏: average Bitcoin price ($) for 3 years from 2018 to 2020 in a $1000

ev
denomination.

2. Parameter 𝑏𝑛: the amount of Bitcoin (BTC) per hour of observation. For BTC

r
calculation, it is as follows:

a. er
A new block is generated every 10 minutes. There are approximately 6 blocks per
pe
hour, 24 hours per day, and 30 days per month, resulting in 6 × 24 × 30 = 4, 320

blocks.

b. The current block reward is 12.5 BTC/block (until May 2020), so

ot

4,320 × 12.5 = 53,200 BTC will be created monthly. Antminer S9 (13.5Th)

tn

yields a theoretical hash rate of about 13.5 Th/s.

c. The current Bitcoin Network Hash Rate5 is about 51,500,000 Th/s, so Antminer S9
rin

(13.5 Th/s) represents about 13.5/51, 500, 000 = 0.00000026213592 of the

Bitcoin Network Hash Rate.

ep

d. So, this particular chip's average monthly personal Bitcoin reward is

Pr

5Hash rate measures the computational power on a blockchain network. It is determined by how many guesses are
made per second. The hash rate helps determine a blockchain network's security and mining difficulty.

34

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 54,000 × 0.000000262135922 = 0.01 𝐵𝑇𝐶, i.e.

ed
0.01/720 = 0.00001966 BTC/hour.

3. Parameter 𝑐𝑖: fixed cost for acquiring a particular chip.

iew
4. Parameter 𝑒𝑖𝑛: A particular ASIC chip consumption in one hour (in M.W.).

5. Parameters 𝑎𝑗 and 𝑏𝑗 are the intercept and the slope of the average 𝑛-cheapest hours

ev
of LMP (in a $1000 denomination).

6. Parameter 𝑛: hours of the day.

r
4.4.2. Baseline Results

er
We calculate and graph optimal mining hours (𝑛∗) using Bitcoin prices (2018–2020)
pe
and Table 8 parameters for Wind, Solar, Gas p05 with Antminer S9.

Table 8: Calibrated parameters for Antminer S9 (13.5Th)

Parameters Wind p05 Solar p05 Gas p05
𝑝𝑏𝑚𝑎𝑥 (in $ per BTC) 29.4 29.4 29.4
ot

𝑝𝑏𝑚𝑖𝑛 (in $ per BTC) 3.2 3.2 3.2

𝑝𝑏𝑎𝑣𝑔 (in $ per BTC) 8.7 8.7 8.7
𝑏𝑛 (in BTC per hour) 0.00001966 0.00001966 0.00001966
𝑒𝑛𝑖 (in M.W.) 0.001323 0.001323 0.001323
tn

𝑎𝑗 (in $ per MWh) −21.8516 −15.5056 −11.0850

𝑏𝑗 (in $ per MWh) 1.4195 1.2103 1.1039
𝑐𝑖 (in $ per MWh) 0.02495 0.02495 0.02495
Source: Author's Calculation
rin

For different chips, I also utilize the newer ASICs, MicroBT Whatsminer M31S+,
ep

which has a higher hash rate (80 Th/s) but a higher fixed or acquisition cost. Table 8 also

comprehensively summarizes the parameters that calculate crucial measures, including

Pr

35

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 the necessary profit, break-even prices, and maximum mining hours. These parameters

ed
are derived from Equations 4, 5, and 7 based on the regression plot depicted in Figure 8.

The values obtained from the regression analysis form the basis for these calculations,

iew
enabling us to quantify profitability and assess the feasibility of Bitcoin mining operations

during different periods.

ev
In Table 8, the minimum (𝑝𝑏𝑚𝑖𝑛), maximum (𝑝𝑏𝑚𝑎𝑥), and the average price of

Bitcoin (𝑝𝑏𝑎𝑣𝑔) are vital indicators representing the lowest, highest, and average values of

r
daily Bitcoin prices recorded from January 2018 to December 2020. These values serve as

er
reference points for evaluating the profitability of mining operations in relation to the

fluctuating Bitcoin market.

pe
The parameters 𝑎𝑗, 𝑏𝑗, and 𝑐𝑖 also represent the intercepts, slopes, and fixed costs

associated with the respective technologies: wind p05, solar p05, and gas p05. These
ot

parameters capture the unique characteristics and cost structures of each technology,
tn

enabling precise calculations of profit margins, break-even prices, and maximum mining

hours.
rin

The negative intercepts suggest a surplus of electricity, enabling Bitcoin miners to

operate during periods of minimal or negative costs, thereby enhancing profitability

ep

under specific technological and timing conditions.

Pr

4.4.3. Sensitivity Analysis

36

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 This section examines the impact of switching from wind to solar or natural gas

ed
on mining outcomes. It examines miner behavior using LMPs from these sources and

considers varying Bitcoin price scenarios.

iew
Figure 12: Bitcoin prices and the optimal hours to mine each day using ASIC: (a) Antminer S9 (13.5Th),
and (b) the newer chip, MicroBT Whatsminer M31S+ (80 Th)

r ev
(a)
er (b)
pe
Source: Author's Calculation
Figures 12a and 12b show how optimal mining hours vary with Bitcoin prices for

Wind, Solar, and Gas p05 technologies. All three intersect at 16 hours and around $1,800.
ot

Wind remains the most profitable option across price ranges, especially beyond $3,000.
tn

More efficient technologies, as shown in Figure 12b, narrow price gaps and increase

mining hours as Bitcoin prices rise. The differences in the intercept (𝑎𝑗) and slope (𝑏𝑗)
rin

values in the electricity price equation, i.e., Equation (1), for each energy source (wind,

solar, and gas), reflect variations in the cost structure and price dynamics associated with
ep

different technologies. However, the analysis in this study focuses on determining the

optimal mining hours (𝑛∗) for each energy source rather than directly examining how these
Pr

differences in 𝑎𝑗 and 𝑏𝑗 values affect the number of hours miners would use each energy

37

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 source. Moreover, Wind technology supports longer mining hours than solar or gas, as

ed
Bitcoin prices rise, highlighting the price sensitivity of Bitcoin to mining decisions.

However, as chip efficiency improves, this sensitivity lessens. Ultimately, the hypothesis

iew
from Dorsey and Musk—that miners would selectively mine based on energy prices—no

longer holds; miners continue operating until mining becomes unprofitable.

ev
Furthermore, an examination was conducted to determine the break-even point

prices (𝑝𝑏 ) for various technologies, including wind p05, solar p05, and gas p05, in

r
conjunction with different mining chips. The break-even point marks the point at which

er
Bitcoin mining becomes unprofitable—when total and average profits equal zero and

marginal profit ceases. Figure 13 illustrates the relationship between break-even Bitcoin
pe
prices and mining hours, showing that prices decline sharply between 1 AM and 5 AM,

then taper off gradually. This trend is consistent across all observed energy sources,
ot

indicating that early morning hours offer greater mining efficiency. Importantly, the cost
tn

analysis incorporates the one-time purchase of the MicroBT Whatsminer M31S+ chip (80

TH), which is 6.65 times more expensive than the Antminer S9 (13.5 TH). This capital
rin

expense has a significant impact on profitability and contrasts with the previous model

(Figure 12), which excluded fixed costs. Accounting for these differences allows for a
ep

more realistic assessment of optimal mining times and the economic feasibility of various

mining hardware options under different market and energy conditions.

Pr

38

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 Figure 13: Break-even Prices of Bitcoin for each ASIC: Antminer S9 (13.5Th) and the newer chip,

ed
MicroBT Whatsminer M31S+ (80 Th) for Wind p05, Solar p05, and Gas p05

iew
r ev
er
pe
ot

Source: Author's Calculation

tn

The analysis of the maximum achievable profit and corresponding optimal mining

hours (𝑛∗) depicted in Figures 14a, 14b, and 14c reinforces the findings observed in Figure
rin

13. Across all three power sources, the profit maximization analysis indicates that Bitcoin

miners can maximize their profits by mining for more than 4 hours without imposing
ep

restrictions on their mining hours. Miners can extend operations beyond four hours to

boost profits, with wind power offering the highest returns, while solar and gas yield
Pr

comparatively lower returns for mining. Prioritizing wind optimizes earnings.

39

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 Furthermore, addressing the limitations imposed by the restricted availability of

ed
solar power is important. Since solar power can only be utilized for a hypothetical

maximum of 12.43 hours per day, I have set an upper bound of 12.43 on the 𝑥-axis for solar

iew
observations. The theoretical model suggests that mining for 24 hours a day would yield

the best results, but in practice, it may not be feasible or desirable. Thus, I have included

ev
a vertical line to indicate that solar mining operations would cease after 12.43 PM.

Additionally, it is essential to note that while solar power has certain limitations due to

r
its reliance on sunlight, wind power offers a reliable alternative by providing a consistent

er
energy source, regardless of the time of day or weather conditions. This further

underscores the rationale for considering wind power as a viable option for Bitcoin
pe
mining operations.

It is also interesting to note that the optimal profit for Wind p05 starts around 7.50
ot

AM and continues for 24 hours. In contrast, for Solar p05, the optimal mining hours range
tn

from approximately 6.30 AM to 12.43 PM. Gas p05, on the other hand, indicates optimal

mining hours starting as early as 5 AM and continuing for 24 hours. This analysis offers
rin

valuable insights for Bitcoin miners, enabling them to make informed decisions about

their operational strategies and capitalize on the most financially profitable opportunities.
ep
Pr

40

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 Figure 14: Profit maximization with the optimal hour of different chips using (a) wind p05, (b)

ed
solar p05, and (c) Gas p05

iew
r ev
(a)

er
pe
ot
tn

(b)
rin
ep
Pr

41

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 Figure 14: Profit maximization with the optimal hour of different chips using (a) wind p05, (b)

ed
solar p05, and (c) Gas p05

iew
r ev
(c)

er
pe
Several limitations should be considered in this study. Firstly, it assumes that the

observed LMPs from power sources in CAISO are isolated from adjacent grid
ot

interconnections within the USA, potentially overlooking trading dynamics and

fluctuating electricity prices resulting from interconnections with entities such as ERCOT,
tn

Alberta Electric System Operator, or Southwest Power Pool. Secondly, due to time
rin

constraints and the complex structure of the CAISO dataset, the study does not consider

the load of each power source, which can have a significant impact on Bitcoin mining
ep

operations. Additionally, while the Bitcoin mining market is described as oligopolistic

with asymmetric firms, this analysis assumes a free-entry market. It does not directly
Pr

consider the geographical factors that influence mining locations, as discussed by

42

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 Delgado-Mohatar et al. (2019). It is important to note that most bitcoin mining operations

ed
currently colocate near hydropower facilities in China and the United States, as

highlighted by Schinckus (2021) and Zhang et al. (2019), due to factors such as affordable

iew
electricity costs, surplus power availability, and favorable feed-in tariff rates. Wind and

solar power remain viable energy sources for Bitcoin mining. Wind offers renewable

ev
abundance, while solar ensures scalable, eco-friendly production. Incorporating both

diversifies energy sources, promotes sustainability, and reduces reliance on a single

r
source. Despite hydropower's dominance, wind and solar align with renewable goals and

er
support the long-term viability of environmentally conscious mining operations.
pe
5. Conclusion and Policy Recommendation
ot

The analysis of LMPs in California's energy market, using CAISO data from 2018 to

2020, demonstrates how Bitcoin mining can strategically leverage renewable energy
tn

sources, such as wind and solar, to reduce operational costs while supporting sustainable

energy integration. During periods of high renewable generation, particularly when

rin

supply outstrips demand, electricity prices often drop below zero, creating unique

arbitrage opportunities for miners. Wind energy emerges as especially advantageous,

ep

offering consistently lower prices than solar during peak generation windows. This price

differential and the inherent variability of renewables suggest that aligning mining
Pr

43

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 operations with renewable energy surpluses could enhance profitability and incentivize

ed
the development of cleaner energy infrastructure.

However, the study acknowledges limitations, particularly the lack of granular

iew
power consumption data from CAISO, which restricts a complete assessment of mining's

economic feasibility. To refine these models, future research should incorporate energy

ev
demand metrics from reputable sources, such as the U.S. Energy Information

Administration (EIA) or the International Energy Agency (IEA). Additionally, while

r
advances in ASIC technology (such as Bitmain's Antminer T19 Hydro) have dramatically

er
improved mining efficiency, the interplay between equipment costs, depreciation

timelines, and volatile electricity prices requires further exploration.

pe
From a policy perspective, governments could accelerate this synergy by

introducing targeted incentives, such as tax breaks for renewable-powered mining facilities
ot

or streamlined permitting for colocation projects. Miners, meanwhile, should prioritize

tn

wind-rich regions and optimize operations around real-time LMP fluctuations to maximize

returns. Over the long term, the marriage of Bitcoin mining and renewables could evolve
rin

into a stabilizing force for energy grids, turning intermittent generation into a predictable

revenue stream while advancing decarbonization goals.

ep

Ultimately, this research highlights a critical insight: the future of sustainable

Bitcoin mining depends on three key pillars—data-driven operational strategies, ongoing

Pr

hardware innovation, and policy frameworks that incentivize the use of clean energy. The

44

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 industry can transform price volatility from a challenge into a competitive edge by

ed
addressing these areas in tandem.

iew
ev
REFERENCE

Aki, J. (2019, October 16). Layer1 Secures $50M In Funding To Rebalance The Global Bitcoin

r
Hash Power. Bitcoin Magazine. https://bitcoinmagazine.com/business/layer1-
secures-50m-in-funding-to-rebalance-the-global-bitcoin-hash-power.

er
Antweiler, W. (2021). Microeconomic models of electricity storage: Price forecasting,
arbitrage limits, curtailment insurance, and transmission line utilization. Energy
Economics, 101:105390.
pe
Badea, L. and Mungiu-Pupazan, M. C. (2021). The economic and environmental impact
of bitcoin. IEEE access, 9:48091–48104.
Bistline, J. E., & Blanford, G. J. (2020). Value of technology in the US electric power sector:
Impacts of full portfolios and technological change on the costs of meeting
ot

decarbonization goals. Energy Economics, 86, 104694.

Borenstein, S., Bushnell, J., Wolak, F. A., & Zaragoza-Watkins, M. (2019). Expecting the
tn

unexpected: Emissions uncertainty and environmental market design. American

Economic Review, 109(11), 3953-3977.
Burkhardt, J., Bayham, J., Wilson, A., Carter, E., Berman, J. D., O'Dell, K., Ford, B., Fierce,
E. V. & Pierce, J. R. (2019). The effect of pollution on crime: Evidence from data
rin

on particulate matter and ozone. Journal of Environmental Economics and

Management, 98, 102267.
Bushnell, J., & Novan, K. (2018). Setting with the sun: The impacts of renewable energy on
ep

wholesale power markets (No. w24980). National Bureau of Economic Research.

California ISO. (2023). Business practice manual for market operations (BPM-002, rev. 15).
https://www.caiso.com
Pr

Calvo-Pardo, H. F., Mancini, T., & Olmo, J. (2022). Machine learning the carbon footprint
of bitcoin mining. Journal of Risk and Financial Management, 15(2), 71.

45

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 Calvo-Pardo, H. F., Mancini, T., & Olmo, J. (2022). Machine learning the carbon footprint

ed
of bitcoin mining. Journal of Risk and Financial Management, 15(2), 71.
CoinMarketCap. (2021, August 21). What Is a Blockchain? CoinMarketCap Academy;
CoinMarketCap. https://coinmarketcap.com/academy/article/what-is-a-
blockchain

iew
Cretì, A., & Fontini, F. (2019). Economics of electricity: Markets, competition and rules.
Cambridge University Press.
Cullen, J. (2013). Measuring the environmental benefits of wind-generated electricity.
American Economic Journal: Economic Policy, 5(4), 107-133.

ev
Deetjen, T. A., & Azevedo, I. L. (2019). Reduced-order dispatch model for simulating
marginal emissions factors for the United States power sector. Environmental
science & technology, 53(17), 10506-10513.

r
Delgado-Mohatar, O., Felis-Rota, M., and Fernández-Herraiz, C. (2019). The bitcoin
mining breakdown: Is mining still profitable? Economics Letters, 184:108492.

er
Fell, H., Kaffine, D. T., & Novan, K. (2021). Emissions, transmission, and the
environmental value of renewable energy. American Economic Journal: Economic
Policy, 13(2), 241-272.
pe
Flammini, M. G., Prettico, G., Mazza, A., & Chicco, G. (2021). Reducing fossil fuel-based
generation: Impact on wholesale electricity market prices in the North-Italy
bidding zone. Electric Power Systems Research, 194, 107095.
Greenberg, P., & Bugden, D. (2019). Energy consumption boomtowns in the United
ot

States: Community responses to a cryptocurrency boom. Energy Research & Social

Science, 50, 162-167.
tn

Greenberg, P., & Bugden, D. (2019). Energy consumption boomtowns in the United
States: Community responses to a cryptocurrency boom. Energy Research & Social
Science, 50, 162-167.
Hakimi, A., Pazuki, M. M., Salimi, M., & Amidpour, M. (2024). Renewable energy and
rin

cryptocurrency: A dual approach to economic viability and environmental

sustainability. Heliyon, 10(22).
Hallinan, K. P., Hao, L., Mulford, R., Bower, L., Russell, K., Mitchell, A., & Schroeder, A.
ep

(2023). Review and demonstration of the potential of bitcoin mining as a

productive use of energy (PUE) to aid equitable investment in solar micro-and
mini-grids worldwide. Energies, 16(3), 1200.
Pr

Hirth, L. (2018). What caused the drop in European electricity prices? A factor
decomposition analysis. The Energy Journal, 39(1), 143-158.

46

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 Hogan, W. W. (2010). Electricity wholesale market design in a low-carbon future. In

ed
Harnessing Renewable energy in electric power systems (pp. 113-136). Routledge.
Hughes, S. D. (2017). Cryptocurrency Regulations and Enforcement in the US. W. St. UL
Rev., 45, 1.

iew
Hutabarat, S. P. H. (2023). Essays on Bitcoin Mining and Renewable Energy: Exploring
Sustainability and Profitability (Doctoral dissertation, Colorado State University).
Jiang, S., Li, Y., Lu, Q., Hong, Y., Guan, D., Xiong, Y., and Wang, S. (2021). Policy
assessments for the carbon emission flows and sustainability of bitcoin blockchain
operation in china. Nature communications, 12(1):1–10.

ev
Joskow, P. L. (2019). Challenges for wholesale electricity markets with intermittent
renewable generation at scale: the US experience. Oxford Review of Economic Policy,
35(2), 291-331.

r
Mills, A., Wiser, R., Millstein, D., Carvallo, J. P., Gorman, W., Seel, J., & Jeong, S. (2021).
The impact of wind, solar, and other factors on the decline in wholesale power

er
prices in the United States. Applied Energy, 283, 116266.
Niaz, H., Liu, J. J., & You, F. (2022). Can Texas mitigate wind and solar curtailments by
leveraging bitcoin mining?. Journal of Cleaner Production, 364, 132700.
pe
Panhans, M., Lavric, L., and Hanley, N. (2017). The effects of electricity costs on firm
relocation decisions: insights for the pollution havens hypothesis? Environmental
and Resource Economics, 68:893–914.
Schinckus, C. (2021). Proof-of-work based blockchain technology and anthropocene: An
ot

undermined situation? Renewable and Sustainable Energy Reviews, 152:111682.

Shynkevich, A. (2021). Bitcoin arbitrage. Finance Research Letters, 40, 101698.
tn

Stoll, C., Klaaßen, L., and Gallersdörfer, U. (2019). The carbon footprint of bitcoin. Joule,
3(7):1647–1661.
The causes and effects of negative power prices. (2018, November 26). Clean Energy Wire.
rin

https://www.cleanenergywire.org/factsheets/why-power-prices-turn-negative.
Velický, M. (2023). Renewable energy transition facilitated by Bitcoin. ACS Sustainable
Chemistry & Engineering, 11(8), 3160-3169.
ep

Vlachos, E., Kaushik, A., & Thompson, J. (2018, June). Energy efficient transmitter with
low resolution DACs for massive MIMO with partially connected hybrid
architecture. In 2018 IEEE 87th Vehicular Technology Conference (VTC Spring) (pp.
1-5). IEEE.
Pr

Wang, G., Sbai, E., Wen, L., & Sheng, M. S. (2024). The impact of renewable energy on

47

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010
 extreme volatility in wholesale electricity prices: Evidence from organisation for

ed
economic co-operation and development countries. Journal of Cleaner Production,
484, 144343.
Willms, J. (2019). Bitcoin Mining In North America: A New Gold Rush In The New World.
Bitcoin Magazine, December, 3.

iew
Winton, B. (2021, April 21). Solar + Battery + Bitcoin Mining. Medium.
https://wintonark.medium.com/bitcoin-mining-impact-on-renewable-uptake-
fc91c5aa9be0.

Zhang, H., Chang, J., Gao, C., Wu, H., Wang, Y., Lei, K., Long, R. & Zhang, L. (2019).

ev
Cascade hydropower plants operation considering comprehensive ecological
water demands. Energy Conversion and Management, 180, 119-133.
Zribi, W., Boufateh, T., & Guesmi, K. (2023). Climate uncertainty effects on bitcoin

r
ecological footprint through cryptocurrency environmental attention. Finance
Research Letters, 58, 104584.

er
pe
ot
tn
rin
ep
Pr

48

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5360010