Source: Galaxy; Compiled by: Bai Shui, Golden Finance
Summary
Bitcoin miners with large-scale land, cooling water, dark fiber, reliable electricity, skilled labor, electrical approvals, and critical long-term delivery cycle infrastructure components are well-positioned to enhance the value of their assets by meeting the rapidly growing demand of the AI/HPC data center market.
Goldman Sachs Research predicts that by 2030, the demand for data centers in the U.S. will reach 45 GW, with electricity demand growing at a compound annual growth rate of 15% from 2023 to 2030 driven by AI.
JPMorgan expects that by 2038, capital expenditures for hyperscale AI will reach $370 billion, an increase of 127% compared to the projected AI capital expenditures in 2024.
The surge in connection requests for facilities of 300 MW to 1,000 MW or more has stressed the ability of local grids to deliver power at such a rapid pace, resulting in interconnection and construction timelines extended by 2-4 years.
Traditional data centers do not have the high power capacity to support high-density computing operations. Server racks that once had a maximum power of 40 kW per rack now need to support over 132 kW per rack, which is necessary for cutting-edge systems such as the GB200 NVL72.
Predictable cash flow, an active financing market, and significant valuation upside potential for AI/HPC operations make this opportunity highly attractive and value-generating for miners with the right assets.
Miners can unlock tremendous value by transitioning to the AI/HPC market, arbitraging their 6-12 times EV/EBITDA multiples against the typical 20-25 times multiples of current leading data center operators.
Introduction
The rise of artificial intelligence (AI) is creating unprecedented demand for high-capacity computing (HPC) facilities. This surge has led hyperscale companies to make significant investments in new data center capacity. However, due to limited power capacity, the construction timeline for new facilities has stretched to 2-4 years, making it difficult for traditional data centers to meet these demands.
Bitcoin miners have a unique advantage to capitalize on this market opportunity as they have already secured the critical components required for large-scale power infrastructure and data center operations. Although not all mining facilities can be converted into AI data centers due to specific requirements for cooling, networking, and redundancy systems, those with the right assets and expertise stand to benefit from the high cash flow margins and tremendous valuation potential of AI/HPC operations. This report examines the current landscape of traditional data centers and highlights the specific barriers to meeting AI computing demands. It then analyzes why certain types of Bitcoin miners are well-positioned to fill this gap and explores future trends at the intersection of Bitcoin mining and AI infrastructure.
What are the opportunities for AI data centers?
AI is set to flourish in 2024, driven by the widespread adoption of generative AI (GenAI) technologies. According to Pitchbook, over 100,000 deals have invested more than $680 billion in AI and machine learning startups since 2016, with $120 billion invested in just 2024.
The surge in artificial intelligence and high-performance computing (HPC) is placing immense demand on data center capacity. Data centers are critical for the operation of AI/HPC, providing the necessary infrastructure and power for GPU-intensive computing. New AI applications, such as large language models (LLMs), are particularly energy-hungry. The International Energy Agency states that a single ChatGPT query requires 2.9 watt-hours of electricity, while a Google search only requires 0.3 watt-hours.
The emergence of energy-intensive AI/HPC businesses in the U.S. is driving the growth of demand for data centers. Goldman Sachs Research estimates that by 2024, demand for data centers in the U.S. will reach 21 GW (a year-over-year increase of 31%). For reference, the estimated compound annual growth rate for data center demand in the U.S. from 2022-2033 is 15.8%. Based on the significant year-over-year growth in data center demand projected for 2024, Goldman Sachs Research expects that U.S. data center demand will increase to 45 GW by 2030. By 2030, U.S. data centers will consume 45 GW of electricity, accounting for 8% of the total electricity capacity in the U.S.
The market opportunity for data centers in the U.S. will be supported by the increased investments from hyperscale companies in AI infrastructure, with hyperscale players like Google Cloud and AWS being able to rapidly scale data center capacity to serve other business clients. These hyperscale companies have committed to investing over $100 billion in AI data centers over the next decade to meet the growing demand for data center capacity. JPMorgan Asset Management estimates that by the end of 2024, $163 billion will be invested to expand hyperscale businesses, representing a year-over-year increase of 28%. The report predicts that by 2038, capital expenditures for hyperscale AI will reach $370 billion, up 127% from the estimated AI capital expenditures in 2024.
The current and expected growth of AI and HPC technologies is transforming the data center landscape. As processing demands increase, hyperscale data centers and data centers are gradually evolving from traditional computing facilities into advanced AI infrastructure hubs. These facilities are becoming the foundational infrastructure driving breakthrough technologies such as autonomous vehicles, advanced medical research, and next-generation AI applications. The future of digital innovation will largely depend on the continued development and expansion of these critical computing facilities, marking a new era of technological infrastructure.
Current Data Center Market Overview
The current data center market is composed of numerous public and private enterprises that collectively manage a large number of data centers. Notable companies in this space include Digital Realty, Equinix, Vantage, EdgeConnex, and QTS. According to CBRE, the largest data center region in the U.S. is currently located in Northern Virginia, but growth is very rapid across all regions, leading to historically low vacancy rates.
Data centers are the backbone of multiple industries, supporting everything from streaming services like Netflix to cloud computing, AI, and many other applications. However, not all data centers are the same. Each data center can be customized for specific functionalities and can be categorized into different types, including hyperscale, edge, cloud, and enterprise data centers. Data centers are growing larger and achieving higher power densities. The competition to provide infrastructure for rapidly expanding industries like AI has led to an arms race among hyperscale companies to accelerate the expansion of data center capacity.
Obstacles faced by traditional data centers in meeting AI demands.
Traditional data center providers servicing non-AI industries typically utilize a combination of smaller, geographically dispersed data centers, many of which were initially built for low-density applications. In the past decade, traditional data centers have had relatively low operational energy consumption. Despite Digital Realty (market cap of $62 billion) and Equinix (market cap of $94 billion) being the two largest data center companies globally, they primarily operate smaller-scale data centers. For example, Digital Realty's data centers typically range from 0.5 MW to 40 MW in power per facility. Similarly, Equinix's xScale program consists of a global data center network with a total operational capacity of only 292 MW across 20 facilities (Equinix Q3 2024 Investor Presentation, November 8, 2024). In contrast, some mining operations can secure substantial energy capacity at a single site.
Historically, operators had little incentive to scale up quickly due to limited compute density from streaming services, telecom, data storage, and many cloud applications. However, with advancements in AI and the increasing complexity of these algorithms, data centers now must use the latest generation of GPUs and operate state-of-the-art facilities at scale to optimize training execution.
The increase in scale is driven by advancements in GPU computing capabilities and the advantages of parallel computing, allowing data centers to build larger clusters with greater computational power. Parallel computing enables workloads to be seamlessly distributed across other GPUs, allowing for efficient scaling by adding more units. Crucially, large clusters in a single site reduce latency between GPUs, thereby improving the performance of parallel computing. This advantage makes a single 200MW cluster significantly more efficient in AI training compared to four geographically distributed 50MW clusters, as low-latency communication between GPUs is essential for maximizing computational efficiency. Thus, hyperscale companies prioritize acquiring high-power capacity at a single location to meet the demands of advanced AI workloads.
Currently, this capacity is in short supply, and many traditional facilities struggle to meet the massive energy demands required by modern AI/HPC workloads. Old facilities cannot be easily retrofitted due to differences in networking, cooling, and rack density requirements between low and high compute use cases.
Today, hyperscale companies require data centers with higher energy capacities to support the training of their high-energy models (such as large language models). According to a Uptime Institute article from December 2020, the average rack density that year was 8.4 kW/rack, excluding high-performance outliers of 30+ kW/rack. The power of server racks in these data centers that once peaked at 40 kW per rack now needs to support over 132 kW per rack, which is the power required for cutting-edge systems like NVIDIA's GB200 NVL72, having increased more than threefold in just a few years. Industry experts predict that improvements in computing density and advancements in Moore's Law may push the power demands of server racks to unprecedented levels.
As a result, traditional data center operators have shifted their focus to greenfield development to accommodate the next generation of AI/HPC dedicated data centers, whose energy approvals and construction will take years. According to a recent report from the U.S. Department of Energy, the surge in connection requests for facilities of 300 MW to 1,000 MW or more has stressed the ability of local grids to deliver power at such a rapid pace, resulting in interconnection and construction timelines extended by 2-4 years.
Hyperscale data center operators now aim to build the largest possible GPU clusters to train AI/HPC models, with several companies targeting gigawatt-scale data centers to accommodate hundreds of thousands of next-generation GPUs. While hyperscale data center operators are building their own facilities, they still rely heavily on third-party providers with established power capabilities to accelerate the time to power GPUs. However, only a few existing data centers can handle such massive power demands and high rack energy densities. This shortage is largely due to the rapid growth of demand for data centers being underestimated.
Why Bitcoin miners can fill critical gaps.
Bitcoin miners possess large-scale, power-ready facilities, thus capable of meeting the energy demands of hyperscale miners. For years, miners have sought locations that are energy-rich and reasonably priced, ensuring significant power capacity at a single location, as well as long-term infrastructure projects such as substation components and medium-high voltage equipment. Some mining sites are already equipped with power-ready capabilities, addressing one of the biggest constraints faced by hyperscale miners: securing reliable large-scale power.
By entering these power-ready Bitcoin mining sites, hyperscale miners can bypass the lengthy process of ensuring energy availability and focus on retrofitting and customizing infrastructure to meet their specific needs. Many miners control hundreds of megawatts of sites, a scale few traditional data center operators can achieve at a single location. Several large mining companies have already established access to industrial-scale power infrastructure, securing energy pipelines exceeding 2 gigawatts (GW), enabling miners to benefit from the growing demand for power capacity. Despite significant differences between traditional Bitcoin mining sites and AI data centers, miners have valuable experience in large building and data center management, often possessing established electrical, mechanical, facilities, and security teams. This expertise can further streamline the transition for hyperscale companies seeking rapid expansion.
Only a portion of miners can benefit from AI.
Not all miners can capitalize on AI/HPC opportunities. To build a data center suitable for AI/HPC, several key factors must be met, including securing large-scale land, cooling water, dark fiber, reliable electricity, and skilled labor. Unfortunately, even if these conditions are met, companies that have not obtained the necessary approvals (i.e., power capacity, land, and zoning) or do not possess critical long-term infrastructure components will encounter obstacles and delays in the development process.
Another key reason not all Bitcoin miners can capitalize on AI/HPC opportunities is that existing infrastructure cannot be directly transferred or adapted to AI data centers due to differences in design and operational requirements. While there are some similarities in critical electrical infrastructure (including high-voltage substation components and distribution systems), AI data centers have specific requirements that demand detailed expertise and skilled labor.
The complexity of AI data centers elevates nearly all operational aspects, including mechanical, cooling, and network systems, making the retrofitting of Bitcoin mining facilities into AI/HPC data centers a daunting task. Below, we outline some of the major upgrades miners will need to make to retrofit existing facilities into AI data centers:
1. Network Infrastructure:
AI/HPC workloads require high-speed, low-latency connections between data center GPUs. Therefore, the internal network architecture of AI/HPC workloads is far more complex than that of mining, due to the constant communication between GPUs. The key to successful AI operations is developing an optimal network backbone to ensure rapid execution of workloads. Additionally, connections to dark fiber must be established from the site to meet latency requirements, which mining sites do not require.
2. Cooling Systems:
Miners employ various cooling designs, including air cooling, water cooling, and immersion cooling systems. Cooling is primarily focused on the actual machines themselves, with less emphasis on supporting infrastructure. In contrast, AI data centers will require more advanced cooling solutions, such as liquid cooling directly to chips, to cool the latest generation of power-intensive NVIDIA servers, combined with additional air cooling systems to support networking and mechanical infrastructure.
3. Redundancy:
Compared to Bitcoin mining data centers, AI data centers have stricter redundancy requirements. Mining operations are inherently flexible and thus do not require robust backup power facilities. In contrast, AI data centers typically employ at least N+1 redundancy throughout the entire operation, while more critical components such as core networking and storage components require even higher levels of redundancy to ensure uninterrupted operation or at least to properly cache and check data in the event of equipment failure. This means that for each critical infrastructure component (such as cooling equipment), there must be a backup (N+1 redundancy). For example, while one cooling unit is undergoing maintenance, an additional unit must be available to maintain continuous operation. This level of redundancy is rarely found in mining facilities that do not have such uptime requirements.
4. Redesign of form factors:
AI data centers utilize rack-mounted servers, which differ significantly from the shoebox-style ASICs used in Bitcoin mining. Retrofitting the internal physical infrastructure of facilities to accommodate AI hardware requires a complete redesign to support rack-mounted systems and their specific cooling, networking, and electrical needs.
5. Other Distinctions:
Overall, these factors indicate that retrofitting mining facilities to meet AI/HPC data center requirements is a design and engineering challenge. Enhanced infrastructure requirements have also led to a significant increase in capital expenditure costs for AI/HPC data centers relative to Bitcoin mining construction costs.
Miners capable of leveraging AI data center demand have upside potential.
While miners may possess suitable infrastructure and locations, transitioning to AI/HPC operations requires more than just physical assets—it necessitates expertise, different technology stacks, and new business models. Those with experienced management teams capable of successfully building AI/HPC operations have a tremendous opportunity to bring significant incremental value to their companies. Below are some key advantages that can bring added value to companies choosing to allocate their power and data center resources from Bitcoin mining to AI/HPC:
High cash flow margins and predictability: AI/HPC data center operations, especially in host colocation/custom models, feature long-term contracts that often stipulate fixed and recurring cash flows even before construction of the data center begins. These are predictable and high-margin cash flows, typically partnered with reputable counterparties, allowing data center operators to pass on most costs to tenants, including energy and operating expenses (depending on the leasing structure).
Diversification of cash flow: Revenue is not only more predictable than Bitcoin mining but also independent of the cryptocurrency market, which can balance the income profiles of companies with higher risks in volatile cryptocurrency markets. This can enhance financial stability during Bitcoin bear markets, allowing miners to continue raising cash through equity or debt without excessive dilution or interest burdens.
Deep capital markets can help expand operations: Although infrastructure is significantly more expensive than Bitcoin mining, underwriting investments is more straightforward due to the predictability of cash flow, thereby opening new sources of debt and equity capital for data center projects. Numerous companies, including private equity firms, infrastructure investments, pension funds, and life insurance companies, are eager to enter the data center space to achieve returns. Data center operators with leases signed with reputable counterparties can lease that lease and raise substantial project financing to build data centers.
According to Newmark's (2023 Data Center Market Annual Overview Report), the volume of regular debt financing in 2023 reached an all-time high and shows no signs of slowing, with $18 billion in development financing underwritten in just the first quarter of 2024. Interest rates are also reasonable, with Newmark's rate range approximately between 2.25% - 4.50% over SOFR, depending on the lender.
Tremendous valuation upside potential: Once assets are established and stabilized, there exists a significant valuation disparity between mining and AI/HPC, making AI/HPC a highly attractive opportunity. Historical transaction prices for Bitcoin miners fall within the range of 6-12 times EV/EBITDA, while some of the largest data center operators in the world have valuations of 20-25 times EV/EBITDA. Given the industry's high margins, growth trajectory, predictable cash flows, and reduced market volatility compared to cryptocurrencies, this is reasonable. To further understand the scale of the current disparity, the total EV of hybrid mining/AI companies is 23% of Digital Realty's EV, despite having a total potential MW capacity 3.5 times greater.
Thus, predictable cash flows, an active financing market, and significant valuation upside potential make AI/HPC opportunities extremely attractive and value-generating for miners with the right assets. These miners are expected to make significant strides in the traditional data center market and become one of the largest operators in the industry.
Outlook for Bitcoin mining
In recent months, AI/HPC has garnered significant attention, but we still expect the hash rate and growth of Bitcoin mining networks to continue to rise. The growth of mining will be synchronized with the growth of AI/HPC. Rising Bitcoin prices enhance miners' profitability, and if prices continue to rise and exceed network difficulty growth, mining may become more lucrative. But what will the future mining landscape look like with the rise of Bitcoin and AI/HPC? Below we outline some major trends that may emerge at the intersection of AI/HPC and Bitcoin mining in the foreseeable future:
Maximizing the value of electronics:
Most Bitcoin miners consistently prioritize maximizing the value of their energy usage. Currently, for those with adaptable sites, AI data centers represent the most profitable avenue. Considering the value growth of AI/HPC sites, mining sites capable of converting to AI/HPC data centers are likely to follow this path to maximize shareholder value. However, this does not necessarily mean a disadvantage for Bitcoin miners. We still expect the network hash rate to grow, but at a slower pace than if no major U.S. miners were converting sites to AI/HPC data centers. These conversions benefit the remaining miners on the network by eliminating competitive hash rates.
Bitcoin mining is the driving force behind monetizing idle power:
As AI/HPC becomes increasingly prominent, we expect miners to further focus on deploying their capacity in more remote areas, as hyperscale companies have large sites available for AI/HPC in more developed markets, thus bidding higher than miners. Bitcoin mining is unlicensed, location-agnostic, and flexible, making it one of the best ways to monetize idle power capacity.
We anticipate that a larger portion of Bitcoin mining will be pushed to the margins to monetize idle power capacity—especially in remote areas of the U.S. and in emerging international markets like Ethiopia and Paraguay, which have abundant cheap surplus energy.
Bitcoin mining as a strategic bridge for infrastructure investment and AI/HPC optionality.
Additionally, as different regions in the U.S. strive to build transmission infrastructure and fiber connectivity, Bitcoin mining can serve as a bridge to underwrite larger capacity energy infrastructure projects, such as substations and power plant construction, even in the absence of immediate or explicit opportunities to leverage AI/HPC capacity. By employing Bitcoin mining for opportunistic real estate and power-related investments, investors can gain returns while waiting for other long-term energy use cases to materialize, positioning it as an attractive strategy for infrastructure growth and investment.
For miners that cannot be converted into AI/HPC data centers, Bitcoin mining sites can still operate as long-term profitable businesses. Several miners have purchased large-load facilities without existing AI/HPC tenants and have also invested in sites at various stages of development. As previously outlined, some of these sites may lack the optimal characteristics required for AI/HPC, but remain useful for Bitcoin mining. Other miners lack the teams or in-house expertise to contract with major acquirers and undertake challenging engineering and large construction projects. The hope of miners seeking to maximize value is to lock in an AI customer, but in cases where AI/HPC opportunities cannot be realized, these miners can still choose to establish a profitable BTC mining business.
Emerging synergies between AI/HPC data centers and mining.
ASIC manufacturers like Bitmain have begun developing ASICs with form factors similar to GPUs for data center racks. Further coordination of ASIC form factors with next-generation GPU sizes will allow data centers to monetize their underutilized server rack space by installing server-sized miners in idle rack spaces, which will help streamline the process of retrofitting data centers for AI/HPC. Looking forward, miners may be more inclined to purchase these machines as they maintain flexibility in data center design and could help miners pivot to AI/HPC more easily if higher-value opportunities arise.
As the capacity of AI/HPC data centers grows, their impact on the grid is also increasing. While these data centers must remain online almost continuously, this does not necessarily mean that total energy consumption is constant. In fact, the load curve for AI/HPC training may be very unstable, as more power is consumed during intensive computation execution and less power during checkpoint periods. The frequency of checkpoints varies, and depending on the deployed infrastructure and the size of the models, this process may take anywhere from a few minutes to several tens of minutes. As the model size scales up, more data needs to be stored, increasing the time needed to retain all data.
Similarly, for AI/HPC inference workloads, load profiles are expected to closely align with customer demand, as each model query is processed directly in the data center. Initially, these profiles may exhibit significant volatility as model demand fluctuates. However, over time, as specific models gain widespread adoption, loads may become more predictable, peaking during the day and declining at night. This daily load cycle offers an ideal opportunity for Bitcoin mining, as mining operations can dynamically scale up or down to complement the fluctuating energy demands of AI inference processes.
Therefore, in the future, Bitcoin mining can serve as a load balancing mechanism, increasing during low loads and decreasing when AI loads recover. Tenants may also not need to utilize the full GPU capacity, allowing miners to accelerate.
The benefits for data center operators are clear, as they can derive more value from available online capacity, while for tenants, this provides a degree of load stability for data centers and the entire grid. As data center clusters grow in size, power consumption and their impact on the grid will face increasingly stringent scrutiny, making load stability essential.
Shifting MW to AI/HPC should slow the growth rate of hash rates.
Miners transitioning into AI/HPC operations are actively reallocating capacity that could have been used for Bitcoin mining, which should slow the growth rate of the network hash rate. This is especially significant considering the potential bull market for Bitcoin, as rising Bitcoin prices will not correlate with an equal and offsetting growth in network hash rates, thereby driving up hash rates. That said, we still expect network hash rates to rise as more efficient mining machines come online, whether to replace older generation machines or to make new net investments at sites that are not conducive to AI/HPC business.
Conclusion
Demand for data centers in the U.S. may surge at an unprecedented rate, with a projected year-over-year increase of 31% in 2024 alone. These forecasts also suggest that the capacity of U.S. data centers will more than double in the next five years, jumping from the current 21 GW data center capacity to an estimated 45 GW. This explosive growth, combined with the hundreds of billions of dollars in investments that hyperscale providers have committed over the next 5-10 years, creates highly attractive opportunities for companies that can provide two critical resources: abundant cheap energy and robust infrastructure capable of supporting AI and HPC operations.
The current AI and HPC boom has exposed a key weakness of traditional data centers: their inability to retrofit existing facilities to meet the significant power demands of modern AI workloads. This gap in the market creates significant opportunities for Bitcoin mining operations, which already possess what AI/HPC companies urgently need: large sites with expedited power connectivity plans. The choices for hyperscale providers are limited as they struggle to rapidly scale their operations to meet the explosive demand of AI/HPC. Bitcoin miners are becoming a feasible option for hyperscale businesses to expand and remain competitive in an increasingly growing market. However, this generation of opportunity for Bitcoin miners remains selective. Only a small fraction of Bitcoin mining operations possess the necessary infrastructure and capabilities to successfully support the rigorous demands of modern AI/HPC workloads. Those who have these scarce assets and seek to maximize their value will pivot to AI/HPC data centers.
While some critics argue that Bitcoin miners diversifying into AI/HPC services may weaken network security by reducing the computational power dedicated to mining blocks, this shift may actually benefit the broader mining ecosystem. Miners unable to meet AI/HPC site demands can achieve higher profitability from rising hash prices. As more miners go offline, Bitcoin prices rise, and the increase in hash prices will significantly boost the profitability of all Bitcoin miners. With Bitcoin prices having increased by as much as 143% this year, and a pro-Bitcoin president taking office in the White House, Bitcoin mining in the U.S. is poised to enter its strongest period yet.
The intersection of cryptocurrency and artificial intelligence can be considered one of the hottest cryptocurrency sectors in 2024. As of December 2024, the total market capitalization of cryptocurrency projects building AI applications using liquid tokens is approximately $33 billion. Furthermore, Galaxy Research estimates that over $382 million in venture capital will be allocated to early-stage crypto AI startups in 2024. Although most crypto AI projects lack product-market fit, the overlap between Bitcoin mining and the growth of AI/HPC businesses is evident. Bitcoin mining's entry into the AI space stands out compared to other overlapping areas in both fields because it has the potential to supply the most critical component for AI/HPC businesses at scale—energy. Therefore, Bitcoin miners holding convertible assets for AI/HPC may be one of the only pure and scalable cryptocurrency x AI investments in the industry today.