Bitcoin Mining Powering the AI Data Center Revolution: Q&A with Capital Innovations’ Michael Underhill 

The demand for high-capacity computing (HPC) facilities has surged due to the rapid adoption of artificial intelligence (AI) technologies, which require immense processing power for training and deploying large-scale models. This trend is leading hyperscalers (e.g., Amazon, Google, Microsoft) to invest heavily in new data center capacity. 

Many existing data centers, however, lack the power capacity to support the energy-intensive workloads of AI and HPC systems. These workloads often demand high-density server configurations that traditional facilities were not designed to accommodate. 

Bitcoin miners are uniquely positioned to seize opportunities in the HPC market, especially in sectors like AI and other power-intensive computational workloads. Their operational models align with many of the infrastructure and energy needs of HPC facilities. 

To address this growing demand, AI companies are actively on the hunt for affordable power sources and large expanses of land for data centers. Having sufficient power available now is vital when considering it takes years to build high-performance computing (HPC) data centers from scratch, with current wait times for electrical-grid connections extending up to six years. 

Michael Underhill, CIO of Capital Innovations, shared his views with Connect Money about the burgeoning opportunity in AI data centers, why Bitcoin miners are gravitating to AI and what they’re doing to adapt their existing infrastructure to serve the data storage and computing needs, and any obstacles facing data centers in meeting demand, among other topics. 

CM: What is the AI data center opportunity? 

MU: Given the current supply / demand imbalance with AI workloads and constraints on the power grid, data center development yields are at historically high levels (and underpinned by long-duration leases). We believe a highly focused approach to data center portfolio investment will position investors well to take advantage of the current opportunity set. 

CM: What hurdles, if any, do traditional data centers face in meeting the demand from AI? 

MU: According to a recent DOE report, data centers consumed about 4.4% of total U.S. electricity in 2023 and are expected to consume approximately 6.7% to 12% of total U.S. electricity by 2028. The report indicates that total data center electricity usage climbed from 58 TWh in 2014 to 176 TWh in 2023 and estimates an increase between 325 to 580 TWh by 2028.   

Data center load growth has tripled over the past decade and is projected to double or triple by 2028. U.S. electricity demand is projected to account for data center expansion and the rise of AI applications, domestic manufacturing growth, and electrification of different industries. 

CM: Why are Bitcoin miners pivoting to AI? 

MU: There is increasing interest among cryptocurrency mining facilities to actually switch over to doing AI because of their service level agreements. Their willingness to do one and not the other depends largely on the price of Bitcoin or whatever other cryptocurrency they’re mining.  

A data center’s stated purpose as a crypto-mining operation had very little bearing on whether it would remain as one in the future. Reuters reported in August that technology companies are seeking the energy assets held by crypto miners as they race to secure electricity supply for AI and cloud data centers, estimating that about 20% of cryptocurrency could pivot to AI by the end of 2027. 

CM: Can Bitcoin mining solve AI’s looming energy crisis? 

MU: Bitcoin mining and AI both have significant energy demands, but they operate in different realms. Bitcoin mining is known for its high energy consumption due to the computational power required to solve complex mathematical problems. This process is energy-intensive and has raised concerns about its environmental impact. 

On the other hand, AI’s energy consumption is driven by the need to train and run large models, which also requires substantial computational resources. The energy crisis in AI is more about finding sustainable ways to power data centers and improve the efficiency of AI algorithms. 

While Bitcoin mining itself is not a solution to AI’s energy crisis, there are some overlapping areas where advancements in energy efficiency and renewable energy sources could benefit both fields. For example, using renewable energy to power data centers and mining operations can help reduce the overall carbon footprint. 

In summary, Bitcoin mining and AI face similar challenges regarding energy consumption but solving AI’s energy crisis will likely require a combination of improved hardware, more efficient algorithms, and a shift towards renewable energy sources. 

CM: What are miners doing to adapt their existing infrastructure to serve the data storage and computing needs of the AI sector. 

MU: Bitcoin miners are pivoting to AI data center operations to leverage their existing infrastructure and expertise. Here are some key adaptations they’re making: 

Leveraging Advanced Equipment: Bitcoin miners have already invested in high-capacity GPU infrastructure, which is also suitable for AI applications. 

Securing Low-Cost Energy: Miners are experts in finding and utilizing cheap energy sources, which is crucial for the energy-intensive demands of AI. 

Building Data Centers: Many miners have built large-scale data centers in remote locations with access to renewable energy, making them ideal for AI operations. 

Cooling Solutions: Advanced cooling techniques, such as direct-to-chip cooling, are being used to handle the high-power density required for AI. 

Diversifying Revenue Streams: By transitioning to AI data centers, miners can achieve more stable and predictable revenue streams compared to the volatile cryptocurrency market 

CM: As we try to move to net-zero, or something resembling that, does the electricity demand from AI pose an obstacle? 

MU: AI’s electricity demand does pose a challenge to achieving net-zero goals. The global electricity demand of AI-relevant data centers is expected to double by 2029. While many data centers are sourcing energy from renewables, the need for continuous power means it will be difficult to phase out fossil fuel-based generation completely. 

However, AI also has the potential to enhance energy efficiency and optimize electricity infrastructure. By leveraging AI to improve energy use and integrate renewable sources, we can mitigate some of the increased demand. The key will be balancing AI’s energy consumption with its ability to drive efficiency and support the transition to a more sustainable energy system. 

It’s a complex issue, but with careful management and innovation, AI can be part of the solution rather than just an obstacle.