Bitcoin miners diversify into AI to sustain profitability

As the cryptocurrency market evolves, digital asset mining companies—specifically, bitcoin miners—are adapting and diversifying into artificial intelligence to ensure long-term profitability. This strategic shift—which involves leveraging existing infrastructure as well as investing in versatile hardware such as graphics processing units (GPUs)—not only mitigates the risks associated with relying solely on cryptocurrency mining but also opens new avenues for growth and innovation.

Through partnerships with AI firms and independent exploration of AI applications, digital asset mining companies are transforming their operations. This is especially important following the April 2024 bitcoin halving event, which reduced the block rewards for mining companies.

This strategic pivot is not without challenges, particularly the limitations of application-specific integrated circuits (ASICs), which are traditionally used in bitcoin mining and not easily repurposed for AI modeling.

ASICs are designed with a singular purpose: to execute the hashing algorithms necessary for cryptocurrency mining, such as the SHA-256 algorithm used in bitcoin mining. This specialization makes ASICs exceptionally efficient for their intended task but significantly limits their flexibility. AI modeling requires a diverse set of operations, including matrix multiplications, convolutions and activation functions, which ASICs are not equipped to handle. Consequently, repurposing ASICs for AI applications is impractical, necessitating alternative strategies for mining companies looking to enter the AI space.

Leveraging existing infrastructure

Despite the limitations of ASICs, large-scale mining companies possess valuable infrastructure assets that can be repurposed to support AI ventures. These assets include extensive data centers, access to low-cost power and sophisticated cooling systems. Mining companies also have staff with experience in optimizing hardware and data center layouts to ensure maximum efficiency.

By repurposing these facilities and employees, mining companies can pivot to AI operations with minor modifications to their existing business operations. This infrastructure, originally designed to support the high power and cooling demands of cryptocurrency mining, is well suited to the similarly intensive requirements of AI workloads.

Mining companies are stepping up at a time when many traditional data center operators are failing to meet the surging demand for AI computing capacity due to power shortages, long lead times to bring new capacity online, and the extensive upgrades required for existing data centers to fulfill these needs.

Strategic partnerships and AI exploration

To capitalize on their existing infrastructure, many digital asset miners are forming strategic partnerships with AI firms. These collaborations enable mining companies to use their data centers for AI training and inference tasks, providing a new revenue stream.

One example is bitcoin mining company Hut 8 Corp., which received a $150 million investment in mid-2024 to build AI infrastructure. Another example is Iris Energy, where AI and bitcoin mining technologies share the same walls, exemplifying the seamless integration of AI infrastructure within existing mining operations.

By hosting AI workloads, mining companies can leverage their existing investments in infrastructure without the need to repurpose ASIC hardware directly. When a mining company hosts an AI developer, the mining company’s job stops after providing the infrastructure. From there the AI developer will provide staff with the technical expertise needed to pilot the software programs that will handle tasks such as AI modeling.

The concept of hosting is not new to mining companies—many operations offer this service to individuals and competing miners alike to allow more people to own and operate miners without the need for the supporting infrastructure in their homes. The difference is in the form of hardware needed to handle the AI workloads. Instead of purchasing additional ASIC machines traditionally used for mining, companies are opting to purchase workstation GPUs that support the latest generation of high-density computing demand. In contrast to powering ASIC miners, this approach relieves rather than increases the mining company’s power demands.

While it is difficult to compare the revenue generation from an ASIC that is mining digital assets to a GPU that is used for AI computing, companies can repurpose GPUs and use them for mining various proof-of-work cryptocurrencies; ASICs, on the other hand, can only mine coins and must utilize the ASIC’s native algorithm.

In addition to partnering with AI firms, some mining companies are independently exploring AI applications. This involves investing in research and development to understand how AI can enhance their operations, from optimizing mining processes to exploring entirely new business models centered on AI. This also requires the mining companies to employ staff members with the technical expertise needed to execute AI improvements, which differs from the expertise required to build and operate ASIC miners.

It’s getting hot in here

Because heat generation correlates directly with power usage, ASICs, with their higher power consumption, naturally generate more heat compared to GPUs, and their cooling solutions vary accordingly. GPUs can often operate safely using their stock cooling components. However, ASICs, especially when used in data centers surrounded by other ASICs, require additional or auxiliary cooling systems to maintain safe operating temperatures.

The need for enhanced cooling solutions for ASICs also means they typically have greater maintenance needs. Data centers using ASICs for digital asset mining often implement more sophisticated cooling methods, such as liquid or immersion cooling, due to the significant heat generated by miners. In contrast, data centers using GPUs can generally rely on air-cooling solutions to manage temperatures effectively.

Aside from the cooling aspect, the maintenance of GPUs and ASICs is relatively similar. Both types of hardware need regular cleaning to ensure optimal operation and keep connections secure and free of debris. If either type of hardware begins to malfunction or experiences reduced performance for reasons beyond cooling or driver updates, component-level repairs are often necessary. For ASICs, this could involve the hash board or power supply, while for GPUs, it might relate to the onboard GPU chip, memory, or power circuitry such as voltage regulators or capacitors.

Managing risks

Entering the AI sector introduces several risks for digital asset mining companies. Here’s a high-level look at some of those risks and ways to mitigate them:

Planning will be paramount for digital asset mining companies eyeing a move into the AI space. Leadership teams and other decision makers will need to balance potential opportunity with these risks in order to diversify effectively.