AI Firms Turn Excess Data Center Power Into Bitcoin Mining Revenue

AI companies are increasingly leveraging excess data center power to mine Bitcoin, a strategic move that analysts believe could stabilize electricity grids and enhance energy efficiency. This convergence of AI and Bitcoin mining is driven by the need to monetize unused energy and optimize power procurement strategies.

Mara Holdings and Riot Platforms are among the first companies to adopt this approach. In June, Mara Holdings launched a project that utilizes idle power from an AI data center to mine Bitcoin. Similarly, Riot Platforms invested $1 billion in energy infrastructure, which will support both Bitcoin mining and AI operations.

Daniel Batten, a climate tech investor and prominent analyst on Bitcoin’s environmental footprint, described Mara’s initiative as "hugely significant." He noted that while it was previously Bitcoin mining companies diversifying into AI compute services, the trend is now reversing with AI firms becoming Bitcoin miners.

AI operations demand substantial computing power and energy. For instance, training the GPT-3 model required nearly 1,300 megawatt hours (MWh) of electricity, equivalent to the annual usage of about 130 households in the U.S. Training more advanced models like GPT-4 is estimated to consume 50 times more energy. Much of this energy is used by data centers, which often over-purchase power to avoid downtimes.

The AI/Bitcoin mining nexus addresses long-term inefficiencies in data center power dynamics by turning excess energy into a new source of value. For AI companies, mining Bitcoin during off-peak hours could reduce operating costs. Mara Holdings, for example, uses AI infrastructure to mine Bitcoin, cutting waste and generating new revenue while creating a responsive energy consumer for the grid.

Fred Thiel, CEO of Mara, emphasized that meeting the demands of today’s compute infrastructure is not just about adding more energy but making better use of the power available. He noted that Mara’s flexible data centers tap into unused or underutilized energy resources to secure the world’s preeminent blockchain ledger, converting clean energy into economic value.

Brian Morgenstern, head of public policy at Riot, echoed similar sentiments, stating that Bitcoin mining can be a first mover on sites suitable for AI data centers to build power infrastructure and monetize capacity. Riot Platforms has invested in a 400MW power plant in Corsican, Texas, using 40% of its capacity to mine Bitcoin.

The implications of this convergence extend beyond corporate profits. AI data centers create erratic power demand, which is challenging for grid operators, especially with the increasing use of intermittent renewable energy sources like solar and wind. Bitcoin mining, however, can stabilize these unexpected spikes in AI power demand because it can be dialed up or down in seconds, responding to real-time grid needs.

Batten highlighted that Bitcoin mining can make the grid operator’s job of stabilizing the grid much easier, allowing for more variable renewable energy to be integrated without risking blackouts or brownouts. He described Bitcoin mining as the most modular, scalable, and interruptible load-balancing technology available today.

Morgenstern further explained that Bitcoin’s decentralization allows mining facilities to curtail power use when the grid needs them to, providing grid stability services. This convergence is also critical for cutting carbon emissions, as better grid stability and less energy waste make renewable energy integration more feasible.

The scalability of the Mara model is a significant question. Batten believes that any AI company forward-purchasing large amounts of energy, from mid-sized to hyperscalers, can benefit from this model. This strategic pivot by AI firms to Bitcoin mining with data center power not only addresses energy inefficiencies but also opens new avenues for economic value and grid stability.