Bitcoin Miners Shift Focus to AI Power Contracts Amid Demand Surge

Bitcoin miners are increasingly pivoting towards artificial intelligence (AI) power infrastructure, capitalizing on a surge in demand for electricity from AI companies. Public Bitcoin miners have collectively announced contracts for AI and high-performance computing (HPC) worth an estimated $70 billion to $100 billion, effectively transforming their roles from traditional miners to power landlords.

The economic advantages of this shift are compelling. AI operations can generate approximately $1.5 million per megawatt annually, compared to Bitcoin mining's $500,000 per megawatt. This disparity has been exacerbated by the upcoming Bitcoin halving, which has tightened profit margins for miners, prompting them to seek lucrative tenants.

Major contracts include Core Scientific's $10.2 billion deal with CoreWeave for 590 MW of capacity, TeraWulf's $12.8 billion in contracted HPC revenue, and IREN's $9.7 billion agreement with Microsoft for GPU cloud services. Hut 8 has also secured multiple long-term leases valued between $7 billion and $9.8 billion.

One of the key advantages Bitcoin miners have is their established infrastructure. They have spent years securing power purchase agreements and building connections to the electrical grid, which AI companies require urgently. This infrastructure is now being repurposed to meet the power demands of AI workloads.

However, there is a significant caveat: only about 550 MW of the 4 GW of contracted AI capacity is currently generating revenue, indicating a utilization rate of roughly 14%. While the potential revenue from these contracts could reach $1.1 billion to $1.5 billion annually, this is only a fraction of what full deployment could yield.

As a result, Wall Street is valuing AI-focused miners at significantly higher multiples compared to traditional Bitcoin miners. This pivot is also expected to impact Bitcoin's hashrate, with an estimated 35 exahashes per second of mining capacity shifting towards AI tenants.

New Insights on AI Data Centers and Energy Demand

• Goldman Sachs warns that next-gen AI servers, expected to deploy around 2027, could consume over 500 to 600 kilowatts each, significantly increasing energy demand.
• Projected US data center power demand could reach approximately 108 GW by 2030, with 72% of capacity concentrated in just 1% of counties.
• Reserve margins across most US grids are expected to dip below 15%, a threshold that indicates potential rolling outages during peak demand.
• Infrastructure delays in connecting new power generation to the grid can exceed seven years due to interconnection queue backlogs and equipment shortages, particularly for transformers.
• As a result, one-third of future AI data center capacity may operate as independent power systems, bypassing the public grid entirely.
• Natural gas peaker plants and renewable energy sources are likely to benefit from increased investment to meet the rising demand for energy from AI data centers.

New Insights on AI Bubble Risks

MIT Technology Review has highlighted the potential consequences of the AI bubble bursting, suggesting that hyperscalers may need to nearly triple their productivity by 2030 to recover their substantial investments in infrastructure, estimated at nearly $1.1 trillion through 2027.

Wharton finance professor Jessica Wachter, who previously served as chief economist at the SEC, indicates that without a significant increase in productivity, the current investments could represent the largest misallocation of capital in history.

Alphabet recently reported a $5.9 billion free cash flow deficit, marking its first such deficit since going public in 2004, raising concerns among investors about the risks associated with AI spending.

Additionally, Morgan Stanley projects that hyperscalers will finance over half of their anticipated $2.9 trillion data center spending through 2028, increasingly relying on external capital rather than cash reserves.

Crypto strategist Arthur Hayes has suggested that a potential AI credit bust could lead the Federal Reserve to print more money, which might push Bitcoin (BTC) towards $1 million, highlighting the interconnectedness of capital markets and economic stability.