Bitcoin Miners’ $70bn AI Pivot Divides Sector Over Long-Term Lease Risk

The Bitcoin mining industry’s rapid pivot towards artificial intelligence infrastructure has entered a period of open disagreement, with public miners having signed contracts worth more than $70 billion to host AI and high-performance computing workloads even as a contrarian faction warns the underlying economics may not justify the shift. Analysts now expect the sector to derive roughly 70% of its revenue from AI-related activity by the end of 2026, up from around 30% earlier in the year, according to Crypto Briefing.

For an industry historically valued as a leveraged proxy for the Bitcoin price, the scale of the reallocation marks a structural break — one that carries implications for how investors, lenders and regulators assess miner balance sheets going forward.

Pre-secured power draws hyperscalers

The appeal of miners to AI operators rests on a simple asset: pre-secured electricity capacity, which is increasingly scarce as data centre demand accelerates. Industry estimates cited by Crypto Briefing suggest miners can deploy AI-ready facilities up to 75% faster than entirely new builds, a timing advantage that has translated into some of the largest infrastructure deals in the sector’s history.

TeraWulf has secured a 20-year lease with Anthropic covering approximately 401 megawatts of capacity, due to come online in 2027. Cipher Mining has signed a 15-year agreement with AWS valued at $5.5 billion. Jefferies has responded by initiating Buy ratings on several miners tied to the AI transition, including Cipher Mining, Hut 8, TeraWulf and Core Scientific.

The urgency behind these deals is compounded by weak conditions in mining’s core business. Hash price — the revenue a miner earns per unit of computational power — sits near cyclical lows, with the post-halving margin squeeze continuing to erode profitability and pushing operators towards alternative revenue streams as a matter of survival rather than opportunism.

The case against conversion

Not every operator is following the trend. A cohort of miners has pushed back on the wholesale conversion of facilities, arguing that Bitcoin mining and AI workloads have fundamentally different infrastructure requirements that the market may be underpricing.

Mining rigs can tolerate remote locations and interruptible power supply, whereas AI inference and training clusters require stable, high-density power delivery and cooling systems that most existing mining sites were never designed to provide. Converting a site typically means committing to a long-term lease with a single hyperscaler or AI laboratory — a concentration of counterparty risk that leaves miners exposed should that tenant renegotiate terms, build its own infrastructure, or exit the market entirely.

A bifurcating industry

The result is a mining sector splitting into two distinct investment cases. Companies such as Core Scientific, Cipher Mining and TeraWulf are effectively becoming data centre operators that happen to also mine Bitcoin, drawing in traditional technology investors and seeing their valuations re-rate accordingly.

Pure-play miners without credible AI pipelines, by contrast, continue to trade largely as leveraged bets on Bitcoin’s spot price, rising and falling with the coin’s movements rather than with any diversified revenue thesis.

Two forthcoming milestones are likely to test which model proves durable. TeraWulf’s Anthropic facility is expected to begin operations in 2027, offering the first substantive evidence of whether converted mining sites can meet the uptime and performance standards AI customers demand. Cipher Mining’s $5.5 billion agreement with AWS represents a similar bellwether: smooth execution would lend credibility to the broader thesis that miners can operate as genuine AI infrastructure providers, while any stumble could reinforce the contrarian case for staying focused on core mining operations.

Read more: $326m Leverage Flush Renews Questions Over Crypto Derivatives Risk Controls