Bitcoin fell 12% this year. Its miners' stocks rose 70%. Here's why

In 2026, something happened that would have seemed strange just three years ago: bitcoin's own price fell roughly 12%, while shares of the public companies that mine it jumped 70% or more. The explanation is simple — bitcoin mining has quietly, and rapidly, turned into infrastructure for artificial intelligence.

Who Signed What, and How Much

TeraWulf (WULF) — the fastest-growing public miner this year, up 73.58% year-to-date — has locked in more than $12.8 billion in contracted high-performance computing (HPC) revenue through deals with Google-backed Fluidstack and Core42 across sites totaling more than 1 GW of power. The company separately signed a 20-year lease with Anthropic for a Kentucky data center supporting 401 megawatts of critical IT load, expected online by early 2028.

Hut 8 followed a similar path: the company anchored a $7 billion, 15-year lease at its River Bend campus with Anthropic and Fluidstack as counterparties, while building an 8.5 GW development pipeline across various stages, from due diligence to active construction. Separately, it signed another 15-year, $9.8 billion lease for the second phase of its Beacon Point campus in Texas, after which its stock jumped 17%.

IREN signed a $9.7 billion deal with Microsoft for 76,000 Nvidia GPUs, plus a separate $2.8 billion multiyear cloud computing services contract with AI developers. Cipher Mining, Core Scientific, and Applied Digital are showing similar transition patterns.

Why Miners Specifically Were Ready for This

The economics of the pivot look almost indecently favorable: by some estimates, AI compute generates the equivalent of roughly $25 per kilowatt-hour, versus about $1 per kilowatt-hour for bitcoin mining. But the key factor isn't just the margin gap — it's that for AI data centers, the binding constraint has become available power, not chips. That's exactly where miners hold a decisive edge: they spent years locking in large power contracts, securing land, building cooling systems, and clearing regulatory approvals — meaning they already hold assets that normally take years to build from scratch.

Converting an existing mining site for AI use costs roughly $700,000-$1 million per megawatt, versus $8-15 million per megawatt to build a full AI-grade data center from the ground up. An added benefit is operational flexibility: bitcoin mining can act as a "switchable" load when AI compute demand temporarily dips or mining becomes more profitable due to power prices — a flexibility pure-play AI data center operators simply don't have.

How Much This Has Already Changed Miners' Businesses

In the first quarter of 2026, TeraWulf generated $21 million in high-performance computing revenue out of $34 million in total revenue — meaning its AI business had already become a larger, more stable part of the company than bitcoin mining itself. Industry analysts project that by the end of 2026, public miners could derive as much as 70% of their revenue from AI infrastructure, up from roughly 30% today. Across the industry, public miners are now pursuing more than $70 billion in AI infrastructure contracts combined.

What This Means for the Industry

The story of bitcoin miners turning into AI compute infrastructure providers is a rare example of a single skill set — running energy-intensive facilities at industrial scale — becoming valuable across two completely different industries at once. For investors, that means public miners' valuations are increasingly driven not by bitcoin's price, but by the growth of their AI infrastructure business. And for the AI industry as a whole, it means that a shortage of available power — not just a shortage of chips — has become one of the main factors determining who can scale compute quickly.

This material is for informational purposes only and is not investment advice.