Dan Loeb's Third Point LLC disclosed positions in at least three former cryptocurrency mining operations now repositioning their infrastructure for artificial intelligence compute, according to the fund's latest 13F filing. The aggregate disclosed stake totals approximately $47 million across companies that spent 2021-2022 building data centers for proof-of-work blockchain validation and are now marketing those facilities to hyperscalers and AI labs.
The timing follows a six-quarter collapse in Bitcoin mining economics. Network hashrate difficulty increased 340% between Q1 2021 and Q3 2023 while Bitcoin traded sideways, compressing margins for all but the lowest-cost operators. Companies that raised capital at $60,000 Bitcoin now sit on stranded assets: purpose-built facilities with cheap power Purchase agreements, existing cooling infrastructure, and GPU clusters originally spec'd for SHA-256 hashing. Third Point's entry suggests Loeb sees the pivot from consensus algorithms to transformer inference as more than narrative—these are live conversations with Azure, AWS, and private AI labs hunting for non-hyperscaler capacity.
The second-order effect matters more than the headline position size. Third Point runs $15 billion in AUM and maintains a reputation for operational activism, not passive position-taking. When Loeb discloses a sub-$50 million stake in micro-cap infrastructure plays, the signal is exploratory positioning ahead of a larger thematic build. His 2012 Yahoo campaign started with a $1 billion stake that grew to board seats and strategic reorientation. The Bitcoin-miner pivot is a different scale, but the pattern recognition is identical: stranded assets, mispriced optionality, a sector transition the market hasn't yet modeled.
What Third Point likely sees: data centers already permitted, power contracts signed at pre-inflation rates, and management teams desperate to avoid equity dilution by monetizing existing infrastructure. AI inference doesn't require the same chip density as training, but it does require low-latency regional compute and cheap electrons. A former mining facility in upstate New York or rural Texas with a 15-megawatt power allocation and fiber connectivity can host inference endpoints for multimodal search, video generation, or edge AI without the $800 million capital expenditure of a greenfield hyperscale build. The companies trading at 0.6x book value suddenly have a call option on $4-per-watt data center valuations if they can sign a multi-year capacity contract with a Tier 1 buyer.
Allocators should watch for follow-on moves in two directions. First, whether Third Point increases its stakes above the 5% threshold that would trigger activist disclosure and potentially board representation—a signal the fund believes it can drive operational improvements or M&A. Second, whether other multi-strategy funds or infrastructure specialists follow into the same names, which would confirm the thesis is moving from exploratory to consensus. The next catalyst is likely Q1 earnings calls in April, when these companies will either announce signed AI compute contracts or admit the pivot is still speculative. Third Point's filing covers positions as of December 31, meaning Loeb has had 90 days to build or trim since then.
The forward fact that matters: if even one of these former miners signs a $20 million annual run-rate contract with a hyperscaler or AI lab, the entire sub-sector re-rates, and Third Point's $47 million exploratory stake becomes the anchor for a $200 million thematic build.