Bernstein is pushing back on the idea that bitcoin miners can broadly replicate Core Scientific's CoreWeave-driven 75% return on AI compute contracts. A new returns analysis puts TeraWulf and Cipher Mining at just 5% and 4% stabilized return on assets, a fraction of the headline figure that sent CORZ soaring last year.
Why it matters
The 75% number set off a stampede of bitcoin mining names chasing hyperscaler and AI compute pivots. Bernstein's note argues the market is conflating one outlier deal with a replicable playbook. Cipher and TeraWulf, the two next-most-cited candidates, post returns closer to a typical industrial power-and-heat business than to Core Scientific's converted data-centre economics.
Market impact
Miners have collectively contracted roughly 7 GW of power capacity to hyperscalers, neoclouds and chipmakers across 19 deals worth more than $135 billion over the past two years, per Bernstein's tally. The scale is real, but the economics look uneven. Names priced for Core Scientific-style AI optionality without Core Scientific-style infrastructure are the ones at risk of a sharp re-rating if the sector narrative cools.
Frequently asked questions
-
What return does Core Scientific post on its CoreWeave AI deal?
Bernstein puts Core Scientific's stabilized return on its CoreWeave AI compute contract at roughly 75%, the figure that originally set off a sector-wide AI-pivot stampede.
-
How do other bitcoin miners compare on AI compute returns?
Bernstein's analysis puts TeraWulf at about 5% stabilized ROA and Cipher Mining at 4%, well below Core Scientific's 75% and closer to a standard industrial power business.
-
How much power have bitcoin miners contracted to AI and hyperscaler buyers?
Bernstein tallies 19 deals over the past two years covering roughly 7 GW of capacity and worth more than $135 billion in total contract value across miners, hyperscalers, neoclouds and chipmakers.
-
Why is Bernstein pushing back on the AI-pivot narrative?
The firm argues the 75% Core Scientific return is an outlier deal, not a replicable playbook, and that most miners lack the data-centre-grade infrastructure needed to capture similar economics.
-
Which miners are most exposed if the AI narrative cools?
Bernstein flags names priced for Core Scientific-style AI optionality without equivalent infrastructure as the most exposed to a sharp re-rating if the sector narrative loses momentum.
TheBlock