MARA Holdings and CleanSpark both reported double-digit revenue declines this earnings season, but Bernstein responded to the two bitcoin miners' results with sharply different ratings — keeping CleanSpark at Outperform with a $24 price target while holding MARA at Market-Perform with a $17 target, a divergence the firm attributed directly to execution on AI infrastructure rather than underlying mining performance.
MARA posted second-quarter revenue of $174.9 million, down 27% from $238.5 million a year earlier, with a net loss of $611.3 million driven largely by a $343 million fair-value writedown on its bitcoin holdings — even as the company mined 3% more bitcoin than a year earlier and grew its energized mining capacity 22% to 70.3 exahash per second. CleanSpark's fiscal third-quarter revenue fell 30.5% to $138 million, with a $239.8 million net loss. The split in Bernstein's ratings comes down to a single fact: CleanSpark has already signed a $6.6 billion, 20-year AI hosting lease at its Sandersville facility, while MARA has yet to land its first commercial AI contract despite pursuing the same pivot.
Research from Blocksbridge Consulting, examining 25 AI and HPC transactions announced by bitcoin miners between June 2024 and August 2026, found that average same-day stock price gains on announcement have fallen from roughly 24% for early deals to about 10% for recent ones. Core Scientific's first CoreWeave hosting deal sent its stock up more than 40%; TeraWulf's first Fluidstack agreement produced a nearly 60% surge. Neither reaction has been matched since. That decline suggests investors have stopped rewarding miners simply for announcing a pivot and are now waiting to see whether the infrastructure actually gets built and leased — precisely the distinction Bernstein's split ratings are pricing in.
This is the maturing phase of a story Blockhead has followed since the pivot accelerated this summer. Bitcoin miner Ionic Digital's Nasdaq debut in late July was pitched almost entirely on its 234-megawatt Texas AI buildout, and shares jumped roughly 26% on debut — a reaction that, per Blocksbridge's data, likely wouldn't repeat today. Around the same time, Blockhead covered how Poolin's bankruptcy sale of its Texas mining sites drew interest from AI and hyperscale bidders rather than only traditional mining buyers, and Bernstein separately flagged how Texas's data center interconnection audit in early August would make already-secured power capacity more valuable precisely because new supply was about to get harder to build.
Every miner in this cohort is telling investors some version of the same story: mining alone doesn't justify the valuation anymore, but the power and land already secured can be repurposed into AI infrastructure worth more per megawatt than bitcoin production ever was. Bernstein's split verdict on MARA and CleanSpark is the market's first real attempt to separate which companies are actually converting that thesis into signed, revenue-generating leases from which are still narrating it. With stock reactions to new AI announcements now roughly a third of what they were a year ago, the remaining unconverted miners have a shrinking window to close that gap before the market simply stops giving them credit for the story alone.