As tech companies consider how best to scale the buildout of data centers over the next few years, Morgan Stanley has identified what it views as an up-and-coming, underappreciated opportunity: converting bitcoin mining sites to data centers. “We believe we are at a very early stage in terms of investor recognition of the upside potential stemming from converting Bitcoin mining facilities to HPC Data Centers,” analysts led by Stephen Byrd said in a note on Wednesday, referring to high-performance computing. “We think it is plausible to expect that a REIT structure is the ‘end game’ for Bitcoin mining assets converted to [data centers], and that the potential upside may be far more significant than appreciated.” The Wall Street investment bank’s calculation is as follows: its indicative valuation of a U.S. bitcoin mining facility ranges from between $2 and $3 per watt of capacity, while the value to a hyperscaler — like Microsoft, Alphabet and Amazon — of three years of time savings to power up a new data center is more than $10 per watt. Over the past few months, Morgan Stanley has seen more Bitcoin miners’ management teams focus on such a conversion opportunity, to about 80% from roughly 50%. These managements that have been engaging with data center developers and tech hyperscalers say demand and pricing feedback has been positive, Byrd said. This opportunity is even more attractive to investors, considering there could be a high likelihood of a significant shortage of new data centers between 2025 and 2027, driven by a lack of available grid interconnection, Morgan Stanley said. Bitcoin miners that Morgan Stanley thinks could benefit from this opportunity include Bitdeer Technologies Group , Cipher Mining , TeraWulf , Iris Energy and Galaxy Digital (None of which are covered by Morgan Stanley research.) Each stock is buy-rated from analysts polled by FactSet, and their consensus price targets indicate large potential upside. Cipher Mining, a potential standout, has actively pursued data center conversion opportunities for hyperscaler customers over the past few months, according to Morgan Stanley. “We like management’s commercial judgment and focus on achieving the best risk-adjusted outcome for shareholders,” the bank said in the 19-page note. Cipher “has large sites that appear to be well suited for the conversion opportunity, and the company has significant expertise in site development.” Cipher’s shares are down roughly 3% for the year and slid more than 35% over the past month, as the cryptocurrency market plunged in early August. The company’s second-quarter earnings and revenue released last week were also worse than analysts had estimated. Analysts polled by FactSet expect shares could gain 79% over the next year, based on consensus price targets. Infrastructure-focused bitcoin mining company Terawulf could also announce data center development transactions, according to its management commentary, per Morgan Stanley’s research team. “The company has some of the most attractive sites for Data Center development, and management has deep expertise in developing complex power projects, which we view as a plus. We also believe management has deep expertise in developing optimal financing strategies,” Byrd said. He noted, however, that the company’s data center potential could be limited because of its potentially smaller footprint in terms of megawatts. Shares are up 92% this year, but are down about 29% since their 52-week high in mid-July. Bitdeer Technologies Group is another potential bitcoin miner that could enter the AI data center market. Its shares are up 8% so far this week, but slid this month after the company sold $150 million of convertible senior notes due 2029. The stock is down 29% this year, but analysts polled by FactSet expect nearly 111% potential upside. Bitdeer announced better-than-expected earnings on Aug. 12, but failed to meet analysts’ revenue estimates. Bitdeer lost 14 cents a share in the second quarter, narrower than its loss of 36 cents a share in the year-ago period, while revenue of $99.2 million missed the FactSet revenue consensus.