Shares of Riot Platforms ( RIOT ) rose nearly 9% on Wednesday after activist investor Starboard Value LP released a letter pressuring the company to accelerate its transition from bitcoin mining to AI infrastructure provider. The goal is for Riot to pursue high-margin artificial intelligence and high-performance computing (AI/HPC) hosting deals.
Riot’s 1.7 gigawatts of fully available power capacity makes the company “well-positioned to execute high-quality AI/HPC deals,” Starboard said, highlighting two of Riot’s Texas-based sites, Corsicana and Rockdale, as “prime” locations for data center development.
Starboard said that if Riot can monetize its power in line with recent transactions in the space, “it could generate more than $1.6 billion” in annual EBITDA. The group praised Riot’s recent deal with AMD, which is expected to provide $311 million over 10 years.
With a market capitalization of $4.25 billion, Texas-based Riot is the fifth largest bitcoin mining company in the United States. Its shares are up 19% in the past year, but remain down about 80% from highs hit during the bitcoin bull market in 2021. They have also underperformed miners such as IREN, Cipher Mining and Hut 8, which were quicker to recognize and move to AI strategies.
Starboard was Riot’s fourth-largest shareholder at the end of last year, and this isn’t the first push at the company. In December 2024, Starboard requested Riot to convert some of its bitcoin mining sites into data centers capable of hosting HPC machines to support large tech companies.
While Riot Platforms has built its business around bitcoin mining, the pivot toward AI infrastructure could diversify revenue as power-hungry models like OpenAI’s GPT-4o and others drive demand for data centers. Riot’s power access, a rare commodity in the current energy-constrained data center market, could be used to lease capacity to large AI companies.
Starboard urged CEO Jason Les and executive chairman Benjamin Yi to act “quickly” and position Riot as a long-term infrastructure provider for AI workloads.
