Starboard Value is pressing Riot Platforms to fast-track a shift from Bitcoin mining into AI and high-performance computing (HPC) colocation, arguing the company’s Texas footprint could unlock $9 billion to $21 billion of equity value. The activist’s mid-February 2026 letter reframed Riot’s 1.7 GW of permitted, grid-connected power capacity as a monetizable data-center asset rather than a mining-only input.
The message landed with markets, with Riot’s shares rising roughly 6–9% after the letter as investors leaned into the idea of contracted colocation revenue versus mining volatility. The push also built on Riot’s January 2026 25 MW proof-of-concept lease with AMD, which Starboard positioned as an early signal that demand exists beyond pure crypto.
Why Starboard sees AI/HPC colocation as the higher-value lane
Starboard anchored its thesis in a market-structure claim: high-density data-center space with adequate power is scarce, and Riot already controls the power layer at scale. The letter’s central contention is that Riot’s 1.7 GW of permitted, grid-connected capacity in Texas can be repurposed into long-term colocation contracts for AI/HPC customers.
The activist also emphasized the gap between mining economics and the alternative margin profile it believes colocation can deliver. Starboard cited Riot’s all-in production cost at $89,000 per BTC against a cited market price of $67,000, while pointing to 80–90% margins on long-term contracted AI colocation leases.
Starboard framed urgency as a competitive necessity, arguing that “time is of the essence” to secure meaningful AI/HPC agreements beyond the AMD pilot. The implication is that speed-to-contract, not just power capacity, becomes the key driver of whether the projected $9–$21 billion value range is achievable.
Execution risks and near-term milestones
The letter did not treat the pivot as automatic, and it outlined material delivery risk in moving from mining operations into a services-led colocation business. Starboard acknowledged that converting mine-oriented infrastructure into AI/HPC colocation requires different operating capabilities, a different contracting cadence, and stronger data-center dealmaking than Riot has demonstrated so far.
It also flagged exposure to changing market sentiment in AI infrastructure, where valuation expectations can compress if pricing or demand assumptions soften. Starboard explicitly warned that AI infrastructure multiples could be stretched, meaning upside targets remain sensitive to execution and market repricing.
From an operator’s standpoint, the near-term playbook is straightforward but execution-heavy: translate permitted megawatts into high-density racks, then translate racks into multi-year contracts. The market will be watching whether Riot can move quickly from a 25 MW proof of concept to larger, durable AI/HPC commitments while proving stable delivery at scale.
