The importance of companies, especially $IREN, that are former BTC miners shifting to AI/neocloud, stems precisely from the energy bottleneck I've described in the appendix:
AI growth is no longer just about GPUs; it's about electricity, land, cooling, transformers, grid connectivity, and permits. According to the IEA, data center electricity consumption was around 415 TWh in 2024 and is projected to more than double to around 945 TWh by 2030; AI is the main driver of this growth.
Therefore, former BTC miners have suddenly transformed from "ordinary miners" into companies offering "energy + infrastructure options." Bitcoin mining already provided them with experience in finding inexpensive power, acquiring large plots of land, building high-density infrastructure, working with the power grid, and running uninterrupted operations. AI customers want all of these at once; meaning the scarcest asset these companies have is often not GPUs, but readily available capacity. This conclusion is also supported by the fact that companies highlight phrases like “power-first,” “AI-ready site,” “data center development,” and “AI infrastructure platform” in their investor presentations.
It’s also easy to see why hyperscalers are more willing to deal with these. Large cloud companies and AI buyers don’t want to wait, as grid connectivity and new power capacity can take years. In contrast, these older mining companies offer “ready-made power” and “ready-made space,” giving them the chance to deploy an AI data center without waiting.
Examples – IREN’s site shows 6 GW of power secured, 2,100 MW under construction, and AI Cloud infrastructure; and its $9.7 billion deal with Microsoft includes plans to build GPU clusters on a 750 MW campus in Childress. This is exactly what the market is looking for: gaining momentum through a partner with power, cooling, and scalable site.
The same logic is seen in Hut 8 and TeraWulf. Hut 8 says it has commercialized the first phase of its Beacon Point campus at a scale of 1 GW, secured a 15-year, 352 MW AI data center lease, and accumulated a total of $16.8 billion in contractual lease revenue across its two hyperscale AI campuses. TeraWulf, meanwhile, announced a 200+ MW, 10-year AI hosting deal with Fluidstack, generating approximately $9.5 billion in contractual revenue, while attempting to mitigate financing risk with the Google backstop. These types of contracts represent the most powerful step in transforming a “story on paper” into a “cash-flow infrastructure business.”
CleanSpark and Riot are other examples of this transition. CleanSpark says it controls a portfolio of over 1.8 GW of power, land, and data centers; and has developed a site capacity of up to 600 MW and utility potential capacity of up to 890 MW in Texas. Riot, strengthening its identity as a digital infrastructure company, announced a 25 MW, 10-year data center lease with AMD at the Rockdale site. On the Core Scientific side, the increase in colocation revenue from $8.6 million in the first quarter of 2025 to $77.5 million in the first quarter of 2026 shows that the demand for AI-focused capacity is beginning to translate into revenue for the company.
What makes these stocks significant isn't just the "AI narrative"; there's actually a three-tiered value. First, they possess scarce and strategic resources: power and land. Second, they have the operational capacity to quickly convert these resources into revenue. Third, the same asset pool can be used for BTC mining, AI hosting, and potentially other high-density computing tasks in the future. In other words, the investor is paying for power-backed platform optionality, not just a single business model. Therefore, these companies are often priced not like classic mining stocks, but like "AI infrastructure platforms."
But the most critical point here is: they aren't all automatic winners. In this model, financing costs, construction risk, customer concentration, electricity prices, permitting processes, and contract quality are decisive factors. Therefore, the market should view these companies not simply as "miners," but as "infrastructure operators that monetize access to energy"; otherwise, the gap between the narrative and actual cash generation could widen. For investors, the real question is no longer "Will AI grow?" but "Who has the ready capacity and resources to operate it without waiting for this growth?"
This is not investment advice.
