KI trading is booming: Why hyperscalers and Bitcoin miners dominate data center capacity

As concerns grow about a possible weakening of the AI hype, a fundamental business model remains untouched: the demand for power capacity. According to Joe Nardini, head of investment banking at B. Riley Securities, Bitcoin miners, large technology companies, and AI infrastructure developers continue to compete intensely for data center capacity—partly extending well into December 2025. The reason is simple: hyperscalers, the massive cloud computing centers operated by tech giants like Amazon, Google, and Microsoft, require continuous enormous amounts of energy.

What is a hyperscaler and why do they need so much power?

A hyperscaler is a large-scale data center operator—typically a tech corporation—that runs interconnected, state-of-the-art infrastructures for cloud computing, data processing, and AI workloads worldwide. Amazon Web Services (AWS), Microsoft Azure, and Google Cloud are the most well-known examples. These hyperscalers differ fundamentally from traditional data center operators: they have enormous financial resources, international presence, and a constantly growing demand for GPU capacities for AI models and high-performance data processing (HPC).

GPU-intensive data center capacity has become a bottleneck. Nardini observes that GPU-enabled facilities attract multiple creditworthy tenants at attractive terms. Bitcoin miners, increasingly aligning their business models with HPC and AI hosting, are receiving higher valuations and better access to capital—indicating that the market is rewarding this diversification. The BTC price is currently around $77.31K, illustrating that even at these levels, margins for pure mining operations are strained after the 2024 halving. The strategic response: repurposing data center spaces for higher-value AI and HPC workloads.

Billion-dollar deals: how hyperscalers and miners compete for megawatts

Deal-making on Wall Street is sustained by a simple economic reality: power is the new gold. In highly competitive situations with top-tier energy sources and favorable locations, dollars per megawatt can appear extremely attractive. Nardini reported processes involving valuations over $400,000 per megawatt, with potential up to $450,000 per megawatt, depending on negotiation outcomes. He has even seen earlier transactions at $500,000 to $550,000 per megawatt—prices that underscore the scarcity of high-quality capacity.

This was concretely demonstrated in December 2025, when Hut 8, a publicly traded mining company, signed a 15-year lease agreement with FluidStack. The deal included 245 megawatts of IT capacity at the River Bend Campus and was valued at $7 billion—a signal that top-tier operators with hyperscaler-standard facilities are achieving surprisingly high valuations. The company’s stock subsequently rose by up to 20 percent.

However, the market differentiates by location quality. Less attractive or distressed sites continue to attract bids, but at significantly lower levels: between $100,000 and $250,000 per megawatt. Buyers in this segment value the power resource but remain skeptical about market potential or geographic location. This has triggered a surprising phenomenon: old industrial facilities, some 160 years old, are suddenly becoming valuable assets if they have power connections.

Buyers and sellers: a new market ecosystem is emerging

Who are the driving forces behind these deals? On the buyer side, three groups dominate: hyperscalers (cloud infrastructure providers), AI companies, and Bitcoin miners. On the seller side, the circle is widening. It’s no longer just crypto-related actors but also traditional industrial companies converting their older or unused facilities with energy connections into data center capacity.

A case example: a private seller of a similar property attracted interest from about 25 potential buyers—miners, hyperscalers, and AI firms—who all requested confidentiality agreements. Another private investor converted older office buildings into modular power capacities, each with “30-megawatt units,” and is currently seeking additional financing for expansion. This illustrates an unusual strategic decision for asset holders: sell to a hyperscaler-like operator or become a developer themselves?

Another indicator of market tension: in at least one negotiation, a tenant was willing to pay rent in advance before the facility was completed—a strong signal of how scarce high-quality, available capacity truly is. This underscores that hyperscalers and AI developers are willing to accept premium prices and upfront payments to secure guaranteed capacity.

2026 in the sign of demand: why hyperscalers will continue to invest

Looking ahead to 2026, Nardini sees a favorable environment. If interest rates fall, a “risk-on environment” will emerge, favoring risk assets—good for his investment banking business, as he openly admits. But his optimism is based on operational facts: tenants are present, prices remain robust, and if a developer cannot use their site themselves, another buyer can quickly be found.

His core caveat: if developers cannot rent out the built capacity or cannot achieve the necessary prices, then concern is warranted. Currently, he observes the opposite. “The fundamental structure of the business remains intact,” he summarized. The demand for data center and AI/HPC capacities remains “unabatedly high.” Developers with data center spaces are receiving inquiries from multiple creditworthy tenants at attractive terms—a scenario that strengthens the business fundamentals.

The fact that hyperscalers, miners, and AI companies are fueling this competition indicates that the fundamental problem—power shortages—has not been solved. Infrastructure is expanding faster than demand, and this asymmetry is driving valuations upward. Nardini succinctly concludes: “The AI market is still active.” In other words: as long as hyperscalers and miners compete for megawatts, deals will stay alive.