Bitcoin mining companies are being valued less for the Bitcoin they produce and more for the underlying power infrastructure they control. Riot Platforms' recent 20-year, $9.1 billion compute lease with Anthropic demonstrates this shift, converting a Texas mining campus into contracted AI infrastructure revenue that has nothing to do with block rewards or Bitcoin's price.

What Is Changing About How Miners Are Valued?

For years, Bitcoin mining companies were assessed almost entirely through mining economics: how much electricity they could secure, how efficiently they converted it into hashrate (a measure of computing power), and how much Bitcoin that capacity could generate. That framework is being disrupted, not because artificial intelligence is replacing Bitcoin mining, but because AI is creating an alternative market for the infrastructure miners accumulated during industry expansion.

The distinction matters because a Bitcoin mining facility's underlying components are broader than just crypto hardware. They include secured electricity, grid connectivity, land, substations, buildings, cooling systems, networking, and the ability to deploy computing capacity at scale. Bitcoin was the original customer for that infrastructure, but the infrastructure itself was never inherently tied to Bitcoin.

Keel Infrastructure, formerly Bitfarms, illustrates this transition. The company shut down its U.S. Bitcoin-mining operations as it shifted toward high-performance computing and AI infrastructure, with a pipeline of up to 2.2 gigawatts of energized, secured, and expansion power capacity. This move shows what happens when the market begins to value underlying infrastructure independently of the activity it was originally built to support.

How Are AI Companies Competing for Mining Infrastructure?

The attraction of mining sites to AI companies starts with something both industries require in enormous quantities: electricity. AI data centers face growing power demands, while securing new grid capacity can take years. Existing sites with substantial power allocations and established grid connections therefore have value beyond their original purpose.

Riot Platforms' agreement with Anthropic provides one of the clearest examples. The 20-year deal covers 191 megawatts of computing capacity at Riot's Rockdale, Texas campus and is expected to generate $9.1 billion in revenue, with extensions potentially taking the value to $16.1 billion. The agreement runs through June 2048 and carries two five-year extension options that could push total contract value to as much as $16.1 billion if both are exercised.

Riot shares jumped 25 percent to $24.40 in after-hours trading on the news, though the stock initially soared more than 20 percent in regular trading before giving back almost the entire move, a reminder that even a landmark contracted-revenue deal does not automatically produce a clean, durable re-rating in a single session.

The mechanism here is closer to commercial real estate than mining economics. Riot is not operating GPUs or selling AI compute directly; it is leasing power access, land, and data-center shell capacity to a tenant that brings its own hardware and workloads. That structure makes the $9.1 billion figure fundamentally different from mining revenue. Bitcoin mining income fluctuates with BTC price, network difficulty, and the fixed schedule of halving-driven reward reductions. A 20-year lease with fixed or contracted pricing insulates that portion of Riot's revenue from all three variables, trading mining's volatility for a bond-like, long-duration cash flow stream.

What Makes Mining Infrastructure Suitable for AI?

The AI opportunity should not be overstated. A large power connection does not automatically make a site suitable for AI workloads. AI data centers can require sophisticated cooling, high-density electrical systems, advanced networking, and significant reliability. Converting a mining facility can therefore require substantial capital, engineering, and time.

The quality of a mining company's physical footprint may matter more than simply its total megawatt capacity. Grid access, location, permitting, connectivity, expansion potential, and the ability to support different workloads can determine whether a mining site has genuine optionality or simply excess power.

Data Center Dynamics reported that initial capacity from Riot's Anthropic deal is not expected to go live until late 2027, with full deployment targeted for mid-2028, meaning the revenue ramp is gradual rather than immediate. There is also a regulatory variable specific to Texas. CNBC's report cited Compass Point analyst Michael Donovan noting that ERCOT's tightened scrutiny of new power projects could slow speculative build-outs across the state, even as it makes already-approved, grid-connected capacity like Riot's more strategically valuable to tenants racing for scarce power.

How Should Investors Evaluate Mining Companies Now?

Bitcoin mining still has an advantage that AI does not automatically overcome. Mining operations can be curtailed when electricity prices rise and restarted when economics improve. AI workloads generally require far greater availability and performance consistency. The trade-off is therefore not simply Bitcoin versus AI. It is flexibility versus contracted capacity, protocol-driven revenue versus enterprise revenue, and variable upside versus potentially more predictable infrastructure cash flows.

Optionality may become the new mining moat. The most strategically valuable mining companies may not be those that abandon Bitcoin the fastest. They may be those with the ability to choose. A site that can support Bitcoin mining today and potentially another high-density computing workload tomorrow has something more valuable than a single revenue stream: optionality.

Keel represents one end of that spectrum, moving its business away from Bitcoin and toward high-performance computing and AI. Riot represents another, retaining its mining operations while monetizing part of its infrastructure through long-term data-center agreements. The difference matters. It suggests there is no single template for the industry's evolution. Some miners may become dedicated AI infrastructure operators. Others may maintain Bitcoin mining alongside data-center businesses. The common factor is that their physical assets can increasingly serve more than one computing market.

Steps to Understanding the New Mining Valuation Framework

• Assess Power Access: Evaluate a mining company's ability to secure and maintain grid connectivity, as this determines whether infrastructure can serve multiple computing markets beyond Bitcoin.
• Evaluate Site Quality: Consider the physical footprint, including cooling systems, electrical density, networking, permitting status, and expansion potential, not just total megawatt capacity.
• Compare Revenue Models: Analyze whether the company generates income from volatile mining rewards or from contracted, long-duration leases that insulate revenue from Bitcoin price fluctuations and network difficulty changes.
• Monitor Conversion Costs: Track the capital, engineering, and time required to convert mining infrastructure for alternative workloads, as this determines the feasibility of optionality.
• Track Regulatory Environment: Watch for grid operator scrutiny and permitting changes in key regions like Texas, which can affect the strategic value of already-approved capacity.

AI is giving Bitcoin mining an external benchmark that the industry previously lacked. Mining economics have traditionally been assessed against Bitcoin's price, network difficulty, block rewards, energy costs, and hardware efficiency. Now miners can increasingly ask: What would another computing business pay for the same infrastructure ? The answer will depend on conversion costs, contract terms, utilization, customer concentration, and the economics of both businesses. But the comparison itself matters. It creates a market-based test for the opportunity cost of using scarce infrastructure for Bitcoin mining.

That could eventually change what investors look for. Hashrate and energy efficiency will remain fundamental. But alongside them, investors may increasingly consider power access, grid connectivity, site quality, expansion potential, and the ability to support alternative workloads. The result is a broader valuation framework in which the infrastructure supporting Bitcoin production can be assessed independently of the Bitcoin it produces.

Anthropic's appetite for infrastructure leases is not limited to Riot. The company has also struck a roughly $10 billion agreement with the months-old infrastructure startup Volta Infra Holdings and agreed in May to buy close to $45 billion in computing from Elon Musk's xAI. This pattern of diversified, multi-vendor sourcing from a lab whose Claude models sit at the center of the growing overlap between frontier AI and crypto-adjacent infrastructure suggests that the competition for power-backed computing capacity will intensify.

For years, miners competed to secure cheap power so they could produce more Bitcoin. The next phase may be about something broader: controlling power and infrastructure that multiple digital economies want. Bitcoin may have been the original customer. AI is forcing the market to ask whether it will remain the most valuable one.