Hundreds of billions of dollars in funding: Bitcoin mining companies collectively shift toward AI

Author: Wu Talks Blockchain

TL;DR:

• Mining companies collectively pivot to AI infrastructure: North American publicly listed Bitcoin mining firms are undergoing a dramatic identity transformation. They are moving away from “mining companies” that are highly volatile, strongly cyclical, and deeply tied to Bitcoin price, and are repositioning themselves as energy infrastructure operators and AI data center platform providers—backed by project financing in the billions of dollars and long-term contracts.

• The core moat is large-scale power plus rapid delivery capability: The fundamental reason AI cloud service providers choose to partner with mining companies is that miners control the most scarce resources in the AI era—ready-made grid interconnection, mature substation and transmission/distribution infrastructure, and rapid delivery capability. This can significantly shorten the GPU cluster deployment cycle.

• Top companies lock in ultra-large long-term agreements: Typical miners such as Core Scientific, Hut 8, Iris Energy, and TeraWulf are already deeply integrated with leading AI customers. They have signed long-term IT capacity leasing agreements totaling hundreds of MW, and for some firms, the base-term value of long-term contracts reaches a tens-of-billions-of-dollars level, with even greater potential for renewals.

• Evolution of financing structures drives a logic switch in valuation: Mining companies are increasingly adopting project-level debt financing, Triple-net (three-net leasing), and Take-or-pay terms. This shift makes their revenue model increasingly resemble traditional data center REITs. As a result, capital markets are repricing them—from commodity-cycle companies to infrastructure assets with stable cash flows.

• The pivot faces major tests of capital and execution risk: While embracing infrastructure, mining companies also face high-cost capex upgrades (on the order of millions to tens of millions of USD/MW), overly high customer concentration, and challenges in transitioning from pure compute deployments to AI technology operation and maintenance. The market’s current valuation is based on future successful delivery; if execution falls short of expectations, the valuation logic could reverse.

Industry background: halving pressure and AI infrastructure demand emerge at the same time

In 2026, North American publicly traded Bitcoin mining companies are going through a significant identity shift.

Once viewed by capital markets as highly volatile, strongly cyclical “mining companies” that are deeply tied to Bitcoin prices, these firms now frequently show up in discussions about AI data centers, power infrastructure, and the revaluation of energy assets. From long-term AI hosting contracts, to billions of dollars in financing, to ultra-long-term leasing agreements—some traditional miners are trying to reposition themselves as energy infrastructure operators and AI data center platform providers.

This change is not accidental. It is the result of two forces acting at the same time: on one side, the continued pressure on Bitcoin mining economics after the 2024 Bitcoin halving; on the other, the rapid expansion of AI infrastructure demand.

After the 2024 Bitcoin halving, miners’ profitability was squeezed even further. Many listed mining firms faced substantial pressure with their all-in costs (including energy, depreciation, interest, taxes, and equipment) amid Bitcoin price volatility. More and more miners began seeking diversified growth paths.

Meanwhile, demand for AI training, inference, and high-performance computing is growing rapidly. Large-scale data centers are facing new constraints—not a shortage of GPUs, but a shortage of electricity, a shortage of grid interconnection, and a shortage of infrastructure that can be delivered quickly. For hyperscalers and AI cloud service providers, even if they have sufficient capital, they may not be able to complete the construction of hundreds of MW of campus-scale facilities within a reasonable timeframe. This supply-demand mismatch is what gives miners—with mature power infrastructure, ready industrial parks, and large-scale energy interconnection capacity—new strategic value.

The real question is not whether these companies can do AI well, but rather: why is the capital market willing to provide these companies with billions of dollars in financing? The answer is that they control one of the scarcest resources in the AI era—large-scale power and infrastructure delivery capability that can be deployed quickly.

Repricing from “mining machines” to “power assets”

In recent years, the market’s valuation logic for miners has been very clear. At its core, it is a highly leveraged Bitcoin beta. When the Bitcoin price rises, miners often benefit from higher elasticity; when halving arrives, profit margins get compressed; when the Bitcoin price falls, they enter survival mode. For most investors, miners have long resembled a highly cyclical commodity asset, with their key variables consistently revolving around the Bitcoin price, network difficulty, and energy costs.

But the explosion of AI is changing this loop.

AI training, inference, and high-performance computing require massive computing power, but the fundamental bottleneck is not the GPU itself. What is truly scarce is the ability to support GPU clusters with large-scale, stable electricity; grid interconnection; substation, transmission, and distribution infrastructure; industrial land; cooling systems; and rapid delivery capability. These are precisely the assets that North American miners have kept investing in during previous bull-market cycles. The large-scale mining farms they built in pursuit of Bitcoin hash rate are now being repriced as the foundation of AI infrastructure. For an increasing number of miners, their business model is starting to shift from simply selling Hashrate to gradually selling power capacity and data center capacity.

Why AI clients are willing to work with miners

For hyperscalers and AI cloud service providers, choosing to partner with miners is not only because electricity might be relatively cheaper. Their core bottleneck right now is that even if they can procure GPUs, they may still not be able to obtain enough scale of power interconnection and data center capacity that is available immediately within a reasonable time. Compared with building from scratch, miners with existing grid interconnection, industrial parks, and mature power infrastructure can significantly shorten deployment timelines. Therefore, what the market is buying is not only electricity itself, but “large-scale power capacity that can be delivered quickly.”

More importantly, in today’s North American data center market, the truly scarce resource has been shifting from the GPU itself to Time-to-Power. For large AI clusters, waiting years to complete grid approvals, transmission construction, and park development often means missing model training cycles and business windows. In contrast, some miners already have scalable power interconnection, mature parks, and potential development capacity in the hundreds of MW—allowing them to compress deployment cycles from years to a shorter period of time. Against this backdrop, what AI clients are effectively purchasing is not just power capacity, but an infrastructure delivery capability that supports fast deployment and ongoing expansion.

Breakdown of typical company cases: contracts, financing, and valuation switches

Core Scientific (CORZ)

Core Scientific (CORZ) and CoreWeave’s long-term cooperation is an early, representative case. The contracted capacity expanded to about 590 MW of key IT load. The disclosed base-term contract value has most recently reached approximately $8.7B+. With subsequent capacity expansion, total revenue potential is expected to exceed $10B. In July 2025, CoreWeave’s all-stock acquisition of CORZ (valued at approximately $9B) was announced, but it was officially terminated on October 30, 2025, due to shareholders not approving the merger agreement. CORZ is still continuing its HPC/AI business as an independent company.

Hut 8 (HUT)

A standout in execution. Hut 8 has signed long-term leasing agreements with leading AI clients. At the River Bend campus, it signed a 15-year, 245 MW IT capacity lease agreement, with a base-term contract value of $7B (including renewal potential, which is higher). At the Beacon Point campus (Texas), it added a 15-year, 352 MW IT capacity lease; the base-term contract value is $9.8B. Combined, the two contracts represent contracted AI capacity of 597 MW, with a base-term total contract value of approximately $16.8B. The company supports project development through project financing and uses a triple-net lease structure to enhance long-term cash flow predictability. The market is also beginning to reassess company value more around long-term cash flows and infrastructure attributes.

Iris Energy / IREN

The company has signed an agreement with Microsoft to deploy about 200 MW of IT load at the Childress 750 MW campus. The five-year contract value is approximately $9.7B (corresponding to about $1.94B annualized, based on company disclosures and market estimates). The related deployment arrangements are driving the company’s shift toward an AI cloud infrastructure provider. The market has started to reevaluate its value more based on long-term contract revenue and infrastructure cash flow logic. The company has also signed hardware procurement agreements with Dell and is leveraging its renewable energy advantages to advance the deployment.

TeraWulf (WULF)

TeraWulf (WULF) HPC/AI business is gradually becoming one of the company’s important growth engines. The company has partnered with Fluidstack, including the Abernathy campus 168 MW AI compute joint venture project (a 25-year agreement, with contract revenue of about $9.5B). It has also completed the related project financing to support the development of AI infrastructure, accelerating its transition into an AI data center platform.

Summary

The cumulative publicly disclosed AI-related contracts, project revenue potential, and market estimates across the industry have already reached the hundreds-of-billions-of-dollars scale. For some leading companies, the contribution from AI-related revenue is starting to rise, and financing structures are gradually incorporating more project-level debt, long-term notes, and infrastructure financing tools—strengthening the infrastructure-attribute valuation logic.

Most AI projects are expected to be delivered and come online gradually in 2026–2027. As of mid-2026, they have not all completed deployment, and actual revenue contributions are still in the ramp-up stage.

The real logic behind the financing wave: infrastructure-based valuation

What is most worth watching in this round of financing is not just the contract size itself, but the fact that the financing structure is changing.

Previously, mining-company financing usually relied heavily on equity financing, equipment-backed loans, or cyclical financing tools tied to Bitcoin prices—meaning financing costs were deeply bound to volatility in the crypto market. But as some miners start to sign long-term AI hosting agreements, ultra-long-term lease contracts, and data center projects with clearly defined cash flow structures, the capital market has begun to look at these assets through a different lens.

Some companies have started to obtain project-level debt financing, non-recourse or credit-enhanced structures, triple-net long-term leases, and take-or-pay contract arrangements. The core significance of these financing tools is not merely “borrowing more money,” but rather that the revenue structure is becoming more long-term and more predictable—gradually taking on cash-flow characteristics closer to traditional infrastructure assets.

This indicates that the market is trying to switch valuation for mining companies—from typical commodity-cycle companies to infrastructure assets and growth-oriented energy platforms. What the market is truly betting on is not whether these firms can become the next OpenAI, but whether they can continuously deliver hundreds of MW of power capacity and have rapid delivery capability. The recurring keywords in contracts are often not “models,” but power capacity, IT load, and interconnection.

Risks are real and huge

Market optimism does not mean risks disappear. On the contrary, AI transformation itself may become one of the most capital-intensive and most difficult transitions in mining companies’ history.

First is capex pressure. Converting mining farms into high-density AI data centers is far from a simple equipment replacement. It requires more complex cooling systems, higher-density power architectures, and large upfront construction investments. For some projects, depending on project specifications, construction costs per MW can reach millions to over tens of millions of USD. That means even if financing is obtained, the project delivery pace will directly affect returns and pressure the balance sheet.

Second is customer concentration risk. Many contracts currently rely on a small number of hyperscalers, AI cloud service providers, or large model companies. If deployment progress slows, customer demand changes, or AI infrastructure investments enter a correction cycle, the value of long-term contracts may also be reassessed. In addition, what mining companies have historically been good at is ASIC deployment, managing electricity costs, and operating mining sites. But operating hyperscale AI infrastructure requires new sales capabilities, a technical operations and maintenance system, and a more complex partner ecosystem.

To some extent, the high valuations the market gives today are, in essence, pricing in execution success in the years ahead. If delivery speed, customer demand, or financing conditions change, this valuation-switch logic also carries the risk of reversing.

A deeper question: Are mining companies still mining companies?

If long-term selling of power capacity can generate more stable cash flows than simply selling Hashrate; if long-term leasing models can provide higher income predictability; and if infrastructure valuations remain higher than traditional mining valuations—then a more fundamental question begins to emerge: in the future, will these companies still be Bitcoin miners?

In the past few years, markets have tended to treat mining companies as typical cyclical assets, with the core variables being the Bitcoin price, network difficulty, and energy costs. But as more companies build business models around power capacity, data center parks, and long-term infrastructure contracts, their revenue structures, financing methods, and even investor narratives are changing.

Bitcoin halving may not end the growth logic of these firms, but it is forcing miners to redefine themselves. Some leading players may ultimately evolve into infrastructure platforms with AI data center operations at the core. Some companies may continue to maintain a hybrid “mining + AI” model. And companies that transform more slowly will still be influenced by traditional mining cycles.

In the end, this transformation may determine not only the fate of the mining companies themselves, but also become an important case study for how energy assets in the AI era are being repriced.

From this perspective, what capital truly purchases may never be computing power itself, but the ability to deliver computing power continuously—supported by electricity, land, network access, and infrastructure capabilities.