Bitcoin Miners Take On Debt To Fund AI Data Centers

Bitcoin mining companies are looking to convert their power sites into data centers for artificial intelligence. This transformation would require nearly 50 billion dollars in the short term, while lenders impose increasingly variable financing conditions.

AI Contracts Are Not Enough to Finance Infrastructures

Mining companies have an advantage sought after by tech firms: land already connected to the power grid and capable of hosting significant energy consumption. They can therefore offer their capacities to artificial intelligence and high-performance computing operators.

However, this conversion is not limited to replacing mining machines with graphics processors. IA-dedicated data centers require more resistant buildings, advanced cooling systems, fast connections, redundant power supplies, and a permanent availability level.

A classic mining site costs between 700,000 and 1 million dollars per megawatt. An AI infrastructure equipped with liquid cooling can cost between 8 and 15 million dollars per megawatt.

BlocksBridge Consulting explains :

Major lenders have become more selective regarding loans to data centers.

The firm specifies that their decisions now depend on the client, the signed contract, and the company’s ability to complete the project.

Therefore, even a commercial commitment of several billion dollars does not guarantee that the mining company has the necessary capital to build the promised data center.

Bitcoin Mining: IREN Pays Between 6% and 9% to Finance Its Projects

IREN illustrates the gaps that can appear between two financing operations concluded by the same company. In June, the group obtained 3.65 billion dollars to finance graphics processors, with an average interest rate of 6%.

Two months later, IREN signed another financing of 2.4 billion dollars at a fixed rate of 9%. Both operations cover different projects and cannot be directly compared. However, the gap shows that lenders separately assess the risk of each infrastructure.

IREN would still present the largest financing deficit among publicly traded mining companies engaged in AI. An estimate published in June assessed its uncovered needs at 21.1 billion dollars. Riot Platforms followed with 7.2 billion, ahead of HIVE Digital and its 4.6 billion.

In total, publicly traded companies in the sector would need about 50 billion dollars in the short term to carry out their announced projects. Their longer-term needs could approach 221 billion.

This debt adds to a balance sheet already weakened by the decline in mining profitability. At the bitcoin market peak in October 2025, the debt of publicly traded mining companies had increased by about 500% compared to the second quarter of 2024.

Financing Can Also Dilute Shareholders

Interest is not the only cost borne by companies. Some partners request potential shares in exchange for their financial guarantee.

As part of an operation involving Cipher Digital, Google agreed to guarantee up to 1.4 billion dollars of lease obligations. In return, the tech group received warrants that could represent 5.4% of Cipher’s capital, subject to certain adjustments.

If these securities are exercised, existing shareholders will see their stake decrease. This mechanism allows Cipher to access more favorable financing but transfers some of the future value to the guarantor.

Mining companies also use convertible bonds, which can be converted into shares. These instruments sometimes reduce immediate interest at the cost of potential additional dilution.

The risk increases if projects are delayed. The company must then continue to pay its interest before receiving the expected revenues from its clients. A major contract can thus become a source of financial tension when capacities are not delivered on time.

More Than 100 Billion in Contracts but Few Actual Revenues

The studied groups announced over 100 billion dollars in contracts related to AI and high-performance computing. Yet, these commitments currently produce only about 1.1 billion dollars in annualized revenue.

Nearly 4 gigawatts of capacity have been contracted, but only 550 megawatts actually generate billings. Companies have delivered only about 25% of the already leased capacities.

The gap also appears in expenses. Nine miners invested 5.11 billion dollars in the first half of 2026, against only 341.2 million in revenue directly attributed to AI and high-performance computing. The investment-to-revenue ratio therefore reaches fifteen to one.

AI can offer more regular revenues than bitcoin mining, which is subject to the BTC price, network difficulty, and electricity cost. But this stability will exist only after the delivery of data centers.

The transition thus creates a new risk profile. Mining companies become heavily indebted infrastructure operators whose success depends less on their bitcoin reserves than on their ability to build on time and control costs.