Bitcoin mining companies are paying different borrowing rates as they expand into artificial intelligence and high-performance computing. According to BlocksBridge Consulting, IREN’s financing agreements show how customer commitments, project requirements and lending terms can affect the cost of funding new data centers.
In its October 8 Miner Weekly newsletter, BlocksBridge said lenders have become more selective about financing data center projects as Bitcoin miners compete to build AI infrastructure.
Having a major AI customer is only one part of the process. Companies still need enough money to buy computing equipment, build facilities and install the power systems required to deliver their contracts.
The consultancy said lenders are increasingly looking at the customer, contract terms and the likelihood that a project will be completed successfully.
IREN provides one example. The company secured a $3.65 billion GPU financing package announced on June 1 to support its cloud agreement with Microsoft. BlocksBridge reported a blended debt cost of 6%.
IREN also disclosed a separate $2.4 billion financing arrangement carrying a fixed interest rate of 9%. That package supported its air-cooled expansion at Mackenzie.
Although the rates differ by three percentage points, BlocksBridge cautioned against treating them as a direct increase in IREN’s borrowing costs. The two agreements support different projects and carry different risks.
The consultancy also noted that financing for GPUs does not necessarily cover the full cost of an AI data center. Companies must still fund buildings, electrical infrastructure, installation, unexpected construction expenses and working capital.
IREN has continued expanding its AI business through new customer contracts. In July, the company announced an additional $2.8 billion in multiyear AI cloud agreements and raised its year-end annualized revenue target from $3.7 billion to more than $4 billion.
IREN said around 85% of its revised revenue target was under contract. It also reported advance customer payments equal to approximately 45% of the GPU capital spending associated with recent agreements, reducing the amount it needed to finance directly.
The company reported approximately $7.6 billion in cash and cash equivalents, including restricted cash, as of June 30.
BlocksBridge also highlighted Cipher Digital’s financing arrangements with Google as an example of how companies can secure project funding without relying entirely on conventional interest payments.
Under a September 2025 agreement, Google agreed to backstop $1.4 billion of Fluidstack’s lease obligations connected to Cipher’s Barber Lake project. In exchange, Google received warrants representing approximately 5.4% of Cipher’s equity on a pro forma basis, subject to adjustments.
The arrangement helped support project financing, but it also created a potential cost for existing shareholders. If Google exercises the warrants for shares, its ownership could dilute other shareholders. The agreement also allows cash settlement under certain circumstances.
BlocksBridge said investors should look beyond interest rates when comparing AI financing deals. Warrants, equity contributions, parent guarantees and restrictions on project cash can all affect the overall cost of funding.
Cipher has also secured a separate $200 million credit facility backed by financial institutions. The facility matures in March 2030 and includes an option to increase its capacity by up to $50 million.
The company said the credit facility carries an interest rate tied to the Secured Overnight Financing Rate, plus 1.25 to 1.75 percentage points. The rate can fall depending on Cipher’s total debt-to-market-capitalization ratio.
At the time of its announcement, the facility remained undrawn. Cipher said the borrowing capacity could support liquidity, working capital and expansion, including construction and equipment installation at its AI and high-performance computing campuses.
In May, Cipher also disclosed a 15-year lease for its third major AI campus with an investment-grade customer.
IREN has raised capital through convertible debt as well. In May, the company raised $3 billion through senior notes carrying a 1% coupon and maturing in 2033.
The offering included $2.6 billion in base notes and an additional $400 million purchase option that was fully exercised. IREN reported net proceeds of approximately $2.96 billion after fees and expenses.
The company allocated $201.3 million to capped call transactions designed to reduce potential shareholder dilution if the notes convert into shares. The initial cap price was $110.30 per share, twice IREN’s share price of $55.15 on May 11.
IREN said the notes carried a 32.5% conversion premium and could not generally be put back to the company before maturity, except under certain standard change-of-control conditions.
The growing cost of AI infrastructure is also visible across the wider mining industry.
BlocksBridge’s August 20 analysis found that nine publicly traded mining companies spent a combined $5.11 billion on capital expenditure during the first half of 2026, compared with $341.2 million in directly reported AI and high-performance computing revenue.
The spending figure included purchases of hardware, property, equipment and other productive assets. Because some of that investment continued to support Bitcoin mining, it does not represent AI spending alone.
Combined AI and high-performance computing revenue reached $205.8 million in the second quarter, up 52% from the previous quarter.
The shift toward AI has also affected investment products. In August, CoinShares renamed its WGMI exchange-traded fund the CoinShares Bitcoin Mining and Digital Power ETF and expanded its eligible investments to include data center operators, AI semiconductor companies, power producers and advanced computing businesses.
The fund reported 29 holdings and approximately $225.6 million in assets as of August 18. Its investment policy requires at least 80% of net assets to be invested in qualifying companies. The fund does not hold Bitcoin directly or through derivatives.
Despite the growing interest in AI, completing planned infrastructure projects will require substantial additional funding. BlocksBridge estimated in June, using VanEck data, that publicly traded Bitcoin miners needed approximately $50 billion to finish their planned AI infrastructure developments.
IREN accounted for an estimated $21.1 billion of that funding gap, followed by Riot Platforms at $7.2 billion and HIVE Digital at $4.6 billion.
The figures show the scale of the financial challenge facing miners as they move into AI. Large customer contracts can help attract lenders, but companies must still cover construction costs, equipment purchases and other expenses before their projects generate enough revenue.
As a result, financing costs will depend not just on interest rates, but also on the strength of customer agreements, the structure of each deal and the amount of capital required to bring projects online.
