What if the biggest numbers in crypto right now are not the ones that matter most for the companies announcing them? Bitcoin miners have been locking in AI contracts worth billions, and the market reaction has been loud. Yet the cash needed to actually pour concrete, buy GPUs, and power those facilities arrives through a different set of deals entirely. Customer advances, equipment loans, and carefully worded guarantees are doing the heavy lifting while the headline lease values stretch across decades.
I have been watching this transition closely. The shift from pure Bitcoin mining to hybrid AI compute looks inevitable on paper, but the funding mechanics are messier than the press releases suggest. A twenty-year commitment sounds impressive until you realize the first meaningful payments may not start until late 2027 or even 2028. Between those dates sits a long stretch of construction risk, interest costs, and the simple question of who writes the checks while the buildings are still empty shells.
The Real Money Trail Behind Billion-Dollar AI Deals
The contracts themselves can make construction easier to finance. Their value to shareholders, however, still depends on three practical factors: the cost of borrowing, the equipment the operator must buy, and how soon a customer begins paying for an operational facility. That distinction is easy to overlook when the announcement focuses on the total lifetime value.
Take one of the larger examples currently circulating. A major miner secured credit support covering roughly $19 billion in lease commitments spanning twenty years. Simple division produces an average of about $950 million per year. Useful as a comparison tool, sure. It is not a forecast of recognized revenue, operating profit, or cash available to shareholders in any given twelve-month window.
An operator cannot simply take the full lifetime figure and hand it to today’s contractor. Someone has to advance money or a lender has to agree to finance against future payments. Borrowing converts part of the expected cash flow into construction funding, but it also attaches interest and repayment obligations. The commercial agreement also decides who supplies the actual computing hardware. A facility owner leasing already-equipped space carries a different purchasing burden than a cloud-style provider that must acquire the GPUs, servers, and networking gear itself.
How Credit Support Actually Works
Credit support from a large technology company sounds like a blank check. It is not. The relevant backstop typically activates only when the last data hall under the corresponding lease begins operating. Following specified payment defaults or insolvency events, the supporter can assume the lease or, in certain circumstances, permit termination and pay a declining fee. That fee has limits linked in part to project financing and can come with an interest in the landlord subsidiary.
In plain language, the protection concerns a customer failing to meet covered lease obligations. It does not promise reimbursement of every cost the developer incurs before the buildings are finished. A lender examining a construction loan will still ask which event turns the support on. If a contractor needs payment before the guaranteed rent obligation becomes effective, the financing package must bridge that gap somehow.
This distinction has real consequences for the price of debt. Protection against a tenant’s eventual failure can improve a lender’s expected recovery without removing worries about procurement delays or whether the borrower can complete the asset in the first place. One chief financial officer involved in these deals has called the cost of capital fundamental to infrastructure development. Stronger customer credit support can help a company obtain more competitive borrowing terms, and the math is straightforward.
On a hypothetical $3.5 billion average debt balance, dropping the rate from 9 percent to 8 percent saves $35 million a year before fees. Over five years that constant saving totals $175 million without any discounting. Actual savings depend on how quickly the debt is drawn and repaid, whether the rate floats, and the cost of arranging the support itself. The strongest argument for these agreements is concrete: a credible guarantor can improve financing terms and make an otherwise difficult capital requirement more manageable.
A Different Path: Customer Prepayments Plus Equipment Loans
Other companies with Bitcoin mining roots are solving the same problem with a different structure. One operator combined a large customer prepayment with loans secured against GPUs and contracted payments. The original multi-year agreement was valued at $9.7 billion and included a 20 percent customer advance. Twenty percent of that figure is $1.94 billion available before every contracted service has been delivered.
Later the same company closed GPU financing of $3.65 billion. Combined with customer prepayments, the package covered $5.59 billion of a $5.81 billion GPU capital requirement. Roughly 96 percent of that particular equipment bill sat inside the described combination. The residual three or four percent still had to come from other resources, but the structure answers the construction question in a way a pure headline contract value never can.
Some cash arrives from the customer before service begins. Some arrives from institutions lending against equipment and contracted cash flow. The rest comes from whatever other sources the company can access. Customer advances still carry a commercial obligation, of course. The provider must deliver the service for which the money was paid, and the agreement governs what happens if delivery slips or changes.
An advance should never be treated as pure profit received before the work starts. If a customer pays $100 million toward future services, the operator has more cash on hand but still owes the agreed performance. Equipment financing adds another layer: the lender can take security over assets that are expensive to acquire and can change in residual value. The practical question becomes whether the repayment schedule is supported by the customer receipts attached to those same assets.
Why Two Financing Percentages Can Both Be Correct
One announcement presented both a 6 percent blended cost of debt and a 3.31 percent average financing cost. The lower figure included the customer prepayment as a zero-interest funding source. Both numbers describe the company’s calculation, yet they rest on different funding bases.
A simplified example shows the difference clearly. Borrowing $60 at 6 percent produces $3.60 in annual interest. Combining that loan with a $40 advance that carries no interest puts the same interest bill over $100 of total funding, producing a blended 3.6 percent. The lender still charges 6 percent on the actual loan. The smaller blended number describes the overall package, not the rate available if the company had to borrow every dollar.
Fees matter too. A one-percent upfront fee on a $1 billion facility is $10 million before any periodic interest starts. If that facility is used for only one year, the fee carries a different annual weight than the same fee spread across five years. Comparing the coupon alone can hide that difference. Investors comparing financing announcements therefore need consistent units: interest on debt, cost across all funding sources, fees, and the period over which balances remain outstanding.
Construction Timelines Dictate When Cash Starts Flowing Back
Phased ready-for-service dates for one of the larger projects currently run from late 2027 through the second quarter of 2028. Those are planned delivery windows, leaving a substantial period between the credit-support announcement and a fully operating asset. A six-month delay carries an identifiable financing cost in a simplified case. If $1 billion remains outstanding at 8 percent for the extra half year, the additional interest alone reaches $40 million before any other delay expenses.
The effect changes if the facility has not yet been drawn or if the customer begins paying before final completion. The actual documents determine which party finances a delay and whether receipts continue during it. A construction budget also needs a clear distinction between work already completed, committed purchases, and spending that remains uncontracted. Cash paid to a supplier reduces the bank balance while securing equipment that may arrive months later.
Contractors, suppliers, and utilities often operate on separate payment milestones. A developer may need to fund several of them before the customer accepts a completed hall. Guaranteed maximum pricing arrangements can improve cost visibility, yet their scope matters. Equipment changes requested by a customer or work outside the covered package may be treated differently under the contract.
Permitting adds another layer of project-specific risk. Existing approvals and the precise terms of any state-level restriction determine which work is affected. An operator announcing access to power still has to explain the form of that access. A completed connection, a utility commitment, and an early development opportunity all carry different remaining construction requirements. Customers paying for reliable computing capacity need an operating service that meets the agreed specification. The value of a site’s electricity connection is therefore tied directly to the additional work needed to turn it into usable capacity.
Bitcoin Treasuries as a Funding Source with Limits
Selling Bitcoin remains one possible source of development cash, yet there is no automatic requirement that every AI project lean on it. Customer advances, secured borrowing, and equity issuance can cover some or all of a given company’s need. The arithmetic becomes more demanding when the coin price falls. Raising $100 million at a hypothetical $100,000 per Bitcoin requires 1,000 coins. At $60,000 the same dollar amount requires roughly 1,667 coins before transaction costs. That difference of nearly 667 Bitcoin for an identical invoice is not trivial.
Operating mining receipts introduce another variable. A miner can sell newly produced coins as part of ordinary cash management without reducing a previously accumulated treasury by the same amount. New equity carries a different cost from selling existing Bitcoin. It can preserve corporate cash while dividing future earnings and assets among more shares. A company with 100 million shares that issues another 20 million increases its share count by 20 percent. An existing holder’s fraction falls by one-sixth if the holder buys none of the new shares, although the company receives the proceeds and may create value with them.
Evaluating that transaction requires both the price of the new shares and the return on the assets they fund. Dilution alone does not prove value destruction, just as preserving the Bitcoin treasury does not automatically prove wise financing.
Cash Flow After Financing Is What Shareholders Ultimately Care About
A customer can pay every invoice on time while shareholders still earn less than expected if construction costs or financing charges absorb more of the proceeds. Contract performance and equity return are related, but they are not the same measure. Consider an illustrative project generating $200 million annually after operating costs but before financing. If annual interest and principal payments total $150 million, the remaining cash is $50 million before taxes and other obligations. If the same project also requires $50 million of recurring equipment spending, little may remain for distributions despite a positive operating result.
Removing principal repayments from that calculation would make the apparent surplus look much larger. Replacing principal with depreciation would answer an accounting question without showing the same cash requirement. A long lease can support borrowing precisely because payments are expected to continue for many years. Financing the asset allows the operator to spread the initial cost across those receipts. The favorable case is strongest when contracted payments exceed operating and financing expenses by a comfortable margin, construction stays inside budget, and the asset begins service on schedule. A customer advance can further reduce the cash shareholders must supply up front.
Credit support strengthens that case when it improves the amount or the price of available funding. A financing discussion that treats guarantees only as evidence of weakness would miss the practical reason developers seek them. The unresolved question remains the return after the entire funding arrangement has been paid for. Announced capacity and lifetime revenue do not supply every input needed to calculate it.
Avoiding Double-Counting in Funding Announcements
Multiple funding routes can appear inside a single development plan. Prepayments, convertible notes, GPU leasing, and traditional equipment financing may all be listed over a short period. Adding every headline number can double-count money if an earlier facility is later refinanced. A hypothetical $1 billion bond used to repay an $800 million construction loan produces only $200 million of new funding before fees. Reporting $1.8 billion as simultaneously available would count the same project financing twice.
The same discipline applies to cash that is restricted for a particular project. A bank balance reserved under a financing agreement cannot automatically be assigned to another campus without checking the conditions attached to its use. A useful funding reconciliation starts with unrestricted cash, adds committed and drawable financing, subtracts repayments and near-term obligations, and compares the result with remaining spending. It keeps equipment financing separate where the proceeds cannot be used for general construction.
None of those calculations requires assuming that AI demand will suddenly disappear. Even a fully booked project can encounter a mismatch between the dates bills fall due and the dates customer receipts become available. Conversely, evidence of funded construction, accepted facilities, and timely customer payments would weaken concerns about that mismatch. Those milestones are observable. Treating every large AI contract as either guaranteed profit or pure speculation is not.
What to Watch Going Forward
The remaining disclosure questions are specific. How much capital is still needed, from whom, on what terms, and before which delivery date? A company that answers those points lets investors compare the next announcement against the cash requirement already on record. Several practical items deserve ongoing attention:
- The amount actually borrowed versus the maximum facility size, along with repayment schedules, fees, and collateral attached to each campus
- Cash received from customers before service begins, the specific bills it funds, and the delivery obligations that remain
- The exact commencement conditions, covered defaults, and termination-payment provisions inside any credit support agreements
- Procurement progress, approved budgets, accepted data halls, and any shifts to the published construction timetable
- New equity issuance, convertible obligations, and disclosed Bitcoin sales measured against the cash they actually deliver
I have found that the most useful analysis keeps the lifetime contract value, the expected service dates, and the remaining spending requirement in the same frame. Comparing only the contract size with a company’s market capitalization leaves out the capital that must still be supplied to earn those payments. The difference between a signed agreement and an operating, cash-generating facility is measured in years of interest, construction risk, and careful sequencing of advances and draws.
Perhaps the most interesting aspect of this entire transition is how ordinary the funding mechanics actually are once the crypto branding is set aside. Customer advances, secured equipment loans, and credit support for long-term leases appear in many capital-intensive industries. The novelty here is the speed with which former Bitcoin mining companies are adapting those tools to AI demand. Whether the returns ultimately justify the capital depends on execution details that will become clearer only as the first halls come online and the first customer payments begin to arrive on schedule.
Until then the headline numbers will continue to attract attention. The quieter question of who pays for construction, and on what terms, will decide how much of that attention translates into lasting value for the companies and their shareholders. The answer is already visible in the structure of the deals themselves if you look past the total commitment and focus on the cash that actually moves while the concrete is still wet.