Bloomberg reported on 5 November that the Bank of England “has queried market participants about data center projects that are being financed without a tenant in place to pay rent.” The questioning is part of wider efforts to understand lending to the industry, according to people with knowledge of the matter.
The Bank’s July 2026 Financial Stability Report had already discussed related risks. It carries Morgan Stanley estimates in which debt sources add up to $1.75 trillion of data centre funding for 2026 to 2028, about 54.5% of the total, by our arithmetic.
Business Pill · Take-or-Pay
An evergreen Business Engineer explainer on take-or-pay contracts, the kind of signed commitment a tenant gives a lender. It teaches the general idea only and does not discuss the Bank of England or any lender named here.
The key insight: Bloomberg reports, citing people with knowledge of the matter, that the Bank of England has queried market participants about data centres financed without a tenant in place to pay rent; the Bank’s July report shows Morgan Stanley estimates in which debt sources add up to $1.75 trillion of data centre funding for 2026 to 2028, about 54.5% of the total, by our arithmetic.
Bloomberg’s report, citing people with knowledge of the matter, says the questions concern projects financed before an occupier has agreed to lease them. Bloomberg explains that “Loans to projects that start construction before an occupier agrees to lease the completed project are known as speculative financing in the real estate industry.”
We read the headline and first two paragraphs of Bloomberg’s article. In what we read, the report does not name the lenders that were asked, or give a figure for their exposure.
Morgan Stanley estimates of funding sources for global data centre capital expenditure, 2026 to 2028, as shown in the Bank of England’s July 2026 Financial Stability Report (Chart 2.2). The debt total and shares are our arithmetic.
What the Bank Has Already Published
The Bank’s July Financial Stability Report had already discussed AI debt. It says hyperscalers’ declining free cash flows are “leading to their growing use of off-balance sheet financing arrangements”, and that this is “broadening the AI debt footprint across leveraged, structured, and private credit.”
It adds: “The riskiness of this debt depends on the underwriting terms, in particular the quality of the leases and guarantees which back debt holders’ claims.” A data centre with no tenant has no lease behind it.
The report also flags a timing gap. It says most AI debt so far has funded data centre buildings, and that “frontier AI data centres can quickly become outdated if they are no longer able to support the latest chips.” How that risk falls “across lenders, borrowers or tenants, will depend on the terms of the debt agreements,” it says.
On the UK, the report says: “The UK has the largest data centre pipeline in Europe.” It says the Financial Policy Committee will report on its work on frictions to long-term investment in the November 2026 FSR.
The Funding Mix in the Bank’s Chart
Chart 2.2 of the report is titled “More than half of the external financing need for data centres from 2026 to 2028 could be funded via debt”. It shows Morgan Stanley estimates for global data centre capital expenditure: equity capital of $1.46 trillion, private credit of $700 billion, investment grade debt of $650 billion, ABS and CMBS of $200 billion, high yield debt of $150 billion and loans of $50 billion.
Adding the debt sources gives $1.75 trillion of a $3.21 trillion total, about 54.5%, by our arithmetic. Private credit is the largest single debt source, at 40% of the debt.
Why the Tenant Matters to the Rate
A signed tenant changes the cost of the money. In our 4 November explainer on how neoclouds work, IREN said its $3.65 billion facility of June was “Anchored by Microsoft’s offtake” and cost a blended 6.00%, while its August financing for GPUs at a site serving other customers carries a fixed 9.0%.
The two loans differ in date, site and structure, so this is an illustration, not a like-for-like comparison. Between 2 October and 3 November, IREN shares fell 41.4% and CoreWeave’s 40.0%, as our piece on neocloud stocks and CoreWeave bonds showed, and our Microsoft capacity analysis adds up to $70.9 billion of Microsoft contracts disclosed by three neoclouds, by that piece’s arithmetic.
Connecting the Coverage
Our related pieces on how the AI build-out is financed:
As we read the July report, the Bank had already named the variable that matters here: it says the riskiness of AI debt depends on the quality of the leases and guarantees behind it, and a project with no tenant has no lease to point to.
As we read the Morgan Stanley figures the Bank shows, private credit is the single largest debt source, at $700 billion of the $1.75 trillion, which is the part of the market the report says is taking on more AI debt through off-balance sheet structures.
As we read IREN’s filings, as set out in our explainer, the tenant also shows up in the price of money: a facility anchored by Microsoft’s offtake cost a blended 6.00%, against 9.0% for GPU financing at a site serving other customers, though the two loans differ in date, site and structure.
Bank of England, Financial Stability Report, July 2026
“The riskiness of this debt depends on the underwriting terms, in particular the quality of the leases and guarantees which back debt holdersโ claims.”
Three Implications
FOR LENDERS Bloomberg describes speculative financing as loans to projects that start construction before an occupier agrees to lease them, which is the category the Bank asked about.
FOR BORROWERS The July report says hyperscalers’ declining free cash flows are leading to growing use of off-balance sheet financing arrangements, including securitised and special purpose vehicle structures.
FOR THE UK The report says the UK has the largest data centre pipeline in Europe, and that the Financial Policy Committee will report on its work in the November 2026 FSR.
The framework says: “It is a sovereign-scale financing problem dressed in tech-company clothing — and the financial architecture is starting to invert.”
As we read it, a central bank asking lenders about data centres with no tenant is that financing question reaching the regulators. The framework predates this report and does not discuss it.
What Is Not Established
We read only the opening of Bloomberg’s article. In what we read, it does not say which firms were queried, how much speculative lending the Bank found, or whether any supervisory action follows. The $1.75 trillion is Morgan Stanley’s estimate of global funding, as shown by the Bank, not a figure for UK lenders. The 54.5% and 40% shares are our arithmetic.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
Bloomberg reports, citing people with knowledge of the matter, that the Bank of England has asked market participants about data centres financed without a tenant. The Bank’s July report already said the risk in AI debt depends on “the quality of the leases and guarantees” behind it, and its chart shows $1.75 trillion of estimated debt funding for 2026 to 2028, by our arithmetic.
A note on sourcing. Bloomberg’s report was read in its headline and first two paragraphs. The Bank of England passages and the Morgan Stanley funding estimates are from the Bank’s July 2026 Financial Stability Report. The IREN figures are from our 4 November explainer, drawn from IREN’s filings. The debt total and shares are our arithmetic.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
