FOURWEEKMBA TRACKER ยท FROM THE NEWSROOM ยท UPDATED 4 NOVEMBER 2026
Free to share with a link to this page. Figure: Financial Times estimate as reported; the multiple is our arithmetic.
The short version: AI is now built on borrowed money. Big Tech borrows through bonds and loans (Microsoft $25B, Amazon $17.5B, Oracle $10B), AI labs borrow against the chips they lease (Anthropic’s $60B Project North Star), and data-centre developers take loans backed by the AI companies that rent their buildings. This page tracks 21 AI borrowings: who borrows, the instrument, the amount, the lenders and what it costs, each linked to its source. The 14 with a stated amount, excluding deals still in talks, add up to $158.5B by our sum.
The money is getting dearer and some lenders are pulling back: the hedge fund Arini sold about $800 million of debt after margin calls, the story behind this page (Hedge Fund Arini Sold $800M of Debt After Margin Calls). Amounts are as filed or reported; non-dollar deals are shown in their own currency and left out of the sum. Data as of 4 Nov 2026.
17 of 21 rows show an amount in their filing or report; “โ” means not disclosed. Status: priced, filed, closed, drawn, signed or syndicated = done; reported = an outlet’s account; in talks or seeking = not done.
What AI debt costs
The same market prices AI borrowers very differently. A neocloud with Microsoft or Meta as the customer borrows at investment-grade rates; one whose customers are less certain pays double digits.
Who is pulling back
Arini Capital Management (hedge fund). Sold approximately $800 million in debt holdings in recent weeks after margin calls from banks (WSJ). The Wall Street Journal ยท our newsroom
Pimco. Lent billions of dollars into the AI debt boom in 2025, then cut positions before AI bond prices fell (Bloomberg). Bloomberg ยท our newsroom
The market around these deals
AI-linked companies raised an estimated $466bn of debt in 2026 so far, up from $101bn in the same period of 2025 (Financial Times). Financial Times, via our newsroom
Cloud-computing giants and data-centre developers issued nearly $368bn of bonds in 2026 so far, more than 6x the same period last year (Financial News, Korea). Financial News (Korea) ยท our newsroom
Goldman Sachs projects $420 billion of global hyperscaler debt issuance in 2027 and roughly 35% of their capex funded by debt. Axios ยท our newsroom ยท www.goldmansachs.com
Top-rated data-centre bonds yielded below 5% at the start of 2026 and closer to 7% by October (Bloomberg, citing Morgan Stanley). Bloomberg ยท our newsroom
How AI debt financing works
Corporate bonds.Microsoft, Oracle, Alphabet and Amazon borrow on their own balance sheets at investment-grade rates. The money funds data centres and chips; the bonds do not name a project.
Project and asset-backed loans. A special-purpose company owns the chips or the building and borrows against them. Lenders are repaid from the rent a lab or hyperscaler pays, so the tenant’s credit sets the price.
Guarantees. A stronger company (Broadcom in Project North Star, Nvidia in talks with Nscale) guarantees part of the debt so lenders accept a lower rate.
Private credit and leveraged loans. Funds such as Blackstone and Apollo, and banks that sell loans on to investors, take the riskier slices at higher coupons.
Questions
Who is borrowing the most to build AI?
Among the deals we track with a stated amount, Anthropic (Project North Star) at $60B is the largest, followed by SpaceX (Nvidia GPU financing) ($10B loans + $30B bonds, still in talks) and Microsoft ($25B).
Yes, by the measures we track: top-rated data-centre bonds moved from below 5% at the start of 2026 to close to 7% by October (Bloomberg, citing Morgan Stanley), and CoreWeave’s 2032 bonds yielded 14.8% in mid-October (CNBC).
Built by FourWeekMBA from our newsroom’s database of sourced facts (each row lists its sources), the filings those facts link to and reports by the outlets named. Sums and multiples are our arithmetic. A missing amount is shown as “โ” and never estimated. New borrowings are added as the newsroom reports them.
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.
