--- title: "AI Debt: How the Build-Out Is Financed" slug: ai-debt-financing type: data-index sector: finance canonical_url: https://deepstoryresearch.com/data/ai-debt-financing series: [issuance_by_issuer, issuance_trend, projections] series_count: 3 data_point_count: 11 data_as_of: 2026-07 source_quality: curated_snapshot tier: free generated_at: 2026-08-11 --- # AI Debt: How the Build-Out Is Financed _How the AI build-out is financed — not how much it costs_ > Deepstory Research context file · **Free tier** · > Self-contained AI briefing. Drop into an LLM window or RAG pipeline. ## AI research use contract - Treat this file as source-grounded context, not a live database. Quote `source_ref` and `data_as_of` when using numbers. - Separate `FACT`, `ESTIMATE`, `FORECAST`, `TARGET`, and `INFERENCE`. If a row basis is unclear, say it is unclear. - Prefer T1/T2 sources for hard facts; use T3/T4 sources as context that should be checked before money, legal, medical, operational, or policy decisions. - Keep answer structure clean: first say what the data says, then what you infer, then what would change the conclusion. ## Research upgrade checklist - Separate public bond issuance, private credit, leases, and off-balance-sheet structures — they are not additive. - Track maturity schedule, spread, issuer rating, and whether proceeds are explicitly data-centre linked. - Distinguish refinancing from new money; 2026 issuance is the latter. - Add a stress prompt for refinancing risk if AI revenue growth lags debt-funded infrastructure growth. ## TL;DR — index summary The five largest US tech issuers sold $159bn of bonds in the first five months of 2026 — a 47% jump on the $121bn they issued across all of 2025, and new money against data-centre construction rather than a refinancing wave. Sell-side projections put full-year hyperscaler issuance at $250–300bn and global AI-related debt near $570bn. ## What this index covers Three series: 2026 issuance by issuer, 2025 full-year versus 2026 year-to-date, and where projections land. **Series in this index:** - `issuance_by_issuer` — Who is raising the debt. - `issuance_trend` — 2025 full year versus 2026 year-to-date. - `projections` — Where sell-side expects AI debt to land. **Entities tracked:** Amazon, Alphabet, Meta, Microsoft, Oracle, Morgan Stanley. ## Key findings - Amazon and Alphabet together account for well over half the $159bn five-issuer total (cryptobriefing-159b). - $159bn in five months of 2026 against $121bn for all of 2025 — a 47% increase (yahoo-ai-bonds). - Morgan Stanley projects $250–300bn hyperscaler issuance and ~$570bn global AI-related debt for 2026 (forbes-570b). - A further ~$800bn is estimated to sit in private credit and off-balance-sheet data-centre structures (marketplace-ai-debt). ## Evidence basis map The free tables above show `source_ref` but not the row-level `value_basis`. This section mirrors the visible rows only, so AI tools can separate observed values from estimates, forecasts, targets, and derived values. | series | row | source_ref | value_basis | | --- | --- | --- | --- | | issuance_by_issuer | issuer=Amazon; usd_b=57 | cryptobriefing-159b | Amazon ~$57bn of bond issuance, 2026 through early June | | issuance_by_issuer | issuer=Alphabet; usd_b=52 | cryptobriefing-159b | Alphabet ~$52bn of bond issuance, 2026 through early June | | issuance_by_issuer | issuer=Meta; usd_b=30 | cryptobriefing-159b | Meta $30bn bond offering, 2026 | | issuance_by_issuer | issuer=Oracle; usd_b=18 | cryptobriefing-159b | Oracle raised $18bn, 2026 | | issuance_trend | period=2025 full year; usd_b=121 | cryptobriefing-159b | Five largest US tech issuers sold $121bn across all of 2025 | | issuance_trend | period=2026 to early June; usd_b=159 | cryptobriefing-159b | Same five issuers sold $159bn through early June 2026 — a 47% jump on full-year 2025 | | projections | label=Issued YTD 2026 (AI-linked); usd_b=250 | yahoo-ai-bonds | AI-linked bond issuance tops $250bn, 2026 YTD | | projections | label=Morgan Stanley FY2026 low; usd_b=250 | yahoo-ai-bonds | Morgan Stanley expects $250–300bn of 2026 hyperscaler issuance (low end) | | projections | label=Morgan Stanley FY2026 high; usd_b=300 | yahoo-ai-bonds | Morgan Stanley expects $250–300bn of 2026 hyperscaler issuance (high end) | | projections | label=Global AI-related debt, end-2026; usd_b=570 | forbes-570b | Global AI-related debt projected to approach $570bn by end-2026 | | projections | label=Private credit / off-balance-sheet (est.); usd_b=800 | marketplace-ai-debt | ~$800bn estimated in private-credit and off-balance-sheet data-centre structures — ESTIMATE, not an issuance tally | ## The series (per-chart briefings) ### Issuance by issuer — hbar **Direct answer:** Who is raising the debt. **Time bracket:** 2026 YTD | issuer | usd_b | source_ref | | --- | --- | --- | | Amazon | 57 | cryptobriefing-159b | | Alphabet | 52 | cryptobriefing-159b | | Meta | 30 | cryptobriefing-159b | | Oracle | 18 | cryptobriefing-159b | **Read:** Amazon added a further $25bn in July, so its full-year total runs above the figure shown. **Caveat:** Microsoft is inside the $159bn total but no standalone figure is published, so that bar is omitted rather than derived by subtraction. ### Issuance trend — bar **Direct answer:** 2025 full year versus 2026 year-to-date. **Time bracket:** annual | period | usd_b | source_ref | | --- | --- | --- | | 2025 full year | 121 | cryptobriefing-159b | | 2026 to early June | 159 | cryptobriefing-159b | **Read:** Five months of 2026 exceeded all of 2025 by 47%. ### Projections — hbar **Direct answer:** Where sell-side expects AI debt to land. **Time bracket:** FY2026 | label | usd_b | source_ref | | --- | --- | --- | | Issued YTD 2026 (AI-linked) | 250 | yahoo-ai-bonds | | Morgan Stanley FY2026 low | 250 | yahoo-ai-bonds | | Morgan Stanley FY2026 high | 300 | yahoo-ai-bonds | | Global AI-related debt, end-2026 | 570 | forbes-570b | | Private credit / off-balance-sheet (est.) | 800 | marketplace-ai-debt | **Read:** Issued-to-date, a hyperscaler-only range, and a broader global figure. **Caveat:** These three measures are NOT additive. The ~$800bn private-credit figure is an estimate of off-balance-sheet structures, not an issuance tally, and is the least observable number on the page. ## Cross-series synthesis — why this matters This index answers how the build-out is financed, which is a different question from how much is being spent — capex guidance is covered by ai-capex-vs-defense. The distinction matters because debt-funded capex transfers risk from equity holders to bondholders and, in the private-credit tranche, to holders who are not publicly identifiable. Read issuance_trend against the projections chart: the run-rate is already consistent with the low end of the sell-side range, which means the projection is largely an extrapolation of an observed pace rather than an independent forecast. ## Entities & relationships | Entity | What they do | Key stat | Relationship | | --- | --- | --- | --- | | Amazon | Largest issuer | $57bn 2026 YTD, plus $25bn in July | Sets the pace of the issuance wave | | Alphabet | Major issuer | Second-largest of the five | Together with Amazon, over half the total | | Morgan Stanley | Sell-side projection | $250–300bn hyperscaler, ~$570bn global | Source of the forward numbers | ## Timeline of key events & decisions - **2025** — $121bn issued across the full year → the baseline - **Jan–Jun 2026** — $159bn issued in five months → the step change - **Jul 2026** — Amazon adds a $25bn package → shows the run-rate continuing past the charted window ## Cross-industry ripple - **Corporate bond market:** A single sector supplying a large share of high-grade issuance compresses spreads elsewhere and concentrates duration risk. _(inference)_ - **Private credit:** The ~$800bn off-balance-sheet estimate overlaps directly with the definitional dispute in private-credit-growth. _(inference)_ - **Utilities / energy:** Debt raised against data centres pre-commits load growth tracked in ai-power-demand. _(inference)_ ## Non-obvious reads (interpretation) _This section is interpretation, not sourced fact — each item names the observation it is built on and a confidence level._ - **Observation:** The least observable tranche is the largest single estimate on the page. **Read:** If ~$800bn genuinely sits in private credit and off-balance-sheet structures, then the publicly visible bond market — the part that gets priced daily — is the minority of AI financing risk, and the majority has no continuous price signal. _(confidence: low)_ ## Glossary / key terms - **Off-balance-sheet** — Financing structured so the obligation does not appear on the sponsor's balance sheet. - **Private credit** — Non-bank lending arranged bilaterally rather than issued into public markets. - **Hyperscaler** — The largest cloud/data-centre operators — here Amazon, Alphabet, Meta, Microsoft, Oracle. ## How to use this with AI Paste this file into an LLM context window (or a RAG store) and ask cross-series questions. The tables carry a `source_ref` per row; the source registry below maps each ref to a named source and a trust tier. ### Suggested prompts (multi-series) ```text Using issuance_trend and projections, stress-test refinancing risk if AI revenue growth lags the debt-funded build-out. ``` ```text Explain why the three bars in the projections chart cannot be summed. ``` ## Sources & trust | ref | source | trust tier | url | | --- | --- | --- | --- | | yahoo-ai-bonds | Yahoo Finance / PitchBook — Bond issuance backing AI investment tops $250B | T3 · Reputable press / research org / OWID | https://finance.yahoo.com/technology/ai/articles/bond-issuance-backing-ai-investment-135635215.html | | cryptobriefing-159b | Alphabet, Amazon, Meta, Microsoft, Oracle issue record $159B in bonds for AI buildout | T3 · Reputable press / research org / OWID | https://cryptobriefing.com/tech-giants-record-159b-bonds-ai/ | | forbes-570b | Forbes — Bond investors push back as AI debt heads toward $570 billion | T3 · Reputable press / research org / OWID | https://www.forbes.com/sites/robertszczerba/2026/07/17/bond-investors-push-back-as-ai-debt-heads-toward-570-billion/ | | marketplace-ai-debt | Marketplace — AI debt is flooding the bond market | T3 · Reputable press / research org / OWID | https://www.marketplace.org/episode/2026/07/21/ai-debt-is-flooding-the-bond-market | ## Caveats & what this index cannot answer - Microsoft's standalone 2026 issuance — inside the total but not separately published. - The true size of the private-credit tranche; the ~$800bn figure is an estimate, not a tally. - Whether issuance is matched to revenue — this index measures financing, not returns. - Estimates and forward targets are labelled in the working dataset; never read a labelled estimate or forecast as a settled figure. - This is an observational data index, not investment, legal, or medical advice. ## Data freshness & methodology - **Last updated:** 2026-08-11 - **Data as of:** 2026-07 - **What changed most recently:** Amazon added a $25bn package in July 2026, above the charted year-to-date figure. **Methodology (as seeded):** > This index tracks how the AI build-out is FINANCED, which is a > different question from how much is being spent. Capex guidance is > covered separately by the ai-capex-vs-defense index. > > Issuer figures are 2026 corporate bond issuance through early June, > as reported for the five largest US tech issuers. Microsoft is inside > the $159bn five-issuer total but no standalone figure is published in > the cited source, so that bar is OMITTED rather than derived by > subtraction. > > CAVEAT: the projections chart mixes three different measures — > issuance to date, a hyperscaler-only sell-side range, and a broader > "global AI-related debt" projection. They are not additive. The > ~$800bn private-credit figure is an ESTIMATE of off-balance-sheet > data-centre structures, not an issuance tally, and is the least > observable number on the page. > > Scout run: Trending_Candidates/2026-08-06.md. Re-verified 2026-08-10. --- _Free tier. The full row-level dataset and per-source detail live on the live page and in the working file. Canonical: . Deepstory Research · https://deepstoryresearch.com_