Q3 Concentration: Amazon Still ~39% of Hyperscaler IG — Top-3 Holds 74%

Absolute AI financing dollar totals answer how large is the credit stack. They do not answer how concentrated is the system at the top of the distribution. That is the capital-markets question this mid-Q3 concentration lens is built for.

Our theme already maps the vintage path. The July research put hyperscalers on a Goldman path toward ~$250B / ~33% of AI capex funded with IG bonds. The mid-year update answered the theme perimeter question: AI-related debt near $489B, with the five names only ~40% of that stack and AI's share of US IG supply near 18%. The Q3 supply-share refresh pushed theme weight to ~23% of USD IG and ~20% of HY, and quantified private data-centre transactions near $200B since early 2025. The late-Aug concentration companion then answered the distribution question on the Aug YTD spine (~$194B, Amazon ~41% / top-3 ~76%). This post re-asks the same core question on a mid-Q3 YTD vintage — so shares and levels stay on one tape as issuance continues toward the FY path.

The interactive dashboard above is a concentration lens, not another KPI strip. Toggle Issuer ladder, Theme stack, Credit channels, and ETF + lenses. Top-1 share is 39% (Amazon, ~$85B of ~$218B). Top-3 share is 74% (Amazon + Alphabet + Meta). Approximate five-name HHI is ~2,486. Absolute YTD rose from the Aug $194B print; top-1 eased from 41% as peers printed into Q3 — larger oligopoly dollars, slightly less extreme top-1 share.

The Q3 concentration scoreboard

Read the table as a family of market shares, not one number. Issuer concentration inside the hyperscaler cohort remains thick. Theme-stack concentration is milder once utilities, DC JVs, and ecosystem credit enter the perimeter — and the HS bloc share actually rose slightly (~40% → ~42%) as the YTD spine grew faster than the mid-year ecosystem estimate. Channel-stock concentration stays extreme at the senior sleeve even after rebasing for private-DC overflow. ETF concentration remains a sentiment story, not an issuer-proceeds story.

Issuer concentration: Amazon still owns the thick end

Filter the dashboard to Issuer ladder. The mid-Q3 cumulative curve for the five-name YTD universe rises to about 39% at top-1, 59% at top-2, and 74% at top-3. Approximate HHI on five buckets sits near 2,486 — still well above an equal five-way split (~2,000). Amazon's disclosed multi-tranche USD/EUR prints keep it the modal top-1; Alphabet and Meta fill the next rungs; Oracle's releveraging path and Microsoft's still-OCF-heavy posture close the ladder.

The Aug → Q3 → FY path panel is the vintage delta that the late-Aug companion could not show. Absolute YTD climbed $194B → $218B toward the ~$250B FY path. Top-1 eased 41% → 39%; top-3 eased 76% → 74%. That is not diversification in the industrial sense — three names still clear three-quarters of hyperscaler IG. It is peer catch-up inside a still-steep ladder as Alphabet, Meta, and Oracle kept printing into Q3.

Pair this panel with the research post's funding-mix story: debt is still only about a third of hyperscaler capex on Goldman's path. Concentration of issuance can coexist with concentration of cash-flow funding — Amazon can dominate the bond calendar while the cohort as a whole still funds the majority of servers from operating cash flow.

Theme stack: thick inside, still a minority outside

Switch the view to Theme stack. Of the mid-Q3 AI-related debt perimeter (~$520B, up from the Aug $489B mid-year print), hyperscaler IG is about $218B / 42% and the broader ecosystem is about $302B / 58%. The five-name bloc is concentrated internally (top-1 ~39%) but is still a minority of the theme once utilities, industrials, data-centre JVs, HY, loans, and private-DC overflow enter the tally.

That is the hinge the August mid-year update introduced and this Q3 concentration lens sharpens. Desks that underwrite "AI credit = Mag 7 bonds" are reading the thick end of a minority sleeve. Desks that underwrite "AI credit is everywhere" without noting issuer ladders miss how much primary-calendar risk still sits in three tickers.

The supply-path panel underneath shows calendar concentration rising even when issuer ladders stay sticky: AI's share of US IG gross supply climbed from ~1% in 2024 to ~7% in 2025, ~18% at the August desk print, and ~23% in the Q3 refresh — with HY near 20%. Theme weight inside the calendar can concentrate without any single issuer matching Amazon's inside-cohort share. The Q3 update also flags the rating-vs-market gap (Meta and Amazon long-dated paper pricing below official AA) — a pricing concentration of supply technicals, not a default narrative.

Credit channels: senior IG still near half — private DC now counted

Open Credit channels. Rebasing the Booth/Hepp stock map with Q3 private-DC overflow puts funded AI-infra credit near $1.15T (from ~$1.065T in the late-Aug companion). HS senior unsecured alone is about $540B / 47%. Project / data-centre finance (~$260B / 23%) and private credit + DC overflow (~$230B / 20%, including the ~$200B private-DC deals quantified since early 2025) push the top-3 channel share to ~90%. ABS and GPU-secured sleeves remain thin residuals.

This is a different concentration story from issuer YTD flow. Flow asks who printed paper this year. Channel stock asks where the outstanding claims sit. A market can look diversified on monthly deal tapes while the stock remains dominated by senior unsecured held by funds, insurers, and pensions. The Q3 twist is that private DC is no longer a dry-powder abstraction — it is a quantified overflow sleeve that raises the funded perimeter without dethroning the senior IG top-1.

Treat the channel ladder as a funded-stock index, not a stress loss forecast. Uncommenced lease overhang still sits outside funded totals — a parallel concentration of off-balance commitments that this lens flags but does not fold into the 47% / 90% prints.

ETF flows: half the thematic sleeve is still one ticker

Toggle ETF + lenses. FactSet's 2025 US thematic ETF inflow sleeve was about $43.5B. QQQ alone absorbed ~$21.7B — roughly 50%. Semiconductor proxies (SOXX-scale) add another ~20%; narrow robotics/AI thematics remain a thin wedge; the residual thematic sleeve fills the last quarter.

Equity-side concentration is a sentiment-capacity meter. Creations into QQQ are not issuer proceeds for Amazon's data centres; they are secondary ownership of a Mag 7–heavy Nasdaq-100 proxy. Still, for the theme's public-markets question — how is the build-out funded in credit and public markets? — the answer remains asymmetric. Credit funding concentrates in a handful of IG issuers and a senior channel stock. Equity flow concentration concentrates in one broad proxy that prices the same names.

The cross-lens scatter makes the asymmetry visual: issuer and channel lenses sit high on both top-1 and top-3 axes; the AI-debt bloc and IG-calendar lenses are high on top-1 but lack a meaningful top-3 ladder because they are single-bloc / theme prints; ETF flows sit near 50% / 75%.

Who is exposed — and what would change the story

Exposed: IG portfolios that treat "hyperscaler" as five equal credits when three names still clear ~74% of mid-Q3 YTD issuance; desks that size AI theme risk off the $520B perimeter without noticing the 42% HS minority and the 39% Amazon inside share; insurers and pensions whose AI-infra exposure is mostly the $540B senior sleeve and therefore inherit issuer-ladder concentration whether or not they own project-finance paper; equity allocators who read thematic ETF creations as diversified AI exposure when half the sleeve is QQQ; credit desks that ignored private-DC overflow until the Q3 quantification made the channel visible.

Relative winners under current rules: mega-issuers that can still clear multi-tranche books inside AA/A technicals (even when long-dated paper prices through the official rating); intermediaries that intermediate the overflow into project finance and private credit as IG calendars saturate toward 23% theme weight; holders of the senior unsecured sleeve while lease overhang and GPU-secured risk sit elsewhere in the capital stack; broad-proxy ETF complexes that capture Mag 7 ownership demand without needing narrow AI wrappers.

What would change the story: a sustained flattening of the issuer ladder so top-1 falls below ~25% of hyperscaler IG; ecosystem debt growing so fast that the HS bloc share of the AI perimeter falls well below 30% and issuer concentration inside the bloc breaks; a funded-stock mix where project + private credit displace HS senior below ~35%; thematic ETF flows dispersing so no single ticker holds more than ~30% of the sleeve. Mid-Q3 shows absolute dollars rising and top-1 easing slightly — not breaking.

Caveats and methodology

• Issuer shares inside the five-name mid-Q3 YTD spine are estimated from disclosed Reuters/LSEG deal prints and close to a ~$218B desk total on the path toward Goldman's ~$250B FY. Treat top-1 / top-3 as order-of-magnitude concentration, not a prospectus table.
• AI debt perimeter (~$520B) mixes IG, HY, loans, ecosystem credit, and private-DC overflow; the 42% HS share is a bloc share, not an issuer ladder. The prior Aug mid-year print was ~$489B / 40%.
• IG/HY calendar weights (~23% / ~20%) are theme shares of gross supplynot top-1 issuer shares of the calendar.
• Channel stock (~$1.15T) rebases Booth/Hepp with Q3 private-DC quantification; private-credit and some project sleeves are estimated and can overlap deal tallies.
• Uncommenced leases remain excluded from funded-stock concentration by design.
• ETF flows are net creations, not AUM, and QQQ is a Mag 7 proxy rather than a pure AI product.
• Equity raises (e.g. Alphabet's mid-2026 ~$85B print) are excluded from the IG issuer ladder on purposethey concentrate funding without concentrating bond supply.
• Rating-vs-market gaps (Meta/Amazon long-dated paper) are pricing technicals from the Q3 update, not implied default probabilities.

The shareable takeaway

AI financing remains concentrated at the top — and mid-Q3 mostly moved the absolute dollars, not the shape. Inside hyperscaler IG, top-1 (Amazon) is about 39% (from 41% at the Aug print) and top-3 about 74% of a larger ~$218B YTD spine. Of the ~$520B AI debt perimeter, the five-name bloc is still only ~42%. Of funded credit stock, HS senior unsecured alone is ~47% and the top-3 channels clear ~90%, with private DC now a quantified ~$200B overflow. Of thematic ETF inflows, QQQ still takes ~50%. The build-out is funded in credit and public markets through a system that looks diversified in press-release counts and top-heavy once you rank the distribution.

Related reading: Late-Aug concentration companion · Q3 supply-share refresh · AI financing research 2026 · Mid-year $489B update.