Q3 Concentration: Cloud Top-3 Hits 58% — US Holds 54% of Pipeline Seats; Google Leads Tokens

Our 2026 concentration print answered the first distribution question: how thick is the tip across chip ownership, hyperscale capacity, regional power draw, and token throughput — Google 25% of H100e, Top-3 owners 54.8%, cloud Top-3 57%, US regions ~45% of AI DC capacity, and ByteDance leading tokens at 29.2%. The Q3 theme update then refreshed the tape with Synergy’s hyperscale site ledger (1,360 ops / 803 pipeline) and a Top-3 cloud print of 58%. This Q3 concentration lens asks the sharper follow-up desks actually trade: how concentrated is the system when you re-score Top-1 / Top-3 after the site ledger, watch the cloud tip tick, and see the token Top-1 flip?

The interactive dashboard above is built as a Q3 concentration lens. Toggle Scoreboard, Sites & cloud, Ownership, and Tokens vs chips. The punchline is deliberately multi-sided. On chip ownership, the tip is stuck: Google still ~25%, Top-3 still ~54.8%, Big-5 still 71.4% with within-Big-5 HHI ≈ 2,421. On hyperscale cloud capacity, AWS + Azure + Google Cloud print 58% — +1 pp versus the Aug ranking tape. On sites, the United States holds 437 of 803 hyperscale pipeline seats (54.4%) while ops seats are thinner (580 / 1,360 ≈ 42.6%). On tokens, the Q3 brand cohort puts Google at ~27% — ahead of ByteDance at ~21% — so the usage tip flipped even while the ownership tip did not move.

The Q3 concentration scoreboard

Read the table as a family of market shares, not one slogan. Ownership concentration is thick at the hyperscaler tip and still leaves nearly three-tenths of world H100e outside the Big-5. Cloud concentration is a capacity story among operators who rent to frontier labs. Site concentration says where new seats land on the Synergy hyperscale-only ledger — not the Aug large-DC mix of ~915 pipeline sites. Token concentration says who processes the prompts — and in Q3 it finally lines up with Google’s ownership lead, while ByteDance and OpenAI still sit far above their chip stock.

Ownership tip: still sticky, still not the whole story

Toggle Ownership. Epoch’s Chip Owners frame (Q4 2025 anchors, Aug explorer restatement) remains the carried print: Google near 25%, Microsoft 17.3%, Amazon 12.5%, Meta 11.3%, Oracle 5.3%. China as an aggregate owner is still near 5%. That is the same ladder the prior concentration companion published — because Epoch has not yet closed a Q1/Q2 2026 period delta that would force a restatement.

The concentration geometry therefore freezes: Top-1 = 25%, Top-3 = 54.8%, Big-5 HHI on the renormalized five-name perimeter ≈ 2,421 versus an equal-five benchmark of 2,000. The Lorenz panel still rises well above the equal-split line by rank three. What changed in Q3 is not the ownership tip — it is everything around it. Analysts who wait for the ownership tip to “break” before updating risk maps will miss the site and token moves that already printed.

Ownership is still not usage. OpenAI, Anthropic, and other frontier labs rent most operational capacity. A lab can dominate the product narrative while owning almost none of the H100e stock. That is why the token scatter remains the corrective lens — and why Google’s dual lead (chips and tokens in this cohort) is news, not a law of nature.

Sites and clouds: the Q3 tip that actually moved

Open Sites & cloud. Synergy’s hyperscale-only ledger — the same frame as the Q3 update — puts 1,360 operational sites and 803 pipeline sites on the board. The United States takes 580 ops seats (42.6%) but 437 pipeline seats (54.4%). That gap is the concentration story: live ops are more geographically plural; new seats still concentrate into the US corridor. Treat the Aug “~915 large-DC pipeline” and the Q3 “803 hyperscale pipeline” as different universes — mixing them produces fake deltas.

On capacity among cloud operators, the Q3 print puts AWS + Azure + Google at 58%, up 1 pp from the Aug ranking tape of 57%. That is a small move with a large meaning: the tip did not dilute as inland campuses proliferated. Market bands carried from the Aug location rankings still put top-20 markets at 60%, Northern Virginia + Greater Beijing at 17%, and the United States in 15 of 20 largest-market seats. New campus IT size ≈ 2× the current operational average means each pipeline seat is also a larger concentration event than the seat it replaces.

The Gartner electricity slice is the watts twin of the site ledger. World DC power prints near 565 TWh in 2026; the US alone is 204 TWh (36%), with roughly 68 TWh dedicated AI inside the US (~1/3 of US DC power, ~12% of world DC TWh). Regional AI DC capacity by power draw — the synthesis carried from theme research — still puts the United States near 45%, China near 18%, and Europe near 14% (Top-3 regions ≈ 77%). Site counts, capacity shares, and electricity TWh are three meters on the same geography — they will not match digit-for-digit, and they should not be averaged.

Tokens: Google takes Top-1 — the ownership mismatch shrinks, then doesn’t

Toggle Tokens vs chips. The prior concentration print used a June 2026 major-brands series with ByteDance at 29.2% and China-origin brands near 62%. The Q3 cohort — aligned with the theme update — puts Google near 27%, ByteDance near 21%, OpenAI near 15%, and a tighter residual. Token Top-3 (Google + ByteDance + OpenAI) clears about 63%. US-origin brands now dominate the cohort (~56%) versus China-origin (~33%) — a sharp flip versus the June China-heavy series.

That does not mean ownership and usage finally agree. Google is the rare name that sits near the diagonal of the scatter: ~27% tokens and ~25% ownership. ByteDance still prints ~21% tokens against ~1.5% ownership. OpenAI prints ~15% tokens against ~0% disclosed chip stock. Microsoft, Amazon, and Meta remain ownership-heavy versus first-party tokens — Azure, Bedrock, and Llama routing dilute credit. The concentration lesson is therefore two-sided: the identity of token Top-1 changed, but the mismatch geometry between rented usage and owned chips did not disappear.

What the prior → Q3 deltas actually say

The ownership tip is the stable risk factor. The cloud tip is the creeping one. The site ledger is the placement one. The token tip is the volatile one. Desks that collapse these into a single “AI is concentrated” slogan will mis-rank which perimeter actually moved this quarter.

Caveats and reading rules

1. Carried ownership is not a no-news print. It means Epoch has not closed a superseding period. Treat Google 25% / Big-5 71.4% as the last disclosed ladder, not as a Q3 discovery.
2. Hyperscale-only ≠ large-DC mix. Synergy’s 803 pipeline seats are not the Aug ~915 large-DC count. Do not subtract them.
3. Token cohorts are editorial. Vendor keynotes, China NDA stats, OpenRouter samples, and revenue proxies do not form a single audited market. June vs Q3 Top-1 flips can reflect brand-set changes as much as true share migration.
4. Dual-hub 17% is capacity, not sites. Northern Virginia + Greater Beijing as 17% of hyperscale capacity is not the same meter as US 54% of pipeline seats.
5. H100-equivalent ≠ identical utility. Peak 8-bit FLOP/s conversions hide memory, networking, and software differences across TPU, Blackwell, Trainium, and Ascend.
6. Cross-perimeter averages are meaningless. Do not average owner Top-3, cloud Top-3, US site share, and token Top-1 into one “concentration index.”

What would change the story

A finalized Epoch Q1/Q2 2026 Chip Owners period that moves Google off 25% or lifts China ownership well above 5%; a Synergy capacity print that pushes Top-3 cloud below 55% as neoclouds and Meta/Oracle absorb share; a pipeline ledger where US seats fall under 45% as Middle East and APAC campuses clear; or a token series that restores China-origin brands above 50% without a matching ownership move. None of those supersede the Q3 site and cloud prints in the sources we carry here.

Related reading: 2026 concentration print, Q3 site-count update, August location rankings, and theme research baseline.