Jim Chanos — research hub

Chanos & Co. founder/president (ex-Kynikos Associates) · veteran fundamental short seller — running synthesis of his video/podcast appearances, with per-transcript breakdowns and a stock index.

Sections: stock index · overall thesis · the product · transcripts. Last updated 2026-JUL-23.

Stock & name index

▲ Positive

Ticker	Name	Current thesis	Research	Seen in	Total $k

► Neutral / referenced

Ticker	Name	Current thesis	Research	Seen in	Total $k
AAPL	Apple	Arguably right to sit out the AI-capex race — "didn't take the risk that might prove unfounded"; a capital-discipline nod, not a stock call.	QT · SA · STK · FA	2026-JUL-17	—
AMZN	Amazon	Hyperscaler that raised CapEx ~$120B→$200B; part of the group whose return on incremental invested capital is falling 40%→20%→toward 10%.	QT · SA · STK · FA	2026-JUL-17	—
Anthropic	Anthropic (private)	Counterparty on one of xAI's short-dated "spot" compute deals; part of the researcher brain-drain churn between the AI labs.	—	2026-JUL-17	—
APO	Apollo Global Management	Peer reference — did the ~$35B June TPU-financing deal with Blackstone; his one named financial short (an office/low-cap-rate PE entity) is deliberately NOT attributed here.	QT · SA · STK · FA	2026-JUL-17	—
BX	Blackstone	Peer reference — named with Apollo on the ~$35B TPU financing; the unnamed office/low-cap-rate PE short is not attributed to any ticker.	QT · SA · STK · FA	2026-JUL-17	—
GOOGL	Alphabet (Google)	Raised ~$80B (upsized to $85B) and struck a short-dated "spot" compute deal with xAI ("tremendous outs"); a hyperscaler in the ROIC-compression group.	QT · SA · STK · FA	2026-JUL-17	—
JPM	JPMorgan Chase	Banks are tangential to the AI blow-up this cycle and have done well; JPM looks a little expensive (>3× tangible book) but he has "no involvement."	QT · SA · STK · FA	2026-JUL-17	—
MSFT	Microsoft	Profitable hyperscaler, but Azure has stumbled vs AWS/GCP and incremental ROIC is compressing; an OpenAI–Microsoft legal fight is "coming."	QT · SA · STK · FA	2026-JUL-17	—
NVDA	Nvidia	The gatekeeper/price-maker and the valuation ceiling of the AI complex — nothing Nvidia-dependent should trade richer than Nvidia; margins solid a couple years, then "anyone's guess."	QT · SA · STK · FA	2026-JUL-17	—
OpenAI	OpenAI (private)	At the center of the "circular financing" and litigation (Apple's suit, a coming Microsoft fight); the OpenAI/AMD mega-deal that spiked then unwound Oracle.	—	2026-JUL-17	—

▼ Negative

Ticker	Name	Current thesis	Research	Seen in	Total $k
C-Squared	C-Squared / formerly Sixtera (private, data-center re-IPO)	A legacy data-center operator that went bankrupt in 2023 (formerly Sixtera) trying to re-IPO — part of the legacy operators dumping portfolios as maintenance-CapEx reality bites.	—	2026-JUL-17	—
CRWV	CoreWeave	Reportedly seeking derivatives to hedge downside in GPU/asset prices — a tell that contradicts the "assets only go up" neo-cloud narrative.	QT · SA · STK · FA	2026-JUL-17	—
META	Meta Platforms	Off-balance-sheet Louisiana data-center SPV (Blue Owl/KKR, owns 20%, 3-yr out) + an "excess compute" rent pivot after repeated Meta-AI stumbles — "no one wants the hard assets."	QT · SA · STK · FA	2026-JUL-17	3.0
MU	Micron Technology	Skeptical of the LTA/order-book faith — a company with negative gross margins 3 years ago now awarded market cap on 80%-margin promises penciled off spot prices.	QT · SA · STK · FA	2026-JUL-17	—
NBIS	Nebius Group	The most asset-heavy neo-cloud ($4.40 capital per $1 revenue) suddenly going asset-light — "a tremendous admission" the capital-intensive story doesn't work.	QT · SA · STK · FA	2026-JUL-17	—
ORCL	Oracle	Worst incremental-ROIC of the hyperscalers; stock down ~50% (65–70% off the OpenAI/AMD-deal spike) — the group's cautionary tale.	QT · SA · STK · FA	2026-JUL-17	—
SLG	SL Green Realty	No position, but the bellwether of absurdly low cap rates — a 5-cap on NYC office, "gone nowhere in 25 years" — that blows up if rates hit 6–7%.	QT · SA · STK · FA	2026-JUL-17	—
SpaceX	SpaceX (private)	$75B IPO chased by $25B of debt; valued "on promises"; its stock was used to buy xAI at a $250B mark — emblematic of the issuance torrent that starts a top.	—	2026-JUL-17	—
TSLA	Tesla	Cited as a stock "valued on promises" — the archetype of the bull-market "premium on promises" that reverses when reality gets its discount.	QT · SA · STK · FA	2026-JUL-17	—
xAI	xAI (private)	"A money pit" that overbuilt for its own needs and now rents out compute via "highly promotional" spot deals with ~3-month outs.	—	2026-JUL-17	—

Overall thesis

In one line: the AI buildout is a bigger, worse bubble than 1999–2000 — the physical assets (data centers, GPUs) may be uneconomic from day one, funded with off-balance-sheet and near-junk debt and underwritten on spot prices against 20-year asset lives; all three top signs (rich valuations + retail speculation + heavy issuance) are now present, and the real fragility is rates/cap-rates, not power. He runs a ~40-idea hedged model portfolio (own money + institutional advisory), so the book is expressed as shorts of the ecosystem's over-valued/accounting-flattered names, not an outright market bet.

"Much worse than the dot-com" build. In '99–2000 the big spenders (BofA, GE, Coca-Cola) were profitable enterprises that just cut orders; the failures (CLECs, fiber) spent ~$100B over five years. Today's numbers "dwarf" that, and outside the hyperscalers much of the ecosystem "is not necessarily profitable" and VC/debt-funded — projects "that may be completely uneconomic from day one."
Spot prices on 20-year assets = the core "financial crime." Data-center leases pay a 20–25% return "only for a year or two" on a 20-year asset — a duration mismatch (the shale/railroad/GFC-repo mistake). "The sanctity of CapEx" is a contrarian tell: history says CapEx gets cut hard when marginal returns fall.
Construction-in-progress accounting inflates earnings. GPUs not yet plugged in sit in a growing "construction in progress" subaccount and aren't depreciated (the clock may not start for ~18 months). Spenders capitalize rather than expense while Nvidia books the revenue — which is how S&P EPS runs 4–5x its ~6% trend, exactly as in '99–2000. He depreciates over 10 years to be conservative and still "can't pencil out most of the data-center economics."
Incremental ROIC is rolling over — Oracle worst. Return on each new dollar of hyperscaler CapEx has fallen from ~40% (18 months ago) to ~20%, "moving toward 10%." At 10% pre-tax "you could earn that on treasuries," and the neo-clouds are then "in a whole world of hurt." Oracle "has the worst of those ROIC metrics."
The Nvidia valuation ceiling. Nvidia is the gatekeeper/price-maker; "not one company in the hardware space should trade at higher valuations than Nvidia — and yet many, many companies do." Those "specious" data-center/neo-cloud names are the short candidates (MU, ORCL, NBIS, CRWV).
Insider behavior is confessing. Coreweave hedging asset-price downside; Nebius — the most asset-heavy neo-cloud ($4.40 capital per $1 revenue) — pivoting to "asset-light" (a "tremendous admission"); Meta's off-balance-sheet SPV (Blue Owl/KKR, 20% owned, 3-yr out); legacy data centers dumping portfolios (Sixtera→C-Squared) as maintenance-CapEx reality bites.
Rates/cap-rates are the fragility — not power. Everyone underwriting 5–7 caps with the 10-year at 4.5%; rates to 6–7% "blows up asset class after asset class." Watch CCC (already widening) ahead of BB/BBB; 5% Treasuries "with vigor" widen spreads. Contrarian: power is ~5–6% of data-center revenues — "there's one thing we are not short of, it's power" — the bottleneck story is "silly."
Issuance + promotion = the top signal. SpaceX ($75B IPO + $25B debt), Google ($80→85B), near-junk neo-cloud/Oracle debt, circular Nvidia-invests-in-its-customers financing (a Lucent/Nortel echo), and "highly promotional" 3-month-out spot deals (xAI↔Anthropic/Google). "2026 feels like mid-year 2021" — shorts began working after the issuance wave. Everything is "valued as if it has worked or it will work"; Tesla/SpaceX trade "on promises."

The product — Chanos & Co. model portfolio & advisory

What it is: Chanos is the founder/president of Chanos & Co. (successor to Kynikos Associates). Per his own description here, the firm no longer runs third-party money directly — it "runs our own money and advises institutional clients." The centerpiece is a ~40-idea model portfolio, designed to be hedged (clients are told how best to hedge it, "whether through S&P or more custom-hedged"), with the view taken as "our 40 stocks versus the market" rather than an outright market call. There is no retail subscription; his public podcast/event/Twitter commentary is the only retail-accessible surface.

Grounded in his own statements in the archived appearance (2026-JUL-17).

Offering	What it is	How he runs it	Seen in the index
40-idea model portfolio	A concentrated book of ~40 fundamental short/negative ideas, structured to be hedged so it isolates single-name alpha from the market.	"We put together a model portfolio for clients… structured with 40 ideas, designed to be hedged." Advises clients to hedge via S&P or custom overlays.	The Negative names (MU, ORCL, NBIS, CRWV, SLG, META…)
Forensic / fundamental research	"Deep fundamental work" — accounting and cash-flow forensics that surface tells others miss.	Construction-in-progress balance-sheet reads, incremental-ROIC tracking, spot-price-vs-asset-duration tests, depreciation-schedule stress tests.	Drives every short thesis above
Institutional advisory	Running the firm's own capital and advising institutional clients — no outside/third-party money.	"We're no longer running third-party money directly. We're just running our own money and advising institutional clients."	(the client-facing wrapper)
Public commentary	The free, retail-accessible channel — podcasts, events and Twitter/X where he lays out the framework transparently.	"He talks about it in podcasts and events and on Twitter." This Risk Reversal appearance is an example.	(this archive)

How it serves retail investors:

The framework is public even though the book isn't. The model portfolio is institutional-only, but Chanos explains his reasoning openly in podcasts/Twitter — the retail value is the method, not a subscription.
Teachable, rerunnable screens. The Nvidia valuation ceiling, the construction-in-progress tell, incremental-ROIC tracking and the spot-vs-duration test are all things a retail investor can apply to any capital-intensive boom (see the actionable insights).
Short-seller's risk lens. A hedged, forensic-accounting perspective on the most crowded theme in the market — useful as a counterweight even for a long-only investor deciding what not to own.

Transcripts

One dated page per appearance — each has its full stock table, talking points, and the saved transcript. Newest first.

Date	Title / analysis page	Show	Video	Transcript	Actionable insights
2026-JUL-17	The AI Bubble Is "Much Worse" Than Dot-Com	Risk Reversal (Nathan/Adami)	▶ YouTube	transcript	insights

To process — backlog

Jim Chanos appearances discovered via YouTube search (Jim Chanos), not yet processed — verify publish dates & channels, newest first. Limited to the last ~2 years. None queued yet.