The Investor's Podcast Network's flagship show (We Study Billionaires), hosted by Stig Brodersen. This archive covers the recurring Bull vs. Bear episodes: one company per episode, one panelist building the complete bull case and a second given explicit licence to attack it, with the host staying deliberately neutral. A panel archive — stances are the roundtable's net take on each name, with the individual member attributed in the per-name note (Manish Karira = bull seat, Ralph = bear seat and a career CPA/forensic accountant, roles rotating between episodes).
Aggregated across episodes; each ticker links up to the consolidated per-security page. Because this is a multi-member panel, the "Current thesis" reflects the roundtable's net view and names which member held it. Newest mention first.
World's largest battery maker (40% EV share, ~$280B): Manish's bull case is a flywheel moat plus two under-priced engines — AI-data-center storage (H1-2026 sales +88% y/y, higher margin than EV cells) and the LRS license-royalty model (Ford Michigan) — at ~18× EV/EBIT, 17% ROIC, 25% ROE for ~15%/yr; Ralph's bear case is a price-deflation treadmill (volume +21.8% vs revenue −9.7%), a windfall margin regressing to 11–12%, an unwinding supplier float, "LRS leakage" and key-man/political risk.
Nearest battery rival at ~16% share and the world's largest EV maker — described as a benchmark, not a call: vertically integrated (most cells self-consumed), competing with CATL mainly in R&D (10→97% in 9 min vs CATL's 10→98% in under 7), and the reason CATL exists commercially (BYD refused to supply BMW in 2012). Berkshire's Munger-led 2008 stake returned ~20× before a full exit last year.
Private Chinese AI lab; the standout among CATL's 150+ strategic stakes — CATL put ~$700M into DeepSeek's first outside round (~$7B at a >$50B valuation, per press reports) as an investment in a future power-hungry customer. Manish treats the whole stake book as strategic rather than financial (~$11B carried at book, which would take CATL from ~18× to ~17× operating profit).
In one line (as of 2026-AUG-22): This is a process show, not a call show — the recurring output is a fully-built bull case immediately stress-tested by an appointed bear, so the durable value is the shape of the argument rather than the verdict. The first archived episode, CATL, is the template: Manish Karira builds a moat argument that refuses to name a single moat ("the moat is that these advantages feed each other and create a flywheel" — biggest → cheapest → most profitable → biggest R&D → best technology → more customers → bigger still, extended upstream into mining stakes and downstream into car-platform design), adds two engines he says the market has not underwritten — AI-data-center storage (batteries buffering GPU load swings of "hundreds of megawatts in seconds"; H1-2026 storage sales +88% y/y with volume nearly doubled, at higher margins than EV cells) and the LRS license-royalty-service workaround to the US ownership ban (Ford owns the Michigan plant, CATL takes a ~3–4% royalty) — and prices it at ~18× EV/EBIT, ~21× earnings, 17% ROIC, 25% ROE for a business that "roughly doubles in value over the next 5 years," ~15% a year plus a point or two of dividend. Ralph then runs three forensic lenses that generalise far beyond this name: the treadmill (volume +21.8% against revenue −9.7%, because raw-material indexation contractually hands efficiency gains to the OEMs — "running exponentially faster to stay in the same place financially"), the windfall margin (a ~15% net margin earned while revenue shrank is arithmetic, not superiority, and regresses to an 11–12% band), and the supplier-float audit (cash flow at ~2× profit only because payables are stretched — an Amazon-style interest-free loan Beijing is now forcing to be repaid faster, "stripping away the cash used for buybacks and dividends"). His structural objections are the monitoring list: LRS as "defensive capitulation" that turns the licensor into "a low rent IP landlord" and "a ghost in the machine" whose IP a regulator can sever with no physical assets to reclaim, while "LRS leakage" trains the very competitors trying to exclude it; the Pentagon list; EU CBAM; Hungary/Germany fixed overhead against a European subsidy cliff; and key-man risk ("I will remind folks about Alibaba"). The bull concedes every one and answers with lowest-cost-producer resilience, sodium-ion as the commodity hedge, and the fact that Robin Zeng "maintains a low profile and rarely speaks in public" — a risk graded low probability, high impact, unhedgeable. Stig's contribution is the one that keeps the show honest: after a superb grid/load-shape explainer (base load, intermittency, 50 vs 60 Hz, and the stadium analogy for synchronized GPU load) he refuses to let the analogy do the work — "someone is going to sell a lot of batteries. It doesn't prove that it's going to be CATL" — which is what surfaces the real answer (a 97% utilization ceiling, not lost competitiveness).
The format is the method. Roles are assigned, not volunteered: one panelist owns the bull case, another is given explicit permission to attack it, and they plan to swap on the next name. Stig's stated reason is the transferable part — "whenever you are bullish about a company you end up speaking with other people who are also bullish… you end up in the same echo chamber," and an unappointed bear pays a social cost ("that guy doesn't get invited back"). Manish's standard is Munger's: "you have to argue the other side better than the other player." The practical output is that the objections the bull can only mitigate — never rebut — become the monitoring list.
Moats are looked for as loops, not line items. The recurring analytical move is to refuse the checklist answer: cheapest producer, biggest R&D budget, best technology are treated as symptoms, and the moat is the circuit connecting them. A competitor's problem is then restated as "how many years of compounding must be undone simultaneously" rather than "can they match X." Supporting moats are audited physically — switching cost via the qualification gate (crash/safety/durability testing locks a design win to a platform's 5–8-year life), complexity via component count and service life ("more components than a Boeing 747," 20+ years).
The bear seat is forensic, not rhetorical. Ralph's objections are all built from the same three financial-statement reads — volume vs revenue (price deflation), margin expansion against shrinking revenue (a temporary cost wedge), and operating cash flow vs net income (whose money funds the growth). Each is a screen that can be rerun on any manufacturer. His management screen is the other half: run management before the model, look for the narcissist/sociopath framing he learned to spot investigating Ponzi schemes and bank frauds, and weight demonstrated give-back behaviour as the positive counter-signal — a half-dissent from Buffett's "pick the industry over the management."
Geopolitics is priced as an unhedgeable, not argued away. Both sides accept that anyone owning a Chinese champion accepts political risk that cannot be hedged; the disagreement is only about grading it. The useful discipline is the classification — low probability · high impact · unhedgeable — which makes it a position-sizing input rather than a valuation input, plus the alignment counter-test (is the company in a sector the state wants to grow, and does the founder stay off the political radar?).
Access and market structure get real airtime. The show ends its valuation work by asking how a non-domestic investor actually buys the thing — dual listings, float size, fungibility, northbound Stock Connect for institutions versus the retail-only line — and prices the premium as a measure of who is allowed to buy, anchored to a long-history comparable (TSMC's ~15% US premium) rather than to zero. This is the part most single-stock write-ups skip.
Transcripts
One dated page per episode — each has its talking points and the saved transcript. Newest first.