LLM Watch

Bts Field Book No4 Model Money Monitor Wk Ending 2026 10 24

376KB ∙ PDF file

Download

Download

Llm Pricing Top30 Workbook Wk Ending 2026 10 24

1.8MB ∙ XLSX file

Download

Download

Pod Summary

The AI Unwind in Numbers

• The week’s equity weakness = a sharp reduction in crowded positioning, particularly around AI, rather than clear evidence of a broader market collapse.

• This is probably not “the big one”.

  • Net leverage has been aggressively reduced, with positioning measures reaching the bottom percentiles

  • Similar to the deleveraging seen from late June into late July.

• The damage within AI-related trades has been much more severe than the headline indices suggest

• The central distinction is between a modest-looking index drawdown and a substantial collapse in individual speculative trades.

AI financing: competition for capital becomes the constraint

• Only a small number of players can compete effectively for credit-market funding.

• Is access to financing is straightforward even for the largest companies?

  • Borrowers are having to pay more.

  • Beyond a certain point, a higher interest rate may not solve the problem because lenders hit capacity limits.

• This creates a potential dividing line within AI: the ability to fund infrastructure may matter as much as the ability to generate demand for it.

AI capex: compute demand is not external revenue (Minute 4)

• Is the bullish argument that extraordinary demand for compute, including strong Hopper rental rates, is sufficient to justify AI investment.

  • Compute demand is not equivalent to revenue.

  • The relevant question is how much revenue must ultimately come from customers outside the AI industry to support the infrastructure spending.

• Discussion of Paul Kedrosky’s AI capex revenue hurdle calculator

  • It translates AI capital expenditure into a required level of external revenue.

  • It then expresses that requirement relative to GDP, global software revenue and US white-collar salaries.

  • The purpose is to test whether the economic assumptions behind the investment cycle are plausible.

• “Unlimited demand for compute” becomes less useful as an investment proposition unless it translates into real customer spending outside the sector.

• Are credit markets are beginning to confront precisely this issue?

AI usage versus spending (Minute 7)

• Rapidly rising usage need not produce equally rapid growth in expenditure.

  • Better routing, system design and optimization can allow usage to grow geometrically while spending flattens or declines.

  • The key issue is whether customer spending will grow enough to justify the capital already committed.

• The strategic motive for continued infrastructure investment.

  • Large operators may be trying to become indispensable, establish a toll-road position or price competitors out of the market.

  • In that scenario, the infrastructure owners could retain value even if the model developers face difficult economics.

• Own the compute rather than be Anthropic or OpenAI but compute ownership only works if the owner can continue accessing capital markets. That narrows the investable universe substantially.

• The hosts compare the possible outcome with a utility-like market in which a small number of dominant providers control access and capture the profits.

What happened to Anthropic’s S-1?

• Anthropic’s expected S-1 filing did not appear on Friday.

• Delayed disclosure may help preserve the AI narrative.

  • Once detailed financial information becomes public, investors must assess the actual economics rather than an open-ended story.

• Differentiation in private debt markets between OpenAI-backed and Anthropic-backed data-centre projects.

  • Some OpenAI-linked deals are reportedly not getting done.

  • Where financing is available, he has heard of differences of approximately 200 to 300 basis points.

  • He explicitly qualifies these observations as individual data points, not a comprehensive market survey.

The sell-off spreads beyond direct AI beneficiaries (Minute 10)

• The latest week’s weakness has extended further down the AI-related supply chain.

  • Copper miners fell nearly 10%. Iron ore miners were also coming under pressure.

  • The damage is moving from direct AI winners into second- and third-order beneficiaries.

• Even traditional value investors are beginning to feel the effects.

• He also sees increasing concern about slowing growth and a more stagflationary backdrop.

• But do these apparently macro-driven moves simply just instead reflect the mechanics of crowded-position unwinds.

Equal weight S&P versus Nasdaq (Minute 12)

• The ratio has staged a striking rebound over past five weeks.

• For it to extend, the semiconductors probably need to break down - but the weekly chart still in primary uptrend:

• Betting against semiconductors means betting against currently booked revenue and earnings, not just a speculative narrative.

• The hosts agree that “cheap” and “expensive” are difficult labels for highly cyclical businesses.

• A bullish memory-stock thesis requires a clear view on:

  • The timing and scale of additional Chinese supply.

  • How that supply affects pricing.

  • Whether strong earnings can persist for another year or longer.

• Very low multiples could produce attractive returns if current profitability lasts.

  • However, the thesis ultimately requires confidence in earnings further out, including 2028 and 2029.

• The broader point is that the durability of earnings matters more than the apparent cheapness of a single year’s multiple.

Industrials ‘20-for-20’ (Minute 18)

• The traditional portfolio role for industrials is challenged in current pricing environment.

  • Industrials should ideally provide a steady base of compounding businesses.

  • Many now carry substantial premiums associated with AI infrastructure, electricity constraints or defence spending.

• He revisits a hypothetical portfolio built from a 2006 starting point.

  • A portfolio extrapolating then-current narratives might have owned auto companies, handset manufacturers and GE.

  • A more successful hindsight portfolio could have included FANUC, Caterpillar and Schneider Electric.

• The lesson is not simply to identify good businesses, but to distinguish business quality from the price paid.

• Caterpillar is a classic case study - an “accidental bottleneck stock”.

• He sees a similar pattern across industrial businesses connected with physical power constraints or the broader AI theme.

• Defense stocks present a related problem.

  • Their thematic narratives are powerful.

  • That does not automatically make their starting valuations attractive for a portfolio assembled in 2026.

Gold miner cash generation (Minute 23)

• Popular narrative is that gold miners are “printing money” at $4,000 gold.

  • Some companies may be generating substantial cash.

  • That does not justify assuming the same economics across the sector or buying GDX indiscriminately.

• Discussion of research on ASX-listed gold miners.

  • The spread between gold prices and reported all-in sustaining costs can suggest margins of approximately $2,000 per ounce.

  • Yet those apparent margins are not consistently appearing as incremental cash on company balance sheets.

  • A major reason is spending classified as “growth capex”.

• Some growth spending, including mine extensions needed to keep mills supplied, is economically necessary rather than genuinely optional.

• He draws a direct comparison with AI accounting.

  • Ignoring model-training costs (‘EBTIT’) flatters the economics of AI labs.

  • Ignoring recurring mine-development requirements can similarly flatter the economics of gold producers.

• The analytical requirement is to identify which companies actually convert high gold prices into cash after all necessary spending.

The US life insurance industry runs a balance sheet the size of the entire American banking system, at roughly 16 to 17 times leverage, with capital that Nick describes as “completely fake”.

A mere 5 percent write down on assets would wipe out the margin of error for most of the industry. And there no federal regulator with full sight of the system, and no obvious bailout big enough if it goes.

Show Notes

Systemic Risk In Life Insurance

14.9MB ∙ PDF file

Download

Download

Insurance and the data problem

• Why insurance? Nick calls it the most boring world - but the more you look, the more interesting it gets, for scary reasons.

• Filings are enormous. About 680 insurers, and a single annual statutory filing can be nearly 10,000 pages. Mapping holdings across entities requires reading them or using LLMs.

• The paywall problem. There is no EDGAR equivalent for insurance disclosures; the NAIC, the body meant to set guidelines, charges per filing (around 20 dollars each). SEC filings give you “really nothing”: you cannot see capital hardness or exposures.

• Bad actors are pushing the risk spectrum. The highest yielding asset is private credit, so anyone pushing ROEs piles into it, and competitive pressure drags everyone else in.

The $10 trillion balance sheet (Minute 3)

• Scale. Asset side is about $10 trillion, liabilities of $9.6 trillion, once discounted. Functionally a balance sheet the size of the entire US banking system. True in force policies are way above $9.6 trillion.

• Discount rate mechanics. Liabilities are PV’d back and the residual is equity. Contrary to what most assume, there is no standard risk free discount rate. Offshore reinsurance creates “toggles” to change the discount rate and flatter returns. Operators openly call it regulatory arbitrage.

• The black Box. Aggregate in force policies and liabilities are visible, but average duration and granular data are not. Whatever sits in Bermuda, the Caymans or Barbados is legally beyond the reach of US regulators; the fact that Nick cannot answer basic questions after 15 months of reading NAIC filings is itself one of the problems.

• Regulatory structure. No single federal regulator. Fifty one jurisdictions set their own rules, the NAIC is a quasi private standard setter “like the Federal Reserve except no law made it”, and individual state commissioners can grant “permitted practices”, essentially permission slips to break NAIC guidelines.

• Nick’s view: the system is intelligently designed to have no oversight. Fed involvement is limited to SIFIs like AIG, MetLife, Prudential.

Leverage, marks and capital quality (Minute 8)

• Industry leverage is about 16 to 17 times. Huge dispersion: mutuals are generally better, listed companies worse, and private equity backed insurers worse again, with exceptions both ways.

• Athene (Apollo’s life vehicle) has only 8 percent Level 1 assets (cash and treasuries). The rest is Level 2 and Level 3. Level 3 is an internal Excel model, but Nick has concerns about Level 2 as well: observed marks do not match market prices, and top of book liquidity is not the same as clearing prices.

• Most insurers are in significant trouble if assets fall 5 percent, before even considering liabilities and hedging.

• Mental model: an insurer is like a bank. Assets are mostly loans (plus equity repackaged as credit, including the newer “equity ABS”); liabilities are premiums instead of deposits. It is a spread business.

• Leverage is simply assets over equity, defined the same way as for Lehman. The capital and surplus buffer between solvency and insolvency is “super soft, as in completely fake”, inflated via permitted practices and questionable or phony reinsurance. Insurers do go under, failures seem to be picking up, and when they fail there is always less capital than expected.

• State guaranty funds cap protection at roughly 250k (300k in New York; California covers 80 percent up to 250k), on the total policy value, unlike FDIC deposit cover. A 1 million dollar annuity becomes a three quarters write down.

  

• Runoff evidence. Insurers report surrenders running far above expectations as rates rose, despite policyholder inertia. The reputational run dynamic (the SVB, “60 Minutes asks if your insurer is safe” scenario) is unmodeled and could spike extraordinarily. With a 5 percent margin of error, a 15 percent write down on private credit puts some carriers close to the edge.

What ‘Civilians’ Can Watch (Minute 25)

• Equities will not tell you (passive flows distort prices). Watch the credit instead, for example Global Atlantic and similar insurer bonds, and subordinated issues. Practical advice: buy the 20 dollar statutory filing for your own insurer, run it through an AI, and check private credit exposure, leverage, and affiliated reinsurance out of Bermuda.

• History says it can happen. Annuities existed in the city states of the 1500s and 1600s; insurers failed in the Depression, at the turn of the century, Executive Life failed in the 1990s high yield unwind, and failures are occurring now. What has never happened in our lifetime is a critical mass failure, and Nick predicts it will pick up a lot.

• Complexity exceeds banking. Athene’s derivative book alone is a 3,000 page exhibit. Filings are only snapshots, and books can be dressed up for each reporting date. Regulators and commissioners do not have the answers either. Opacity has opened a window for extreme risk taking and, in some cases, fraud.

• Tails hide in the shadows. Actuaries are smart, but almost nobody can jointly model asset correlations in a tail event and liability behavior under stress. People are predictable in normal times, not in crises: in 2008, policyholders surrendered even as the Fed cut rates, because they were broke.

• Correlation is where it dies. Models assume stationary correlation, then in a tail everything goes to one while you need liquidity. That was the basis error behind CDOs, and it repeats in private credit and broadly syndicated loan CLOs, which are less diversified than believed (around 35 percent software exposure, heavy housing, real cyclicality). Outliers get excluded from the stats, which is exactly how things blow up.

The endgame ( Minute 35)

• The bailout question. Government could plug the hole, but Nick’s worry is that mispriced, consumer linked credit is everywhere, including the housing market, so a rescue could spiral beyond the sort of control seen in COVID.

• The required number would exceed COVID (which exceeded World War II and dwarfed 2008), insurers are roughly 150 percent of the Fed’s balance sheet, and a credible response could mean 10 to 15 trillion dollars, expanding M2 by 25 to 33 percent in a year, with the bond market potentially revolting.

• This is global. SocGen, BNP Paribas and Allianz carry similar private credit exposure, as does Asia.

• Authorities always refuse the medicine, which takes gold toward $10,000/Oz. Gold works in both deflation and devaluation; deleveraging crashes are the time to buy it, as in late 2008 before the GFC structurally ended that November. Practical implication: shorting insurers caps you at 100 percent; the bigger money is on the other side of the trade.

The New Subprime Crowd Wears Patagonia Vests (Minute 42)

• Growth cannot bail the system out of a 6 percent deficit, leaving two paths: deleveraging through defaults or inflation. Private equity professionals are borrowing against carry and even their homes to meet capital calls and GP commitments of 2 to 5 percent per fund, with nothing returning, while funding Connecticut lifestyles on million dollar salaries. Asset rich on paper, cash poor.

• The K shaped economy risk. If the top of the K stops feeling wealthy and stops spending, you get a recession that is extremely hard to fix. And this time the bailout politics are different: in 2008 people were in shock and Occupy came later; now everyone knows from day one who would be rescued (Apollo is the named example; Leon Black still owns 7 percent). Nick doubts the political appetite exists, raising the risk of a massive recession instead.

• The historical analog is Japan, but Japan was Asia centric; this could be a global lost decade, with AI generating growth in pockets while everything else stagnates and most people do not work at Anthropic or OpenAI.

• Europe and Japan are huge net creditors to the US. If they need to plug domestic holes, they will be forced sellers of US assets (the S&P and the Nasdaq).

Nick’s New Real Estate Report (Minute 53)

Mispriced Assets

Too Big to Bail: US Real Estate

The scale of it…

Read more

2 days ago · 15 likes · 3 comments · Nick Nemeth

• Hyper levered housing: mortgage payments relative to median income are at the highest level ever, including 2006 and 2007. Commercial real estate, multifamily and US single family are all disasters, and Australia, the UK and especially Canada are as bad or worse. A global real estate problem stacked on top of the insurance problem.

• Nick still tries to pick stocks (”there’s always a bull market somewhere”), but after a year of pulling threads and finding progressively scarier things, he is genuinely concerned about what comes next. The actionable takeaway at the personal level: save, cut spending, build a war chest.

• An optimistic take to pair with the bearishness: an asset price reset would clear the forest, rebuild society on better footing, and put things back on track for his generation, because the current path without a crisis is off the rails.

Latest LLM / Token Chart Packs

Llm Provider Pricing Top30 Analysis

1.75MB ∙ XLSX file

Download

Download

Capital Misallocation Llm Chartbook

3.45MB ∙ PDF file

Download

Download

Show Notes

The week that was, and the AI unwind

• Key dynamics: a continued bid to bonds and growing concern about the AI narrative unwinding, centered primarily on neocloud balance sheets, a theme running since the Amish episode in March.

• The hosts turned bearish in early June, calling the blow-off top in AI CapEx recipients.

• SpaceX segment: the biggest unlock is coming; the market is being warmed up for a possible Tesla-SpaceX merger; SpaceX supply was talked about all summer but nobody is positioned. Shares went from around 210 down to near 100; the company has raised roughly $65bn in credit since listing (including a new $40bn package of $30bn IG plus $10bn term loan) plus around $110bn in equity. Elon has taken roughly $86bn in cash out of the market.

Last week’s price action and positioning (Minute 5)

• First real pain at index level after months of poor breadth: defensives caught a bid, staples up over 5 percent and utilities rallied.

• We doubt this is the flush moment: it looks like an overcrowded trade unwinding on positioning and liquidity, with vol and implied correlation too low, a pattern seen repeatedly over the past fifteen years.

• Midterm history suggests risk trades well afterwards; a Democrat result probably would not be great for stocks.

• Growth-scare data points: Anthropic numbers came in “miss-ish,” average token price hit an all-time low, and rumors that the IPO timetable could slip again. A slipped $100bn Anthropic IPO at a $2tn pre-money valuation could paradoxically be bullish for year-end liquidity.

• Triggers: the pulled Firmus IPO in Australia (long end of the curve bid from 5.70% to 5.40% since), plus UK neocloud Nscale having a major Microsoft contract pulled over the weekend. Amusing tweet: “if I were Jensen, I’d simply buy all my clients.”

Positions, sovereign backstops and China (Minute 9)

• 🐿️ is short Oracle and CoreWeave but nervous about NVIDIA acting as “the central bank of AI” and papering over the stress.

• “Too big to fail” whispers about OpenAI are now circulating, an idea the show covered with Brandon Carl (guest on next week’s show) and in earlier episodes.

• What sovereign intervention would actually mean: backstopping Wall Street’s data center credit at the monetary level rather than taking golden shares. OpenAI’s CFO floated a government backstop in Q4 2025 before walking it back on LinkedIn.

• The circular dependency problem: OpenAI’s commitments underpin Oracle and cloud backlogs plus data center DDTLs; if OpenAI goes down, data centers, debt and collateral go down with it, a GFC-style chain. The flip side is adding huge liabilities to the sovereign just as debt refinancing costs are under scrutiny.

• China keeps making tokens cheaper: the latest DeepSeek model offers near-frontier capability at roughly a 90 percent token-cost discount, so any US sovereign underwriting would be backing a disastrous forward P&L structure and would likely need to be twinned with protectionism.

• Fortune 500 companies increasingly want to build their own AI on open-source models and their own hardware for data security; fraud risk is rising; large corporates are blocking agentic access to their APIs, and technically blocking agents is much harder than it sounds.

Korea trigger and the IPO pipeline (Minute 21)

• Korea looks like the proximate trigger, reacting to the Anthropic numbers. The SK Hynix ADR still trades at roughly a 50 percent premium to the Seoul listing, which cannot be arbitraged. Even international retail can buy locally, so there is no reason to pay the ADR premium.

• OpenAI is definitely not coming public this year; Anthropic price talk is sliding from $2tn toward $1.5tn pre-money, and the view is they should just price at $1tn and get it out the door.

• IPOs are momentum games: publicly talking down valuation before pricing is almost unprecedented (closest comp is Facebook’s 2012 “mobile problem”, but at a $100bn valuation, not $1.5 to $2tn). Too late to halve the target now, and the forward capital problem, roughly half a trillion to raise over four to six years, means they have to start big.

Energy: the real hiding place (Minute 25)

• Energy was again a good place to hide. The standout chart is December 2027 Brent: the market is reaching for long-dated contracts while large spec shorts (COT data) remain substantial.

• Under the “China collar” theory, the move in long-dated crude underwrites the energy equities; the IXC global energy ETF chart looks strong.

• Spike risk is higher now: Jesper Kohl’s point that Japan runs out of reserves by January; China is down an estimated 300 to 400m barrels; global reserves are being drawn. Transport costs now add $20 to 30 per barrel, yet China is still lifting cargoes, suggesting more right tail.

• Demand is low volatility (COVID aside); supply is where the volatility sits. A double negative, soft demand against an even worse supply situation, could still push prices up.

Rotation mechanics (Minute 30)

• The AI trade is 40-plus percent of the S&P; neither host thinks it is truly rolling over yet, and if AI rolls over while the index holds via rotation, their non-AI longs benefit.

• India outperformed the rest of Asia massively last week; Brazil is trading on its own election cycle with the runoff the following week. Both are read as positioning: popular funding shorts (staples, Infosys, TCS-type names) getting squeezed in a degrossing.

• The “causal barbell”: implied correlation near zero with single-stock vol in the 99th percentile forces leveraged pods to degross when correlation rises, so shorts rip and longs get sold, but it does not cascade the way old crash analogies suggest.

Twenty for Twenty - The Industrials (Minute 35)

• Twenty for Twenty portfolio update: work began on industrials.

• The key industrial question is the China scenario (the “panda in the room”)

• BYD - arguably one of the most impressive industrial companies on the planet, but a great company is not necessarily a great stock. Deflationary technologies are hard to monetize compared with regulatory-capture industries and full-contact capitalist competition in Chinese industrials should never be underestimated.

The Extras

Tom’s bearish energy note:

The End Of The Age Of Oil 1

658KB ∙ PDF file

Download

Download

CPI Day and the Rates Selloff

• US CPI came in warmer, but Tom argues risk assets were on edge and looking for an excuse to sell: bond market worries on both sides of the Atlantic plus fragile tech positioning; bond yields spiked and stocks got hit, but he would not extrapolate a new trend from one print

• The painful short-end rates trade acknowledged: US 2yr at roughly 4.85% versus about 3.75 to 3.80% back in February; Tom is not outright receiving rates, preferring steepeners instead

• Rationale for steepeners: heavy long-end issuance including hyperscaler supply, and a Treasury already issuing at the shortest duration in the world with heavy bill reliance; outright receivers are too tough here

Banks, AI Capex and Credit Stress (Minute 5)

• Bank earnings: fixed income trading revenues fell sharply; Benny suggests Q2 risk manager “disciplinary chats” showed up in weak Q3 FICC P&Ls; Tom notes crowded carry trades hurt both the sell side and buy side when funding spreads blew out

• CoreWeave spreads around 1000, neoclouds back at lows, “outrageous” token deflation; Tom sees a credit squeeze on CoreWeave analogous to the squeeze in France, but the market refuses to judge the whole sector, with Meta near all-time highs; two tails playing out and neither winning

• Hyperscalers - even Oracle - will be fine operationally, but the credit market cannot absorb the paper (compounded by Basel hedging feedback); better to be a token consumer than provider; data center buildout faces transformer backlogs, interconnection delays and local political pushback, so Wall Street extrapolation is heading for a moment of truth

• The “free cash flow countdown”: aggregate hyperscalers plus Nebius and CoreWeave are now negative FCF.

• A rising cost of capital environment means revenue expectations must rise in lockstep or stocks fall; expect differentiation rather than all boats rising; sovereigns are competing for the same capital, with off-balance-sheet liabilities (retirement costs) seeping onto national P&Ls and structural deficits keeping everyone in “the price discovery zone”

Emerging Markets and China (Minute 17)

• EM local currency bonds: EM rates have matched DM sell-offs in size this year rather than underperforming, which is abnormal and reflects credible high real rates; risk of contagion from rolling stop-outs remains, but policy quality is doing the work

• South Africa: Tom is still a ZAR bull; around 9% FX carry has made the trade comfortable despite Trump and Musk “throwing stones”; 20yr bonds with a 9.5% coupon, a central bank that cut its inflation target midpoint from 4.5% to 3% and is credibly set up to hit it, fiscal consolidation attempts, and an electricity surplus now being marketed to data centers

• Brazil politics: Bolsonaro Jr’s decisive first-round win points to fiscal consolidation and inflation control; yields around 11% give a thick margin of safety built on creditworthiness rather than pure rates

• Prediction markets on Brazil are distorted by uninformed offshore money and the race is closer than priced; with Bolsonaro’s high rejection rate and family scandals, there is a real tail; advice to EWZ call buyers is to wait for 25 October

• The structural Brazil trade: local pension funds sit at about 5% equity allocation because belly-of-curve real rates beat their actuarial waterline; if fiscal dilution compresses those real returns, they must rotate into a tiny equity free float, a two to three year gift that keeps on giving

• China: no current positions; the yuan should keep appreciating gradually while Beijing hopes for a broader trade deal, but long yuan is poor reward against a Taiwan or trade-war risk-off event; prefers onshore A-shares (ASHR, CSI 300) over Hong Kong listings because giants like Alibaba and Tencent get politically squeezed, while smaller A-share names are “too small to be interfered with”

• China champions debate: BYD, CATL and Tencent are superb businesses but function as policy tools, subject to meddling from Beijing, Washington and Brussels, which blocks investors from extracting the returns the fundamentals would justify; they only flourish if a different, rules-compliant China emerges

Europe, FX and Politics (Minute 34)

• FX board check: DXY at 103 yet the RMB is grinding higher and the yen is holding up better than expected; Europe is taking the brunt of dollar strength, and euro-yen has been a good short

• Europe structurally challenged: Spain’s snap election expected to shift right, with callers of snap elections increasingly losing them; the pro-EU lobby has narrowed to a shrinking center while the center-right turns more Eurosceptic

• France scenario: center is fragmented so a Le Pen versus Mélenchon second round is plausible, and Eurosceptic Le Pen would win it; critically she is no fiscal hawk despite all of the “debt brake” talk - a case of the “dog that caught the car” finally having a crack at the title?

Energy, Gold and Wrap (Minute 41)

• Tom’s thesis (full note above): “the end of the age of oil”: global energy surplus, the US is now an exporter, China has spent a decade insulating itself, and the world absorbed the biggest oil shock since the 1970s without punishingly high real prices; UAE leaving OPEC and the Iran war look like the last gasp of the era, accelerating its end rather than reversing it

• Squirrel’s “China collar” thesis: China, run by engineers, has built strategic storage making it a monopsony price setter, stopping buying near $115 Brent and stepping back in near $80; that caps triple-digit extreme risk and reframes how to value global E&P, midstream, offshore and services

• Tom’s twist: oil abundance incentivizes bad geopolitical decisions (Trump on Iran) because surplus makes risk-taking politically palatable, which can keep eroding that surplus; shipping has become a real constraint, with transport costs of roughly $25 to 30 per barrel versus a historical $1 to 2, plus tanker losses in the Strait; refiner margins are distorted and EM demand destruction (three-hour scooter queues in Bali) is beginning

• Gold and silver: Gold always struggles when real rates rise, so recent performance is no surprise, but the long-term case is intact because China - running a big current account surplus and unwilling to buy Western financial assets - must keep buying gold

Show Notes

AI capex: the pulled IPO

• AI capex free cash flow countdown monitor now reads zero: the complex is no longer FCF positive as a group heading into earnings

• Firmus, the Asia-Pacific CoreWeave lookalike, pulled its ASX listing at a mooted multi-billion valuation despite NVIDIA backing, after local long-onlys refused to wear it; a NASDAQ float is now the plan, where gamma squeezes and smoke and mirrors are possibly available

Hyperscaler CDS and rate transmission (Minute 5)

• Hyperscaler 5-year CDS chart (black line is the 5-year UST as reference):

• CoreWeave near all-time wides above +900bps, having been flagged on the show at 465-470 in late May/early June; Firmus-related paper near 200bp, which is BB/junk territory; Microsoft the only contained name

• Most corporate debt and most non-US mortgages sit at five years and under; non-US mortgages are largely floating, hence the RBA’s near-instant transmission (Squirrel’s “most powerful central bank” line, qualified as best transmission mechanism)

• ECB is chained to headline CPI, hence repeated policy errors (Trichet twice in the hall of fame); the Fed has far more discretion

Best of times, worst of times (Minute 9)

• Bond investors and rate sensitives are hurting while AI capex and SPY/QQQ holders thrive; BofA framed 5% yields as agnostic for equities, but Hendry’s

  Niederhoffer rule says 5% is too much

• At some yield the average boomer rolls into the 10-year at 5% or TIPS at 3% real and coasts; high yields exist to attract capital (hunger games for capital), so always ask why the yield is being offered

• The eight-name capex monitor (Microsoft, Meta, Google, Amazon, Oracle, Firmus, Nebius, CoreWeave) shows aggregate FCF at zero; the “funded from cash flow” cheerleader line no longer holds

• Classic finance mistake: issuing paper when the narrative is best

Capital Misallocation Llm Chartbook

3.45MB ∙ PDF file

Download

Download

Llm Provider Pricing Top30 Analysis

1.77MB ∙ XLSX file

Download

Download

AI valuations: infinity, zero, or negative (Minute 15)

• Silicon Valley mindset: model labs are worth infinity or zero, and the average of the two is still infinity

• Squirrel’s pushback: the downside may be negative, not zero (rogue agents hitting the grid, BP-scale liabilities, no liability framework or social contract for AI), so expected value must include a negative tail

• Even the bull case is capped: Anthropic at half of global public market cap (roughly $120tn) is only a 30 bagger

Market structure: the causal barbell (Minute 21)

• NASDAQ/S&P ratio has reflated after the Q2 leverage washout that smoked the high-beta AI trade; still below the prior line, so plenty of pent-up energy

• Benny’s “causal barbell” essay lands this week: pod shops on one side (short vol, correlation zero), free unlevered beta on the other, with the active middle hollowed out

• A cozy equilibrium that works until it doesn’t; owning CDS insurance has bled premium (CoreWeave the exception) while vol sellers get rich

Single stocks: Home Depot and Rocket (Minute 24)

• Home Depot: world-class operator, ROIC down from 38% in 2022 to sub-20; roughly $14 EPS this year versus $16-17 in 2022; at about $280 the stock is 20x troughish earnings, with a wipeout to 14-15x implying $200-220; a 3.3% dividend yield cushions versus Treasuries

• Rocket Mortgage: an $11-12 stock with potential $2 of earnings; the Mr. Cooper deal (done at $22 last October) made it the second-biggest servicer, and the rate-driven selloff bought a free year for integration plus better AI tooling; a defined-use-case token consumer

• For equity managers who cannot trade macro products, Rocket is the equity version of a SOFR trade; Benny frames it as a trade, not a position

EM: Brazil over Korea (Minute 29)

• Brazil ripped on Flavio Bolsonaro’s unexpectedly strong first round; the 25 October runoff is seen by Brazil-watchers as a formality

• Korea has been the EM bull market on memory and DRAM; Squirrel sat it out (PTSD) but notes it remains cheap, prints surpluses, and the currency looks too weak

• Since the June memory-mania peak: EWZ up 29% versus EWY down 13%, a roughly 40-point pair trade in 3-4 months, with correlation near zero

• Brazil thesis: a bet on the Bloomberg Commodity Index plus 8-9% real rates pulling in global money

• The belly of the curve is the real story: local pension funds hold de minimis equity allocations because real rates meet their actuarial bogey; even a modest erosion of fiscal dominance under Bolsonaro sends them into a very tight free float where there is not enough stock

• Squirrel’s “wait for the runoff” call was overtaken by the strength of round one (last week’s EWZ and BRF moves); the 12-18 month view is pension flows driving equities

Capacity constraints and the bond rotation (Minute 33)

• Capacity cuts both ways: shorting is capacity-constrained with pod shops dominant, and on the long side holders lift offers, so price is whatever clears the system (uranium miners, Brazil)

• Open question: when does money actually rotate out of stocks and into bonds?

• Benny holds bonds only as a hedge: parts of his book he does not control are short bonds, and discretionary shorts in utilities plus long oil need protection; not yet a true believer, and duration is off the table

France: the G7 debt epicenter (Minute 36)

• Squirrel’s drive through France felt incredibly normal, yet bond markets are waking up to an emerging political crisis

• The old pattern (Le Pen scares the market, centrists rally and box her out) may be breaking: the horseshoe is in full effect and an April runoff of Le Pen versus Mélenchon looks increasingly likely

• Marine Le Pen is not her father’s daughter: she expelled him, talks debt brakes, and is tacking to the center on the Meloni template

• Our view: give populists the ball! After 30 years as the ghost opposition, the dog that caught the car should get to drive; if it fails, the left gets to say it told you so

Show Notes

The Bearish Bank Thesis

• Bank stocks peaked between August and September; Charlie still expects roughly another 20% downside in some names by January and a bear market in bank indices

• In bull markets the P&L drives bank stocks; at tops, peak revenue growth weighs, and in the developing bear market the balance sheet becomes the focal point. That transition is now underway

• Within S&P Financials, credit card stocks and exchanges bounced hard after a horrendous Q2, but everything else is red, especially regionals hit by rates

Credit Cards and Consumer Credit (Minute 3)

• Time to short credit card banks again: the 2007-style summer bounce was driven by seasonally bottoming credit improvement, which typically deteriorates into winter and peaks in February/March

• Interchange fees are 90% consumed by rewards costs; the Trump-backed Credit Card Act (rate cap) is a renewed margin threat, even if passage odds are low

• Card loan growth has slowed from double digits to mid-single digits; Citi card data shows high-end luxury spending down 6% YoY in September, an early crack in the top of the K

• BNPL names (Klarna, Affirm) look precarious; Charlie believes the rate-hiking cycle is not over, expects a flatter and possibly inverted curve next year, and notes rising auto delinquencies as the leading indicator

Goldman and Prime Broking (Minute 7)

• Charlie’s Q2 call that Goldman was his favorite short was a winner - Solomon’s bonus and Blankfein’s “God’s work” era

• FT reporting shows prime broking revenues have doubled in five years, with leverage concentrated in roughly the top 50-100 hedge funds; Goldman’s financing revenues have ballooned as its substitute annuity after the consumer banking failure

• Goldman is buying wealth management and alternative asset managers (CRE, CLOs, ETFs) at what Charlie sees as the top of the market; he gives Solomon low marks, citing the Greenpoint subprime acquisition disaster

Bank M&A Track Records (Minute 13)

• Asked who acquires well: M&T Bank (Wilmers, then René Jones) for disciplined blue-collar-market deals; Jamie Dimon’s Bank One to JPMorgan lineage built the best franchise

• Travelers plus Citi: supermarket banking did not work; Citi went from near market-multiple at the dot-com peak to a 10-11x name

• Fraser rebuilt credibility with realistic targets versus Corbat’s financial engineering; Charlie downgraded Citi off his buy list at $147 in August/September and would re-add it around $110-105 early next year

Q3 Earnings Preview (Minute 19)

• Goldman, JPMorgan, Citi and Wells report Tuesday; JPM sets the tone. Base case: mid-teen EPS growth on 9-10% revenue growth, positive operating leverage, flat provisions, and buybacks retiring 4-5% of shares annually

• Surprise risk sits in two places: capital markets (September is the swing month) and corporate non-performing assets, which are very low and vulnerable to episodic spike events

• Goldman pre-announced $350-400m of credit costs tied to write downs of investment banking loans on its balance sheet, possibly Paramount-related; framed as episodic, not recurring

Liquidity and Situational Awareness (Minute 22)

• The corporate credit problem is bigger than NDFIs/private credit: it is an asset-overvaluation problem spanning office, used cars, levered hedge funds, MBS posted by life insurers, Treasuries in repo, and securities-based loans now rivaling credit card outstandings

• On the Situational Awareness liquidation: primes reportedly got out whole; JPMorgan cut ties while Goldman stayed; hedge funds now have leverage over primes, forcing six-month notice periods into prime brokerage funding contracts (previously days, weeks, then months)

• Banks have shifted from stockpiling liquidity (cash at the Fed) to just-in-time liquidity management: lower LCRs, prepositioning at the discount window; Charlie sees liquidity drying up into year-end

Treasuries Basis Trade (Minute 30)

• The Treasury basis trade has grown by $1.5-2T in repo-financed Treasury longs since 2023: levered, not true, ownership is a metastable setup

• Charlie’s macro biases: no Fed hike at the October 27-28 FOMC, oil toward $120, then the market prices the Fed as behind the curve, forcing tightening in December and early next year; inversion would stress the basis trade

• Bond selloff framed as repricing fiscal policy: the $40B SpaceX GPU deal, US deficits of 6-7% plus Europe, Japan and China all running hot

• SDR mechanics: the $5B SDR increases in June and December 2024 look like trial runs; drawing SDRs could add $150B-plus to the TGA for buybacks or stimulus; “grow our way out” is the administration’s route

Capital and Buybacks (Minute 39)

• Banks ran capital down too aggressively via buybacks, then faced AI-funding balance-sheet expansion (corporate loans growing 9-10% vs 6% nominal GDP), unrealized bond losses, and capital relief delayed to H2 2027/early 2028; US Bancorp bought back zero stock in Q3

• Anthropic IPO is the big capital-markets payday; watch whether Dimon cheerleads or flags risks. The pulled neocloud deal (Nvidia-backed) is seen as the beginning of the end for circular AI financing

• Buy-side anger over Paramount Skydance debt secondary performance at BofA/Citi; Apollo leading the SpaceX chip financing shows private credit taking underwriting share from banks

Stablecoins and Deposit Franchises (Minute 48)

• Industry deposits fell about 1% in Q3, surprising bankers who guided to second-half growth; corporate and wealth deposits are most yield-sensitive, while core consumer deposits are what stablecoin legislation (Clarity Act) threatens

• 🐿️’s math: $27T of global non-interest-bearing deposits equates to roughly $170B of NII supporting financial repression; studies suggest deposit repricing to 4% would leave about 80% of banks unprofitable, so damaging deposit franchises is self-defeating even if stablecoins could buy bills

• Charlie expects another big deleveraging event within four months, citing deteriorated asset liquidity, stalled deposits, and elevated leverage

Positioning into 2027 (Minute 55)

• Late-week bid in bank stocks looks like rotation out of tech; Q3 itself offers few surprises, and 2027 guidance will stay vague until December/January

• Potential offsets to the bear view: rate hikes help asset-sensitive NII (JPMorgan, Citizens, PNC, and Wells Fargo, where the $50B NII guide could rise), and Q1 is seasonally strongest for capital markets, favoring the Anthropic IPO leads (Goldman, Morgan Stanley, JPMorgan)

• Closing thought: private equity/private credit book values are opaque and likely overstated, so FRE forecasts of 20-25% growth next year should be down; market turns are multi-year processes (GFC resets visible from 2004-05, dot-com unwound in waves), and the risk has not left the system, only changed hands…

Show Notes

French politics and OAT

• French OATs now trade wider than Italian BTPs

• Framed as an old-fashioned political crisis: markets waking up to the risk of a Le Pen vs Mélenchon second round in next year’s presidential election, with no centrist to rally around

• Exacerbated by Japanese sovereign selling…

,,, and the global diesel crisis

Carry trade unwind and rates deleveraging (Minute 3)

• Second major carry unwind this year; in deleveraging, price is irrelevant and positioning is everything

• Mexican peso/yen carry trade gave back 8 of its 12.5 year-to-date point gains in September alone

• WisdomTree EM currency fund surrendered half its year in 30 days

• Friday’s jobs report: 29k headline miss with wage disinflation, yet rates finished higher on the day, a sign that forced deleveraging overpowered the data

• Hawkish central bank rhetoric viewed as a show; the administration is “saying the quiet part out loud” that inflation is the debt paydown playbook, since no politician will choose debt deflation or austerity

Equity internals: a hidden bear market (Minute 9)

• Median and mean stocks down roughly 25% while indices sit near highs; around 70% of the S&P 500 trades below its 50-day moving average

• Rate-sensitive and non-AI small/mid caps are orphaned; September was brutal for long-only stock pickers

• Hunger Games thesis: in a competition for capital, the outer districts (illiquid periphery) pay first

• Higher yields plus higher energy costs squeezing the levered economy; real estate and private credit/equity players growing visibly nervous

• Only just over 1 of the 3 to 4 hikes priced through next year seen as realistic; December ‘27 SOFR position rallied 30bps (95.07 to 95.36) before reversing with rates

The AI put and Russell Resilience (Minute 16)

• Russell held up during Friday’s rates flush because the index is now loaded with AI second-derivative names, not “our grandfather’s Russell”

• Bank of America’s “AI put” framing: AI capex offsets weakness in the median consumer and corporation

• Friday’s price action read as a rates clearing event (someone carried out), not a fundamental signal

Cracks in Credit (Minute 18)

• Triple-Cs and private credit deteriorating, with dealers effectively making their own markets

• European banks gave back a quarter of their year-to-date returns in 72 hours at the start of the quarter

• New issue stress: Paramount opened 6 points under; SpaceX also traded poorly, while Nvidia and Amazon deals held firm

• Anthropic IPO flagged as a critical test: US markets cannot afford it to be pulled after OpenAI’s withdrawal; the “tourniquet” on liquidity may need to loosen around the October Fed meeting to preserve it

Contested election risk, and positioning (Minute 25)

• Historical pattern of post-midterm equity strength, with the only failures tied to oil/inflation spikes (Suez, 1973-74)

• Key mispricing: options markets underweight a contested election scenario, including a possible refusal to seat new members in January; suggested trade is funding December/January SPX puts with November election vol

• Duration avoided despite 5%+ nominal yields; preference for 2-year paper and reinvestment risk, with TIPS-derived real rates dismissed as too illiquid to trust

• Gold favored over 10s and 30s at $4,100: real rates no longer drive it

Show Notes

The Leaked Anthropic S-1

• The twin questions:

  • how to value the foundation-model companies

  • what the public-market displacement effects will be

• Izabella’s “Merlins” Thesis: AI labs as modern-day wizards whose products, if they work as billed, are equivalent to magic. Outcomes are binary and Tolkien-esque: global utopia or the gates of Mordor, with no middle ground, driven by founders who genuinely think in those terms

• The Atlantic piece on the OpenAI vs Anthropic feud: labs apparently discussed auctioning AGI to the highest-bidding government.

• Izzy sees it as “less Tolkien, more Austin Powers - one billion dollars for my AI” - a stark disconnect from the safety and alignment narrative.

Alignment as Game Theory (Minute 5)

• The calls to slow down were strategic positioning, not principle. Meanwhile the labs are consuming astronomical capital.

• Superintelligence should be a deflationary end-state, but the path there looks inflationary; a transition a hundred times more impactful than the industrial revolution arriving a hundred times faster is a genuine societal shock.

• Consumer power vs the “great unbundling”: Apollo’s chief economist argues AI lets consumers extract more value from producers (think bank deposit flight). Izzy’s counter: old certainties die. Price discovery arrives for previously free goods (news moves to subscriptions), and the dynamic resembles post-communist Russia: good agents versus bad agents, a Wild West trust crisis (you can’t trust your own dad’s voice on the phone), and a likely new cohort who happily trade data for platform protection, a new social contract.

• Marginal cost is not zero: Ben cites Harvey (legal AI) reportedly at negative 50% gross margins, and his own experience of AI spend scaling fast for casual users.

• The productivity puzzle: Solow’s line about computers being visible everywhere except the productivity statistics still holds; the last decade shows no boom. If tech gains had been real and shared, you would not have the inequality and “left behind” politics that produced Trump.

The revenue mismatch (Minute 13)

• Breakeven revenue requirements for the AI labs far exceed the revenue uplift corporate customers (per Wall Street analysis of SMEs) expect from AI tools. Either corporates lack imagination on deployment, or the whole loop depends on labs capturing a serious share of the global wage and services wallet, which then raises the distribution question: who is left to pay for anything, and who arbitrates between labs and the public if not government?

• Chinese ghost cities as the AI capex analogy: developers were implicitly insured until even Beijing pulled the plug (Evergrande). The cities failed on maintenance intensity, not demand; a poor man in a nice flat is worse off than in his village. AI capex is meaningless unless it pays for itself via real productivity, otherwise it is a ghost data centre.

Jeffrey West’s scaling framework (Minute 18)

• Innovation mostly just accelerates the pace of life to a faster equilibrium. Cities grow exponentially and never die because they stay open and do not filter good from bad; corporations plateau and die in about 100 years because closed systems get buried in entropy.

• Which path AI takes - ‘open commons’ or ‘closed kingdoms’ - is the live fight, and it hinges on whether Anthropic/OpenAI get their ultimate wish - a government guarantee.

• Too big to fail by design: per Brandon Carl’s piece, the strategy was to scale before auditors and regulators understood, locking in cloud backlogs (Oracle et al.) and now debt. The AI trade is the major driver of US growth and of the S&P; in a hyper-financialized economy that is a gun to the government’s head.

The Comms Failure (Minute 25)

• AI-bashing is now bipartisan. Izzy reckons that the average person neither knows nor cares who Dario Amodei or Sam Altman are and treats AI as a curiosity app.

• 🐿️ counters with data centers polling below nuclear plants and students believing the white-collar first rung is gone. Izzy concedes partly but points out that manual labor retains a moat until robotics. Her diagnosis is the labs behaved paternalistically, Versailles-style, demanding trust as the wizards who had figured it out.

• Why Washington stays hands-off: Trump flirts with an OpenAI stake, already took Intel. Implicit guarantees (Fannie Mae, AIG) and post-GFC central bank suppression, now blowing up in Japan and French rates, mean true capitalism effectively ended in 2008.

The ‘Three Wars’ Framework (Minute 31)

• US vs China for AI dominance, intra-cloud competition for market share, and local vs cloud. Wars are waged to win at any cost, not for optimal outcomes.

• Crucially, credit is now in the structure, unlike the ‘99 or 2021-22 SPAC wipeouts which were equity losses; the capex and off-balance sheet commitments ratcheting over the next two years make this systemic.

• Is this capex cycle different? Every capex cycle produces bag holders; only state intervention removes them. Silicon Valley Bank’s losses were socialized “in a heartbeat”, so precedents exist.

• The AI value paradox: if gains concentrate too much, the system fails its own intelligence test, and a new social contract becomes inevitable, akin to Putin’s post-1990s deal with the Russian public.

Trust and Design (Minute 35)

• The pace of information already exceeds individual capacity, so people will outsource decisions to AI agents, which requires a social contract that they act in your interest.

• That debate is happening behind closed doors with no public input. Capitalism in this environment, she argues, needs the government balance sheet, otherwise the system becomes a self-consuming Ouroboros.

• Agents, containerization and liability: users carry liability for what agents do inside their accounts, the left tail of his AI ROI framework. Adoption is power-law distributed (0.1% of users in a different league, mirroring wealth), and the advantage fades once everyone has the tool, the Blackberry effect.

• The circular economy problem: producing more decks and podcasts is not monetizable value. Removing bureaucracy via AI just recycles the displaced into new “bullshit jobs” (the communist cloakroom attendants example), and Chesterton’s Fence applies: do not remove it until you know why it exists!

• The influencer economy functions as an algorithmic lottery that happens to distribute wealth.

• Real productivity must show up in the material world, otherwise it is a pointless exercise.

Russia, Poland & Europe’s Eastern Front (Minute 46)

• Izzy rejects the “Russia is winning” war-economy argument: intangible goods like freedom and a conscription lottery for your son are not priced in.

• Anxiety in Poland about escalation is real; Poland is among the EU’s biggest defence spenders and aiming beyond 5% of GDP.

• Downsides of the defence build-out: diminishing returns unless you export arms, immature procurement networks (corruption risk comparable to “the UK PPE scandal times ten”)'

• Legacy communist-era structures in the Polish military, and historical military Keynesianism where spending substitutes for welfare rather than capability.

Europe’s Reconfiguration Play (Minute 51)

• Post-Brexit and with Ukraine entering, the EU is more of a planned economy than people admit. Draghi’s roughly 800 billion euro plan needs a catalyst; hyped Russia fear is a useful excuse for dual-use investment.

• The honest caveat: in a war economy everything is dual use and everyone de facto works for the state, but democratic capitalism is also poor at manufacturing needed social goods, so some over-egging of the fear may be a defensible means.

Show Notes

Correlation chaos

• The theme of the week is market divergences and breaking correlations

• Correlation breakdowns are hurting levered macro and pod shops; unlevered beta investors unaffected; machine models handle slow correlation shifts but not fast ones. Heavy “taps on the shoulder” in fixed income relative value pod shop world

• Benny reiterates his bond bull call from prior weeks despite rates rising since; long end blew through 5 percent and payer swaptions are expensive with few willing sellers

• Treasury market now in “price discovery” mode; the level of yields, not just rate of change, suddenly matters

• Why equities have not blown out despite rising rates: markets run in regimes, the 1970s saw stocks and rates rise together, and the post 2022 regime still has more COVID era liquidity and credit capacity than expected

• “Zombie return of risk parity” - stock bond inverse correlation reappeared this week for the first time in years

• Dominant narrative is US growth beating expectations on data center capex; a cohort believes Warsh has two to three more hikes in him

• Capex booms flatter earnings via capitalized costs versus fully recognized revenue; Julian Robertson’s 1998 fund closure cited as precedent for being right too early

• S&P implied vol calm at 13 while MOVE index exploded; equities falling as bond vol spikes

Positioning and fiscal dynamics (Minute 10)

• The Squirrel bought more bonds this week; over 15 percent of his NAV is in trend followers currently max short fixed income, so a long adds some balance

• Administration likely to intervene in markets; Benny’s biggest career mistake was fighting the machine post GFC

• Bond market signaling it may not fund deficits indefinitely; must think dynamically about how a yield spike forces private capex cuts that then roll growth over

• Squirrel believes hyperscaler spenders will blink earlier than consensus; the economy ex data center capex is in much worse shape

• September hike may, with hindsight, prove the one that pushed things over the edge; consistent with historical Fed “death shot” behavior

AI capex and Meta (Minute 13)

• AI on and off balance sheet commitments now $3.3 trillion, 40% higher than the roughly 2 trillion figure cited on the pod three months ago; Latest weekly AI capex token pricing monitors here 👇.

Llm Provider Pricing Top30 Analysis

1.68MB ∙ XLSX file

Download

Download

Capital Misallocation Llm Chartbook

3.19MB ∙ PDF file

Download

Download

• Maintenance capex keeps rising; hyperscalers need real revenue, hence Anthropic deals with Salesforce and Accenture and OpenAI being the fastest ever to $1 billion of ad revenue; target pools are $1 trillion of global advertising and $1.5 trillion of software

• Tale of two sectors: S&P near all time highs with AI at roughly 51 percent of the index, while consumer, housing and rate sensitive names (Rocket Mortgage owned) sit near April 2025 Liberation Day lows

• Meta’s Muse agent is the story of the week - Meta ticks all four boxes of data, distribution, brand and owned compute; stock ripped from a $519 post earnings low last summer to just under $800, mostly in two weeks, on heavy call buying

• Divergence watch: Nasdaq vol crushed from 28 to 22, but credit is not confirming; CoreWeave spreads around +808bps versus +950bps prior - out from +460bps in May

  

• Oracle declared force majeure on New Mexico power plants; Ellison pledged an extra $9.6 billion of stock against the Paramount deal

Crude Oil (Minute 22)

• IEA flagged a shrinking global deficit and soft China demand; oil sold off while rates rose, another divergence on the chart versus the 10 year note

• Squirrel is much smaller in front month crude than a few weeks ago; holds roughly 5% of NAV in but advises new entrants to be appreciably smaller - more comfortable underwriting energy equities than the physical commodity here

Agentic AI and banks (Minute 25)

• 🐿️ just back from four days at a conference in the Peloponnese, Greece where he presented on agentic AI finance bots and bank deposit franchises

• Around $23 trillion of global bank deposits earn zero interest, worth roughly $170 billion of net interest income to global commercial banks; Muse going mainstream at the beginning of last week made the 🐿️’s month-old slides feel a bit stale

• Apollo’s Torsten Slok chart shows fintechs like SoFi and Interactive Brokers paying around 4% on cash versus legacy deposit franchises; classic Silicon Valley overpaying for market share, but frictionless finance bots do pose a real threat to deposit inertia

• Killer chart: net interest income growth minus deposit growth over four years against price to tangible book; banks in the red circle are most at risk of deposit franchise devaluation; big subscriber note coming

  

• Important to look at the entire financials complex through the prism

• On this note, PayPal is the classic battleground stock: binary outcome justifies the compressed 9x times PE - Benny already owns it.

Plus: BofA's 33% earnings growth met with total silence, the AI credit recycling machine keeping index spreads pinned, the dispersion trade propping up a lifeless VIX while single stocks get murdered, and why energy at 3% of the S&P today could be 8 to 12% within five years.

The weekly Top 30 LLM models chartbook and spreadsheet is below the paywall.

Show Notes

Episode 88 Slide Deck

2.93MB ∙ PDF file

Download

Download

The Fed Week Gauntlet

• Warsh hikes and the Fed comes out surprisingly hawkish. Squirrel bought front-end bonds (levered 2-year note exposure via ) before the meeting but could not bring himself to add after, leaving him in limbo on rates.

• Market reaction: stocks and bonds both sold off post-presser, around a 15bp reversal from the pre-meeting move, followed by a 5 to 8bp rally on Thursday. MOVE index ripping higher.

• The energy spike helped set off the rate volatility, echoing the March war-driven bar event.

Earnings, Energy and Positioning (Minute 3)

• Bank of America posts 33% earnings growth and the stock is flat: can earnings growth ever match prices at these levels?

• Bond sentiment at rock bottom; the AI capex mega trend keeps driving earnings, and index credit spreads stay subdued despite chaos under the surface.

• 🐿️ has trimmed a large crude position that was hedging his long risk assets, but is not ready to add risk back. He describes himself as in no man’s land.

• US Buffett indicator at 244% of GDP, way off trend. Professionals have to be involved, but retail viewers do not.

The Case for Bonds (Minute 8)

• For anyone not benchmarked: sit in bills or buy bonds. No credit risk, capital back, plus an embedded call option if the market rolls over.

• Bonds are in “most hated” status: BofA high net worth allocation to bonds at a multi-year low, comparable to energy in 2020 or tech in 2022. Consensus assumes AI issuers and governments will issue bonds forever.

• Undersupply of duration at the long end as issuers refuse to lock in 5% coupons; payer swaptions are expensively priced, showing the market is positioned one way.

• Wall Street has made a habit of getting forecasts wrong, which supports the bond call option logic.

Priced-In Earnings and Distortions (Minute 12)

• Valuation is not a momentum signal, and at this stage individual earnings prints are almost unhelpful to process.

• AI credit mechanics: hyperscaler credit spreads are blowing out while index spreads stay pinned. Passive inflows hit sector constraints, so money gets recycled into weaker names (Boaz Weinstein publicly complaining about Stellantis spreads versus its terrible credit metrics), and banks hedge for BIS regulatory reasons.

• The shift from active to passive: unlevered beta went from roughly 1% fees twenty years ago to nearly free, removing the human “speed bumps” that used to absorb risk in the system.

• Dispersion and the hollowed-out middle: pod shops and near-free passive dominate, single-stock and cross-sector volatility at 10-year highs while the index sits in the low teens, and implied correlation stays floored by the dispersion trade.

• Pod shops and private equity sell smooth return streams that pensions need, but the old mental and market models no longer match a grossly distorted market reality.

Read more

We get into the well productivity slowdown that signals rising development costs, the “China collar” and the renminbi signal, why big tech’s trillion-dollar capex binge is a replay of big oil in 2013, the refining secondary-unit squeeze, and where we are in the energy equity cycle (inning four or five, with parabolic euphoria still ahead).

Show Notes

Framework: productivity, not barrel counting

• Rob’s view was shaped by three mentors: one on history, one on secular oil price direction, one on stock picking via incentive structures.

• Core philosophy: 99% of Street research is “barrel counting” (supply vs demand), which captures short-term moves but misses long-term cycles.

• Evidence: year-over-year oil supply and demand growth have tracked each other since 1866 (1970s demand growth went from +8% to -1% while oil rose from ~$5 to ~$35; the 2000s saw steady matched growth but oil went from ~$10-15 to $140). Supply always meets demand; what matters is the marginal cost of the barrel called upon.

The Eureka Chart (Minute 5)

• Well productivity (oil output divided by producing well count) is the key long-run metric, built from company disclosures.

• When productivity growth slows, development costs for the marginal barrel rise; non-OPEC well productivity growth (June/July) is the slowest ever outside of recessions, implying development costs must rise and the industry shifts to higher-cost basins.

• Analogy: unit labor costs, i.e. productivity growth of 14% vs cost growth of 5% is a positive spread; when productivity slows below cost growth, the economics flip.

• The development cost index has roughly a 0.9 r-squared to oil prices back to 1932, the strongest model he has found.

Marginal cost framework (Minute 8)

• Production costs are backed out of company disclosures (Exxon, Chevron, BP, Shell, Total and predecessors like Mobil, Texaco, Gulf), combining finding and development costs, lifting costs, and taxes; cost of production vs oil price shows ~0.94 R-squared since 1978.

• Saudi Arabia: lifting plus F&D around $11/bbl, but high tax take. Running large fiscal and current account deficits (as now) raises their marginal cost requirement, supporting the bull case.

• Oil price divided by marginal cost defines the trading range: when price is high relative to cost, four things follow (rising non-OPEC supply, OPEC cheating, demand destruction, central bank hikes); the reverse set signalled his bullishness in June and September last year, before the Iran conflict.

• Current read: oil looks “peaky-ish” around $110, floor near $70 absent a COVID or GFC-style shock; he is not selling because “you never sell oils in September,” with the seasonal exit point being spring.

Dollar, liquidity and the global credit (Minute 13)

• Weak dollar is still needed for oil to do well; a strong dollar sucks global liquidity and acts as global tightening.

• Best gauge is world M2 translated into US dollars; global money supply bottomed mid-2025 and has been accelerating since.

• Risk: the Fed is back in a tightening cycle, and historically oil stocks get roughly an eight-month runway before going wobbly; the 1999-2000 “double tightening” (ripping dollar plus hikes) is the cautionary template.

• Historical digression: the 1990s oil bear market lasted about seven years longer than it should have because former Soviet Union supply (Lukoil F&D costs around $0.40/bbl) collapsed the marginal cost; growth-minded, nationalization-scarred majors reinforced it. Venezuela is the modern parallel supply risk.

Incentive structures: oil vs tech (Minute 20)

• Investor-led reform shifted oil exec pay to free cash flow, returns on capital employed, and TSR benchmarked against the S&P 500, not just peers; this disciplines capital allocation because PMs allocate across all 8,000+ listed companies, not within sectors.

• Tech’s incentive structures (Amazon, Google, Microsoft, Apple, Meta) are 100% growth-based with no capex control or ROCE metrics; over $1 trillion of capex mirrors exactly where big oil was in 2013-14 when ~50% of E&P pay was tied to reserve and production growth.

• Shale wells and GPUs share the same profile: roughly 80% of value captured in the first 18-24 months.

Refining and secondary units (Minute 27)

• Refiners have the best incentive structures in oil and gas (post-2008 bust, separation from majors, capex control, buybacks), which is why no new US greenfield refinery since 1977 and why refining is a “rock star” now.

• Common throughput-over-nameplate utilization misses the point; profits come from secondary units that upgrade bottom-of-the-barrel material into gasoline, jet, and diesel.

• Light product demand is set to grow roughly three times faster than secondary unit capacity; Atlantic Basin closures (Europe, East and West Coast) and China’s EV-driven halt to refinery building compound it.

• The Iran war removed medium-heavy crudes that are best for distillate; Venezuela and Canada help but not enough, which is why distillate cracks are so strong.

China collar and the RMB Signal (Minute 31)

• 🐿️ asked for Rob’s view on his “China collar” theory (monopsony buying power capping and flooring the crude price) - he gives it credence because oil tends to bottom when the RMB strengthens; USDCNY has firmed from ~7.3 in April 2025 to ~6.7.

• Rule of thumb: buy what China is buying (gold, then barrels), stop making things that China starts making - LLM tokens may have the same commoditization risk.

Energy equities: cycles and positioning (Minute 34)

• Energy is ~3.5% of the S&P 500; Rob puts the equity cycle in “inning four or five” of his DOPE (Doubt, Optimism, Parabolic & Euphoria cycle). Doubt wave ran from 2020 to 22. We are in the Optimism phase, anticipating a pullback ahead of Euphoria.

• Drillship fleet consolidation chart: that 1990s bubble was Venezuela’s Lake Maracaibo barge rig build-out, which triggered the Venezuela-Saudi spat and eventually Chavez.

• Transocean’s incentives have finally shifted from uptime/day rates to free cash flow; day rates bottomed in 2019-20 at ~50-55% utilization.

• Offshore operating leverage is underappreciated: late-cycle profits go parabolic as fully depreciated old assets (1970s boats by 1978-79, 1970s jack-ups in the 2000s) return to work; today’s parallel is North Sea semi-submersibles and a re-engaged BP - an opportunity for Transocean .

• Long-term target: oil eventually moves towards $240, the inflation-adjusted 2008 high; development costs historically CAGR at mid-teens (14-18%) through upcycles, and energy could return to ~10% of the S&P 500 (vs ~30% in the 1970s), a threefold move in weight.

• Sell signals to watch: oil to the top of the oil-to-SPX logarithmic channel, and producer returns on capital back to the 20-25% level where every cycle peaks.

• On pullbacks: capital discipline plus rising marginal cost puts the floor near $70; his Crude Chronicles Energy Index shows 2023-25 was the first bear market where earnings declines merely matched (not exceeded) oil’s year-over-year decline.

• Valuation anchor: oil and gas market cap at ~7.5% of US nominal GDP implies 6-8% real (double-digit nominal) annual returns over the next decade; R-squared of 0.62 since 1912 (0.71 excluding the 1970s). Keeps him bullish but grounded.

Rob's Full 3Q Slide Deck

18.3MB ∙ PDF file

Download

Download

Show Notes

Quite a weekend in ‘AI La-La-Land’!

• Recap of a wild weekend in “AI La La Land”, triggered by another AI “safety / off ramp” announcement.

• Surely we are in a world of diminishing returns from these ‘Axios-style’ safety / ‘FUD’ headlines. Can the off ramp play out elegantly from here?

• Sunday night futures open with Nasdaq offered and bonds bid - kinda makes sense!

• Grown-up corporates do not want to hand data to model companies; the move is toward private, self-trained, off-the-grid LLMs, citing the Latham Watkins announcement covered in the FT.

• That trend queers the monetization pitch for the model labs; data security is the number one driver, alongside AI ROI and cost concerns.

• 02:27 Where rent can be extracted in the stack: NVIDIA, other chip makers, and power. LLMs themselves may eventually be given away. A return to a “capital world” with software getting squeezed, though not dead.

• Warren Pies’ chart on software productivity rising exponentially within the sector; the “neoclouds” feel like they are about to get ‘smoked’.

Fed meeting setup and the CapEx monitor

• Positioning ahead of the big Fed meeting on Wednesday, with the AI story the main driver of an elevated S&P.

• Benny’s CapEx monitor: gross dollar spend has been strong but peaked in Q2.

  

• Normalized for GDP: roughly 175 basis points of incremental GDP share added over three years, potentially 3.5 percent of GDP by 2027, and up to 70 percent of incremental growth by some estimates.

• AI Capex is now the dominant growth driver, previously it was government; the 6 to 7% percent deficit now just keeps flow going rather than adding growth, and the AI spend is a key reason bonds have not performed (paper supply plus real rates).

• Reported figures only cover the five big spenders; off-balance-sheet and private credit data center spending (for example APLD building shells) makes the true number even larger.

Rate repricing and cycle analogs (Minute 6)

• The market has repriced from cuts to roughly four to five hikes across the curve since February, the peak being the day before the war started.

• End-of-cycle analogs: oil rips late in cycles. The 2007 comparison (oil versus December 2009 Eurodollars, peak summer 2007, US housing and China commodity boom

  

• A similar sized rate repricing) and the late nineties version (Russia and LTCM, peak hawkishness around summer 2000).

• Crude had a proper move last week; Squirrel trimmed length into Thursday and Friday with no regrets, but it feels like it wants to go higher.

• Nearing the call strike on the China collar; China is back buying but could stop anytime. A peace announcement would knock roughly $10 off Brent quickly. Saudi pipeline outage: four to six days of inventory, roughly a 4% supply hit.

Blind Squirrel Macro

Pricing the "China Collar"

Another early edition. Given our overweight positioning in energy, Part 3 of the 🐿️’s annual stock take considers the implication of China’s emerging control of oil prices…

Read more

3 months ago · 26 likes · 7 comments · The Blind Squirrel

• Inflation is everywhere, but the economy would struggle fast with higher rates - the question is whether Wednesday is the “death shot”.

Bonds: time to buy (Minute 13)

• 🐿️, famously a bond hater, is aligning with “Benny the Bond Bull”.

• Curve choice: avoiding the long end due to term premium uncertainty; the seven to ten year bucket has a supply gap dating back to Paul O’Neill ending 30-year issuance in 2002, plus positive convexity from mortgage hedging flows. Benny wrote this up as the “Mind the Gap” trade.

• Framework: convexity per unit of duration; simple alternatives include (7 to 10 year UST ETF).

• 🐿️’s weapon of choice for the beta portfolio: , a five times levered play two-year note futures, giving duration exposure without a big term premium bet.

• Strategic logic: bonds are a way to play AI rolling over. If CapEx growth stalls and the equity market rolls, the wealth effect (households with years of wages parked in the market) and the physical CapEx economy hit growth simultaneously and feed back on each other.

Short Utilities (Minute 18)

• Utilities are no longer trading like bond proxies; since the war began, XLU has ignored rising ten-year yields and become “AI stonks”.

• The XLU dividend yield now sits about 225 basis points below the ten-year yield, versus the old “Treasuries plus a kicker” relationship; no bond-proxy protection left.

• Politics: bashing AI data centers is now bipartisan. Two-thirds of XLU constituents operate in states with legislation to block or delay data centre development. Only the IPPs (Constellation, Vistra, NRG), around 12 percent of the index, can capitalize on scarcity.

• Retail electricity prices have mooned since the public launch of ChatGPT (inset chart borrowed from Kevin Muir).

• Free cash flow across XLU electricity producers is negative except for the IPPs; most face almost five years of negative free cash flow.

• The trade rounds out nicely: ‘short utes’ paired with the levered two-year note ETF (TUA) - may give the position more ‘staying power’.

Hyperscalers and the model labs (Minute 23)

• Anthropic chooses NASDAQ, OpenAI pushes its timeline out; commitments already made matter for future Magnificent Seven earnings, and paper mark-ups that boosted earnings become headwinds if marked down.

• Tailwinds and headwinds for hyperscalers: a frontier-level slowdown cuts CapEx (market positive) but forces write-downs in equity stakes against a general growth slowdown.

• Advertising: OpenAI was the fastest to a billion dollars of ad revenue ever, in a roughly one trillion dollar global market that is only GDP-plus growth and cyclical; a knife fight with Meta and Google. Stop treating the Mag7 as a monolith!

• Most hyperscaler forward CapEx supports training runs for the two jumbo private customers (OpenAI and Anthropic); inference earns ‘decent vig’ but is not the driver. If local models keep spreading, cloud growth may disappoint.

• The 26 trillion dollar enterprise TAM chart from the SpaceX S-1 “will age terribly”; every Fortune 500 company bringing AI in-house (Latham Watkins trend, more anecdotal cases) takes a bite out of that TAM.

The Fed decision (Minute 28)

• Scenario walk-through: market pricing around 86 percent for a 25 basis point hike on Wednesday; both Benny and 🐿️ see it as a done deal.

• Credit to Warsh for making meetings “live” again; he was hawkish at Jackson Hole and has kept markets guessing versus the Powell era.

• If they did not hike, term premium becomes a problem and the long end is ‘lost’ quickly. Both hosts are buyers of bonds into a rate hike because they think it is the ‘death shot’ (and probably the last one).

He later launched The PaineWebber Europe Growth Fund, the first mutual fund specializing in both Western Europe and in the former communist countries of central Europe. In the mid 1990s, Mr. van Caloen led the international investment effort at Provident Capital Management and later served as First Vice President of Schroders in New York.

Show Notes

Underwriting Venezuela

9.13MB ∙ PDF file

Download

Opportunity overview.

Download

Introducing Herve

• Formative lesson: Scudder’s Korea Fund was the first into a closed Korean market, and his Paine Webber fund was the first US mutual fund investing in Eastern Europe; markets that open up offer outsized potential.

• Poland’s post-1989 “shock therapy” was widely predicted to fail but worked; Poland is now the fastest-growing economy in Europe, proof that rule of law plus free markets can rebuild a country in one generation.

• Unlike Poland starting from scratch, Venezuela is a capitalist system in “suspended animation” for 30 years.

Political Transition

• Current setup: Delcy Rodriguez as interim leader with Washington calling the shots (Rubio as de facto “Viceroy”, ultimately Trump); necessarily temporary.

• Institutions must be built before elections (most expect 2027) so results are accepted by all sides, including Chavistas as the future opposition; a Mandela-style “truth and reconciliation” approach rather than revenge cycles.

• Controversy: hydrocarbon licenses awarded without competition to Chavez-era businessman Betancourt; a pragmatic “he gets things done” logic, but it is costing Trump local popularity and adding pressure to accelerate elections.

Macro setup and banking (Minute 8)

• The math: GDP is roughly one third of pre-Chavez levels and must triple to normalize; market cap is 1.5 to 2% of GDP versus a roughly 40% Latin American average, implying around 60x market cap growth just to reach normality.

• Listing wave ahead: companies are starved of capital, banks are not lending, and selling via the exchange attracts roughly 1% capital gains tax versus 34% on a private sale; Herve’s Caracas partner (the largest local broker) is working on two IPOs.

  

• Why banks cannot lend: inflation still 400 to 500% and reserve requirements raised to 73% have frozen credit; no mortgages, no business loans.

• Workarounds filling the gap: BNPL booming, rapid informal dollarization, a physical dollar shortage, and Thiel-backed Erebor Bank distributing USDT stablecoins, which is channeling dollars into the system.

• Sanctions: no OFAC bar on foreigners or Americans buying Caracas stocks, provided companies are not government-linked; sanctions are being lifted sector by sector, oil first, mining next.

• Oil rebuild will be slow and expensive (heavy crude needing pre-refining before Texas); Chevron has flagged roughly $7 billion and Betancourt is talking about $100 billion, though his deal may breach the constitution; risk of a local backlash against “Yankee” takeovers.

• Diaspora: over 8 million emigrants, many successful abroad, are keen to invest or partially return (Margarita Island property, tourism, pharma distribution).

• The pitch to skeptics: pre-Chavez Venezuela was Latin America’s richest economy, with a US-educated elite (MIT, Harvard) and an America-facing culture (baseball, not soccer).

Stocks and Currency (Minute 24)

• Strategy: the classic EM starter kit (banks, brewers, cement) applies but will quickly broaden; anchor holding Banco Provincial trades on a single-digit earnings multiple and 2.6x book, valued on what the bank could become, not its current state.

• BBVA (Provincial’s largest shareholder) is the only foreign bank that never left Venezuela (even Citi quit) and should benefit from that loyalty.

• Rum not beer: Ron Santa Teresa, already exporting to 80 countries via Bacardi’s network, earning dollars with bolivar costs

• Unusually sophisticated candidates for a frontier market: Ridery and Yummy Rides (local Uber/Lyft equivalents), Socado Coffee (“ten times better than Starbucks”)

• Foreign competition is arriving: Colombia’s Nutresa bought Tio Rico (ice cream, roughly 50% share) and Uber entered food delivery via Delivery Hero’s PedidosYa; local champions need capital to fight back or be sold out.

• Herve’s Mercator fund: equity only, aiming to be first to bring liquidity to the exchange; small first tranche, then a larger one if it takes off; the exchange president is fielding tens of companies seeking capital.

• Third Core holding: the Caracas exchange itself, a listed “toll booth” with only about $150k of daily volume, so pure upside; US market cap to GDP above 100% shows the long-run headroom beyond the 40% normalization target.

• Currency risk answer: dollarization; the gap between the official and parallel rates has narrowed from about 30% to roughly 15% and is heading toward zero.

• Inflation is easing (August monthly under 10%); the real economy transacts in dollars, restaurants use a “priced in euros” loophole to beat the official rate, and stablecoins are likely to play a major role, echoing the Turkey/Bitcoin dynamic.

Politics and Sizing (Minute 40)

• 2027 elections are plausible as locals grow impatient; the goal is a strong constitution and institutions, curbing corruption and army influence, and binding the economy to the US so reversal is hard; in the end “only one person calls the shots today and it’s Trump.”

• Chavistas retain positions of power; tension between reconciliation and revenge (his driver’s father wants revenge, the son wants to move forward); expect punishment for the worst actors and quiet exile for cooperators.

• Portfolio framing: a convex call option; Poland rose about 10x in the six years after its market reopened; buy now and do not look at it for five years, because an election announcement would gap the market with too little liquidity to chase.

• Venezuela Market context: the exchange jumped roughly 140% in dollar terms after January 6, then moved sideways as investors await confirmation of the transition.

• Sourcing stock: hunting secondary blocks and IPO allocations now, with government privatizations to come; listed companies are happy to dilute given how scarce capital is.

Follow Herve on LinkedIn and Twitter.

The Show Notes (plus bonus extras at the end)

🐿️’s “20-for-20” and Resources

• The latest Twenty for Twenty sector installment on resources (energy plus materials) was published Saturday

• A hypothetical 2006 version would have been stacked with financials, energy, materials, dry bulk and housing during the China capex boom.

• CRB spot index had an all-timer four-year run into the 2008 oil spike, but buyers of the DJP commodity ETN at its 2006 launch only broke even at the start of this year.

• Product launches are a late-cycle signal: when ETFs or bank products arrive in a hot theme, the trade is usually over

• Today’s resources pitch is different from 2006: a scarcity, capital discipline, debasement and sovereign resilience story rather than a China demand shock.

• The resources sleeve in Twenty for Twenty will likely be well above the current MSCI ACWI 8.5% index weight.

• Running resources through the mosaic of ten macro themes:

• Valuation caution: Freeport-McMoRan’s forward EBITDA is around two standard deviations above its 20-year average; Southern Copper and peers show the same dynamic.

• Glencore included as the wild card despite its coal business scoring poorly on the macro themes (thumb on the scale).

• Copper futures show a crowded bullish tape on COT positioning (large and small specs heavily long, elevated open interest). But it’s still an awesome chart!

• S&P data shows a 20-year famine in tier-one copper discoveries; recent big finds are in LATAM, Asia-Pac and Africa - i.e., not tier-one jurisdictions.

Subscribe now

DCM (and AI earnings) Bubble (Minute 14)

• Banner year for debt capital markets (DCM) bankers as 2026 is lining up to be the biggest issuance since the GFC (with an inflation caveat) - even if Mag 7 issuers are likely paying charity-level fees for market share.

• Last year’s call that hyperscalers would blow out their balance sheets and free cash flow has played out in underperformance; Meta now trades at 9x next year’s EBITDA

• Mag 7 has flipped to a negative correlation with momentum over the last four weeks, a dynamic only seen in “crash-ish” periods.

• H/T ’s charts: strip out markups on big private AI holdings and much of the “earnings bubble” is mark-to-market

Subscribe now

Xmas Tree of Death (Minute 18)

• Benny has built a mathematical signature model for the “Christmas tree of death”: money and leverage crowd in, shorts step aside, then an air pocket. (e.g., US Steel 20 years ago, silver this January, Palantir last fall and Alcoa this year)

• Kospi check-in: possibly a fifth wave. Korean regulators now require retail traders to pass a ‘driving-test’ equivalent before trading double or triple leveraged products - probably not a bad idea.

• Flows: foreigners buying while locals got wiped out, with 3% of Korean account holders margin called in short order; echoes 1997 hot-money dynamics.

• Similar in Japan : foreigners bullish, locals and corporate Japan wary of Takaichi’s reform promises (Charlie Brown football syndrome)

• The only bid in Korea is buybacks: Samsung and SK Hynix are repurchasing stock they just issued. Hynix did a $26B ADR then a $1.5B buyback weeks later - a corporate finance head-scratcher!!

• Google’s early-June equity raise read as tactically front-running the SpaceX, OpenAI and Anthropic raises; better to pull the Band-Aid off early on funding the capex cycle.

• Commodity bear markets are dark places where no one will give you capital; cyclicals can only raise money when the ‘ducks are quacking’. Hynix’s buyback looks like stock-price management after a rapid drawdown (red flag?).

• Mid-innings Xmas Tree ‘mechanics’: crash, stabilization, then beta decay if flows don’t re-accelerate; heavy supply sits above the right shoulder with 90%-plus momentum concentration, and nobody gets out at the top.

Bonus Extras

20260905 Benny's Weekly LLM Chartbook

2.92MB ∙ PDF file

Download

Weekly update for subscribers.

Download

20260905 LLM Provider Pricing & Top 30 Model Intelligence Analysis

1.66MB ∙ XLSX file

Download

Download

Show Notes

Yen intervention and “Oriental Maggie”

• Is there real coordination between the MOF, US Treasury and others behind the yen moves of recent weeks?- “the last 3 to 4 weeks have seen a weaponization of the yen,” with Bessent, Trump and Tokyo speaking with one voice; 160 in dollar-yen is the threshold of pain.

• The real pressure is political: polls show 90 to 95% of Japanese feel the purchasing power pinch, and every 10 yen of depreciation adds roughly 0.4 to 0.5 percentage points to CPI.

• One in three households are pensioners, and the pension “macro slide” means payouts rise only about 1.9% when CPI rises 3%, so pensioner purchasing power is being decimated.

• Backdrop: inflation expectations, a BOJ behind the curve, and Takaichi’s aggressive fiscal program mean bond issuance keeps rising.

• Portrait of Takaichi: obsessed with power, a ruthless Machiavellian on cabinet control, and un-Japanese in style, deciding by phone and email rather than consensus building; “she is the nail that stands out”.

• In February she ended Japan’s voluntary restraint on exporting offensive weapons in a single day, something Abe tried and failed to do; frigates (to Australia), submarines and missile defense can now be exported, and industrials rank dual-use and defense tech as their number one or two growth opportunity.

• Fiscal revolution: multi-year budget commitments beyond the old carve-outs (defense 5 years, education 3 years), now 10-year programs for quantum computing, robotics, AI and anime/content.

• Scale: Tanaka Kakuei’s 1970s restructuring started at 20% of GDP; Takaichi’s program is about 60% of starting GDP, with decade-long commitments to data centers, robotics plants and submarine plants.

• Versus Abenomics: Abe faced a minus 4% output gap and deflation risk; today the output gap is positive (BOJ estimate around plus 1%), rates are rising, and the fiscal mix is investment-led through public-private partnerships (10% government, 90% private, with debt guarantees). This is not consumption-led reform.

• Onshoring logic: at 160, the yen makes Japan viable as an Asia-Pacific export hub, so corporate savings should come home.

BOJ constraints and capital flows (Minute 14)

• Intervention theater: the BOJ did nothing at the last meeting, then Ueda said he was convinced inflation accelerates in H2 (so why not hike?); Bessent’s staged Reuters photo was “all-time trolling.”

• Why the BOJ is slow:

  • fiscal constraint: debt at 260% of GDP means each 100 basis points of rate rises adds about 1 percentage point of GDP to interest expense; interest already absorbs about 26% of the national budget, heading towards one third; the real constraint is political, since less discretionary budget means less power to allocate.

  • financial fragility: regional banks, credit cooperatives and credit unions lack people, systems and compute after 20 years of zero rates (only seven JGB traders were left when Ueda arrived); credit co-ops just posted their worst results since the bubble collapse, with bankruptcies starting; big insurers reported unrealized losses near 7% of AUM in the June quarter.

• The implosion risk is not the GPIF announcing a shift out of Treasuries (a pipe dream); - it is the forced selling of USTs to cover domestic losses.

• Reality check from Jesper’s presentation to the GPIF allocation committee: the fund must be able to liquidate 50 to 100 billion dollars in two days without moving price, which only the UST market allows; selling even 5 billion of JGBs moves the needle.

• The JGB market has been re-privatized: roughly 80% of transactions are now private-to-private, versus 80% BOJ-dominated under yield curve control.

• Ten-year JGBs around 3% are genuinely attractive: the insurers’ guaranteed rate (yotei) is 1.6%, so 3% lets them clip coupons and play golf, consistent with 1 to 1.5% trend growth plus 2% inflation; a floor under JGBs is established.

Domestic trust and the bureaucracy (Minute 26)

• The dirty secret - the only buyers of Japanese equities are foreigners and corporations via buybacks; domestic asset owners have been net sellers for 2 to 3 years.

• Why: America is seen as the nerve center where capital is treated best (tax, rules, liquidity, pro-capital politics), while Japanese investors fear domestic rules will change, especially capital gains tax, with AI-driven labor displacement creating pressure to tax capital to fund a basic income.

• Takaichi is not capital friendly on tax: she cut the food consumption tax but did nothing for corporations; the governance code update empowers boards to reject takeover bids; the founder minimum tax threshold was cut from 300 to 160 and the rate raised.

• The civil service (“Men in Grey Suits”) is bought in: Abe vetoed about a third of senior ministry promotions each year, while Takaichi rubber-stamped everyone, so METI and agency bureaucrats write the policy detail, which is bullish for the longevity of the reforms.

• Jesper’s meta case for Japan: (1) labor scarcity forces companies to rethink incentives and retention, and (2) both big partners have become unreliable, with 32% of profits tied to the US and 18% to China (Toyota just announced another China factory).

Geopolitics, banks and sector calls (Minute 36)

• Nightmare scenario: a Trump-Xi deal opening the US market to Chinese products and joint ventures (BYD building US factories with Ford) would gut Japan’s most profitable market; Japanese car makers were happy to pay the 15% tariff (effectively 7 to 8%) as protection to keep Chinese cars out.

• Hyper-speculative barter: China gets US market access and technology transfer in exchange for reallocating sovereign reserves into US Treasuries.

• Japan is hedging through Indo-Pacific alliances: Australia (11 frigates, 3 built in Japan and the rest in a Japanese-revitalized Australian shipyard) and a deepening India corridor.

• Banks: forced to become utilities under zero rates, they now have net interest income from the steepener; the bull case is the credit multiplier from PPP loan demand (the 370 trillion yen, 60% of GDP, reindustrialization will be funded by PPPs, not JGBs) plus fee income from MBOs, M&A and IPOs.

• Single names: MUFG-Morgan Stanley is the success story, making 1 plus 1 equal 4; watch Sumitomo (SMFG), whose Jefferies stake is likely to grow into a similarly powerful franchise.

• Deposits: sticky so far, but the mattress money (goldfish stuck in jelly) is becoming unstuck through inflation and deposit competition (Shinsei Bank offering 1.7% on 3-month deposits; SBI and Shinsei very aggressive); roughly 1.3 times GDP shifts as the boomer generation dies and heirs move regional bank money to the cities.

• Bullish Tokio Marine: first Japanese asset manager with pay-for-performance for fund managers, now partnered with Berkshire Hathaway; could become Japan’s first globally relevant asset manager within five years.

• Generational flip: the average outgoing CEO is 68, the incoming one 52; the Ice Age generation is taking over corporate Japan and is comfortable making decisions, doing M&A and even MBOs, happily financed by the mega banks.

• Japan has real assets, Hello Kitty does three times Disney’s revenue, and 38 trillion yen of the 370 trillion program targets anime and content; the 19th-century media companies could consolidate into global powerhouses.

• Energy is the pinch point: Komatsu and peers cite energy cost certainty as the top obstacle to reshoring, nuclear restarts remain slow, and if the Strait of Hormuz stays closed, Japan’s strategic oil reserves run out around January to February

The 2026 Prediction that Missed - Better Luck in 2030!

Benny & The Squirrel (@bennyandthesquirrel) | TikTok

https://www.instagram.com/benny_and_the_squirrel/

The show has expanded on the socials! Follow us on Instagram and / or TikTok (if you do that kind of thing!)

Benny & The Squirrel (@bennyandthesquirrel) | TikTok

https://www.instagram.com/benny_and_the_squirrel/

Benny & The Squirrel (@bennyandthesquirrel) | TikTok

https://www.instagram.com/benny_and_the_squirrel/

The show has expanded on the socials! Follow us on Instagram and / or TikTok (if you do that kind of thing!)

Benny & The Squirrel (@bennyandthesquirrel) | TikTok

https://www.instagram.com/benny_and_the_squirrel/

Jeff learned his trade on the knee of his dad who was the CEO of Target for over a decade. You could say US retail runs through Jeff’s blood. You can find his subscriber content over at StockMarket TV as well as at Wall Street Beats where he is a partner there with Benny.

The discussion argues that the US consumer is materially healthier in observed spending data than the prevailing macro narrative suggests. The core investment implication is that heavily shorted, low-expectation retail, restaurant, leisure, and turnaround names can produce outsized returns when results merely show that conditions are “fine,” particularly while capital remains concentrated in AI and semiconductors.

The show was recorded at 6pm EST on Thursday 13th August.