THIS WEEK · 10 PODCASTS · WEEK OF JULY 27, 2026
This week we bring you ten conversations that put the entire market regime on trial — the AI buildout absorbing capital at a historic rate, the two assets at the center of the retail mania, the physical energy markets being drained while paper prices look the other way, and the credit and housing consequences already visible in the data. Steve Eisman convenes Dan Ives and Gil Luria for the definitive winners-and-losers session on AI: Ives' year-three-of-a-decade-buildout bull case, Luria's dislocation hunting — he stepped off the Oracle bandwagon at the very top and now flags $630 billion of backlog priced at zero — and Eisman's three bear cases on capital intensity, missing moats, and a Chinese-triggered price war, all set against more than $100 billion of real AI revenue that did not exist two years ago. Gordon Johnson takes the other side with a forensic teardown of Tesla and SpaceX, arguing that Tesla's own disclosed robotaxi numbers support an $840 million addressable market against roughly $700 billion of market value. George Noble, on Chris Martenson's Finance U, supplies the system-level diagnosis: rallies are liquidity events, the KOSPI's 31% break in 17 days is the movie about to run in reverse, and the only defensible response is to own what cannot be printed. Wes Gray brings the discipline — you cannot time a bubble, size was never the edge, and earnings yield beats book-to-market — while David Hay makes the case that this resolves as a great rotation into value, regional banks and international rather than an outright meltdown, with the yen and the long bond as the two underpriced risks. Mike Rothman lays out an oil situation without precedent in industry history: roughly 1.5 billion barrels of forfeited supply and the biggest inventory draws on record coinciding with falling prices, a disconnect he attributes to a collapse in the financial demand for oil. Matthew Smith, who has spent eighteen months rebuilding the U.S. gas system well by well, shows why a curve pricing $3.50 gas through 2030 collides with storage that breaks below all historical evidence by 2029. Adrian Day, from the Rule Symposium floor, makes the contrarian case for gold miners down 40% since January against nine straight quarters of rising cash flow. Chris Whalen supplies the rates-and-inflation consequence — double-digit wholesale inflation, a 10-year closing in on 5%, 7% mortgages as the new baseline — and Melody Wright closes on housing, where early-stage delinquencies have risen non-seasonally for four straight months and the buyer of last resort is turning out to be the government. Each summary is designed to be immediately actionable — whether you are allocating capital, running a business, or simply trying to understand the forces reshaping the world around you.
THIS WEEK'S LINEUP
EP 1 AI Revolution: Winners & Losers — Dan Ives & Gil Luria — Veteran Tech Analysts, ex-Wedbush & D.A. Davidson — Watch Full Video
EP 2 Crashes in Progress: Look Out Below — Gordon Johnson — GLJ Research Founder & CEO — Watch Full Video
EP 3 What to Do When Markets Become Casinos — George Noble — Noble Capital Advisors — Watch Full Video
EP 4 Why Size Was Never the Edge — Wes Gray — Alpha Architect Founder & CEO — Watch Full Video
EP 5 The Great Rotation — David Hay — Haymaker Publisher; former CIO, Evergreen Gavekal — Watch Full Video
EP 6 You Can't Own Enough Energy — Mike Rothman — Cornerstone Analytics Founder — Watch Full Video
EP 7 The 2028 Natural Gas Crisis No One Sees Coming — Matthew Smith — Chronometer Partners Founder & CIO — Watch Full Video
EP 8 The Contrarian Setup in Gold Miners — Adrian Day — Adrian Day Asset Management — Watch Full Video
EP 9 5% Yields, 7% Mortgages, Double-Digit Inflation & the End of the Party — Chris Whalen — Whalen Global Advisors Chairman — Watch Full Video
EP 10 No Buyers: Housing's Coming Downdraft — Melody Wright — Housing Analyst, M3_Melody — Watch Full Video
Full summaries with actionable insights and investment focus for each podcast follow on the pages below.
EP 1 - AI Revolution: Winners & Losers
Dan Ives & Gil Luria — Veteran Tech Analysts; ex-Wedbush & D.A. Davidson
Steve Eisman convenes two of the few analysts who cover the entire technology stack — Dan Ives, newly departed from Wedbush to launch his own investment bank, and Gil Luria of D.A. Davidson — for the week's definitive winners-and-losers session on AI. The debate has hardened into two questions: will the unprecedented data-center spend ever earn a return, and does AI destroy or entrench the software incumbents? Ives argues we are in year three of an 8-to-10-year buildout — building the Vegas Strip in 1955 — while Luria, the analyst who publicly stepped off the Oracle bandwagon at the very top, hunts the dislocations that the daily AI-is-good/AI-is-bad whipsaw keeps creating. Eisman presses three bear cases — sudden capital intensity, missing moats, and a Chinese-triggered price war — and gets the most substantive rebuttals of the week.
Actionable Bullet Points
The Two Debates — and the $100 Billion Answer: Luria frames the entire fight as two questions: will the extreme, unprecedented data-center investment earn a return, and what does AI do to everyone else, especially software (2:50, 3:11). His ROI answer treats AI as a value chain, not one business — equipment makers (ASML, TSMC), chipmakers (Nvidia, AMD, Micron), the hyperscalers providing compute, and the model labs (11:42) — with OpenAI and Anthropic alone now running above a $75 billion combined revenue rate, and north of $100 billion once Gemini, Meta, and xAI are included, versus zero two years ago (12:15, 12:30). Roughly a trillion dollars is in the ground against $100 billion of demonstrated willingness to pay — not a great return yet, but real and compounding economic activity (12:56). Microsoft, he notes, gets the raw end of both debates at once — accused of wasting capital and of being disrupted — yet its growth is accelerating with AI a tailwind to Azure, Office, and infrastructure software, and when agents run through Excel, Outlook, and Teams, Microsoft stands in front of the model (3:49, 35:53, 4:39). Score the buildout on paid consumption, not on the day's narrative.
Eisman's Three Bear Cases — Capital Intensity, Missing Moats, and a Chinese Price War: Eisman's challenge: companies that had not raised capital since inception — Google, Microsoft, Meta — are suddenly in a capital-intensive business, with Google just raising $85 billion in equity (7:13, 20:55); every week a new model leapfrogs the last, so nothing resembles Google's old search moat (7:49); and China's Kimi K3 is charging roughly a fifth of what Western labs charge for tokens — if he ran Anthropic or OpenAI, Eisman says, he would be petrified (8:46, 11:27). The rebuttal: models get cheaper and commoditize, but the value migrates to data, install bases, and the ecosystems being built in front of us — enterprises will end up choosing among a handful of full-stack providers and paying the piper, the way Netflix's early content spend became an unassailable position (9:17, 9:55, 10:37). Ives' frame: year three of an 8-to-10-year buildout, with gut-check moments three to four times a year and chip demand-to-supply running 15-to-1 on his recent Asia trip (5:09, 5:42, 5:50). Expect the whipsaw; do not confuse it with the trend.
Luria's Dislocation Playbook — Oracle's $630 Billion Backlog Is Priced at Zero: Luria's rule is to get nervous when every hedge fund and long-only is on the same bandwagon — exactly what September 10, 2025 looked like, when Oracle's backlog jumped from $150 billion to $450 billion in a day and the stock ran from roughly 200 to 330, before the market learned it was one OpenAI deal from a customer with no money and $1.4 trillion of total commitments (29:12, 26:33, 29:43, 29:58, 30:14). At 140, he argues, the market has overcorrected the other way: OpenAI has since raised $122 billion — the largest fundraise in history — clarified that most commitments are flexible, and gone code red on compute, yet Oracle's entire $630 billion backlog is being valued at zero or less (30:34, 30:46, 31:09, 31:26). The same inconsistency runs through semis: Intel at roughly 100 times earnings and the semicap/optical complex are priced as if the cycle runs through 2030, while Micron at about six times and Nvidia are priced as if next year is down — and Luria argues the memory market is now better than the CPU market (33:53, 34:15, 34:27, 34:48). Trade the market's internal contradictions, not its mood.
Platform Risk Is Real — Karp's Warning, the Fable Episode, and the Open-Source Counterweight: Translating Alex Karp's warnings, the pair land on two dangers: feed your data directly into a frontier lab's model and it learns how your business operates — and can compete with you — and build your business on top of one closed model and any change to that model can take you out (18:11, 18:25). Luria points to what he describes as a recent episode in which the government pressured Anthropic over its Fable model as the wake-up call on single-model dependence, and to Palantir's answer: stay model-agnostic, with the value in the data and the ontology around it (18:38, 19:02). The structural counterweight is open source — open weights plus open code you can run yourself (13:58) — where only Chinese labs have played so far, partly because the view is Anthropic and OpenAI are so far ahead that chasing them is like chasing Usain Bolt (15:28, 15:57); but Nvidia is now shipping its own free Nemotron model, with Microsoft and Palantir on board, because open models burn just as much compute (16:08, 16:26, 16:43). The end state: frontier closed models for mission-critical work, open-source and small on-device models for everything else (14:52). Architect for multiple models; treat single-model dependence as a balance-sheet risk.
The Winners-and-Losers Map — and the Political Overhang That Could Decide It: The winners: Nvidia — one chip fueling the revolution, a third-rate Nvidia part still 1.5-2 years ahead of Huawei, with an $8-10 multiplier across tech for every Nvidia dollar (36:31, 36:39, 37:13); Apple as the "easy pass on the consumer AI highway," monetizing 2.5 billion iOS devices with no capex and a model-agnostic new Siri (24:55, 25:16, 25:52); cybersecurity, where budgets double over two to three years as every agent expands the attack surface — CrowdStrike and Palo Alto Networks the picks (38:09, 38:41); and Google, re-rated from AI loser at $180 to "a winner" — cloud growth into the 60s and accelerating search advertising, though its model is no longer state-of-the-art and it sits a distant second in consumer chat, third in enterprise AI (21:10, 21:44, 22:11, 22:40). The losers: Salesforce, charging more every year for less value as CIOs crowd out unimportant software — IBM's pre-announcement the tell (35:42, 35:17, 35:31); Adobe and Intuit, which Ives says miscalculated what AI does to the business model (38:58); and above all private-equity-owned software that stopped investing — Medallia last week — as CIOs consolidate a hundred packages toward thirty, while the public software names the pair cover sit in net-cash positions (40:33, 40:49, 41:21, 42:01). Overhanging it all: data-center moratoriums that in their telling hand the win to China, a degrowth movement, and Luria's charge that Altman and Amodei are "pulling the ladder" — stoking job-loss fears to win friendly regulation that shuts out open source and everyone else (45:10, 47:01, 49:02). Position long the toll-takers; treat the politics as the tail risk.
Investment Focus
The most balanced AI session of the week — a flag-bearer bull, a dislocation hunter, and a skeptical host pressure-testing both. The investment template: (1) score the buildout on paid AI consumption — the $100-billion-plus run rate that was zero two years ago — rather than on capex headlines or the daily narrative whipsaw (12:30, 12:56); (2) trade the market's internal contradictions: Micron at ~6x priced as if the cycle is over versus Intel at ~100x priced as if it runs through 2030, and Oracle's $630 billion backlog carried at zero now that OpenAI is funded (34:15, 34:27, 31:26); (3) own the toll-takers — Nvidia with its $8-10 multiplier, Microsoft standing in front of the model, Apple monetizing 2.5 billion devices without the capex, and cybersecurity as budgets double on agent surface area (37:13, 4:39, 24:55, 38:09); (4) avoid the crowded-out — Salesforce-style share donors, Adobe and Intuit business-model risk, and private-equity-owned software where investment stopped, noting that software-debt distress is a private-equity story since the public names are net cash (35:42, 38:58, 41:21, 42:01); (5) architect and underwrite model-agnosticism — Karp's platform-risk warning and the Fable episode argue for multi-model strategies, with American open source (Nvidia's Nemotron) and the regulatory fight the swing factors to watch (18:25, 18:38, 16:26, 49:02).
EP 2 - Crashes in Progress: Look Out Below
Gordon Johnson — GLJ Research Founder & CEO
Gordon Johnson joins George for a forensic teardown of the two assets sitting at the center of the retail mania. His frame is that Elon Musk made three unforced errors that convert a narrative premium into a liability: launching unsupervised FSD in Austin on a self-imposed clock, insisting a vision-only stack can do what the rest of the industry says requires lidar and radar, and taking SpaceX public in a way that finally exposes its financials to daylight. What follows is an hour of arithmetic done entirely with Tesla's own disclosed numbers — robotaxi miles, FSD subscribers, cash flow, float schedules — against a market capitalization that assumes none of it matters. Johnson's conclusion is a sum-of-the-parts in the fifties and a twelve-month target near $25; George's is that the same liquidity mechanics that inflated this will run in reverse.
Actionable Bullet Points
Three Unforced Errors — and the SpaceX Filing That Ends the Mystery: Johnson's first mistake is launching robotaxis in Austin on a deadline rather than on readiness (1:57, 2:16); the second is the vision-only bet, which he argues cannot work without lidar or radar the way the rest of the industry has concluded (2:34, 2:53); the third is the gap between promises of fifty cities and exponential scaling versus the seven or eight vehicles actually visible on the ground (3:11). The costliest is SpaceX going public, because an S-1 forces disclosure of what has never been disclosed (4:18, 4:46) — thirteen Starship launches that have not reached orbit and a cash burn he puts near $4 billion a year (5:24, 5:50). Treat the filing itself, not the narrative, as the catalyst.
Tesla's Own Numbers Give You an $840 Million Market — Priced at $700 Billion: Tesla has disclosed roughly 380,000 unsupervised robotaxi miles in Austin (13:50); running that at the observed rate and extrapolating to the full country produces a total U.S. addressable market of about $840 million (14:47), against roughly $700 billion of market value assigned to the opportunity (16:05). The FSD math is no kinder: about 1.5 million paid subscribers globally (17:02) at roughly $100 a month is on the order of $1.8 billion in annual revenue, carried at something like $350 billion (17:37). Anchor to disclosed operating data, not keynote slides.
Sum-of-the-Parts Lands at $57.46 — Using Competitors' Own Valuations: Johnson builds the parts generously: Optimus marked at Figure AI's $39 billion (27:46, 28:00), and FSD marked at roughly Waymo's $100 billion — even though Waymo has driven 127 million miles to Tesla's 380,000 (28:16, 28:27). On safety, he reads fourteen NHTSA-reported accidents across those miles as roughly one per 55,000 miles, about ten times worse than the human benchmark (28:50, 29:11), and notes Tesla's own documentation classifies the Austin vehicle as Level 2 with liability capped at the ride cost or $100 (30:31, 30:56). Add an auto business benchmarked near Ford's $57.5 billion (32:43) and the total is $57.46 a share (33:30); GLJ's twelve-month target is about $25, roughly 12x his 2026 estimate (33:35). The stock trades near 310x trailing earnings against a Magnificent-6 average of 33x (36:17, 36:31, 36:48).
The Merger Math and the Float Schedule Both Work Against Holders: On a SpaceX-Tesla combination, Johnson's arbitrage math shows dilution already widening from 50% to 57.8% (20:28), taking Musk's side from half the combined entity to about 42% (20:35), all underwritten by a claim that Optimus is a $25 trillion opportunity (20:58). The practical obstacles — Chinese approval, U.S. review, and what would be the largest S-1 in capital-markets history — argue for a two-to-three-year timeline at best (21:43), and he expects SpaceX to sell off on announcement rather than rally (26:00). Layer on a float engineered to expand from 5% to 25% to 50% to essentially 100% inside a year (52:51), and supply, not story, becomes the price-setting variable.
The Funding Window Is Closing While the Liquidity Drain Starts: Johnson reads Nvidia's contemplated $250 billion commitment to OpenAI as a desperation tell (47:26, 47:39) — vendor financing with an Enron echo, where the supplier becomes the only lender willing to fund its own demand (48:04). Into that he adds a Fed with Kevin Warsh's hawkish record (49:05), roughly $320 billion of net Treasury issuance between late July and early September (49:35), and a reverse-repo facility drained to nearly zero (50:06), which he reads as negative for risk assets broadly, Bitcoin included (50:27). The unpriced political tail is Kalshi's roughly 80% odds on a Democratic House and the subpoena power that comes with it (57:05). Position for the liquidity window, not the headline.
Investment Focus
Johnson supplies the arithmetic; George supplies the portfolio response. The investment template: (1) treat Tesla and SpaceX as the concentrated expression of the retail bubble, with a sum-of-the-parts near $57 and a twelve-month target around $25 defining the downside case (33:30, 33:35); (2) rotate the core — equal-weight RSP over cap-weighted SPY, shorten duration, and move part of the bond allocation into energy and gold as debasement hedges (55:21, 55:36, 55:49); (3) watch free cash flow as the trigger, with capex above $25 billion against roughly $10 billion of operating cash flow and a history of roughly 15% drawdowns in flat-or-negative cash-flow quarters (34:55, 35:16, 35:47); (4) respect the leverage mechanics in reverse — Buffett's weapons-of-mass-destruction warning, and the double-levered Lucid product that closed at negative NAV, are the template for how this unwinds (38:40, 40:10); (5) trade the calendar — net issuance into early September against an empty reverse-repo buffer, then the midterm subpoena risk, with data-center industrials the picks-and-shovels long on the other side (49:35, 50:06, 57:05, 57:37).
EP 3 - What to Do When Markets Become Casinos
George Noble — Noble Capital Advisors, on Chris Martenson's Finance U
George joins Chris Martenson for the system-level diagnosis behind the week's other conversations. His opening frame is Lenin's line about decades where nothing happens and weeks where decades happen — a tech wreck underway, an AI complex he believes is imploding, thirteen truces in eight weeks in the Middle East, a 10-year Treasury at 4.60%, and a market perched on a ridge with zero margin of safety. The argument that follows is mechanical rather than moral: rallies of this kind are liquidity phenomena, the shorts that provided the bid have been cleared out, and the same machinery that manufactured the upside runs in reverse on the way down. What separates this from ordinary bearishness is the prescription — own the things that cannot be printed, and stop measuring wealth in the unit being debased.
Actionable Bullet Points
The Rallies Were Liquidity Events — and the Bid Has Been Removed: George traces the advance to record short squeezes rather than any fundamental improvement (4:06, 4:32), and makes the structural point that shorts are the natural bid in a selloff, so squeezing them out removes the stabilizer just when it is needed (5:08). His exhibit is Korea: the KOSPI ran eight standard deviations above its channel, then fell 31% in seventeen days (5:52, 6:00) — not a healthy correction but Q1-2000-style volatility (6:38), with Bob Farrell's rule that explosive moves do not resolve sideways leaving bag holders at the highs (6:56, 7:09). Triple-levered products dropped 40% in single afternoons, Korean margin accounts were liquidated, and an ETF traded at negative NAV (7:30, 7:44, 8:11). The movie runs in reverse (8:28).
Adoption Is Not Returns — the 2000 Template: Against the "but AI is real" rebuttal, George points out that internet traffic compounded 43% a year for 26 years, a cumulative gain of roughly 25 million percent, and it did not save Lucent, Nortel or JDSU from losing 90% or going to zero (20:25, 20:36). Buffett's textile-loom parable is the mechanism: the benefits accrue to consumers, not shareholders (21:12), and Peter Berezin's 1999 parallel makes the same point from the strategist's chair (20:12). Meanwhile Chinese labs are undercutting at a fraction of the price (38:33), Gary Marcus and Ed Zitron argue the economics cannot work (38:46), Nvidia's financing of CoreWeave is "legal but wrong" vendor financing (30:40), Oracle CDS is the tip of the spear (38:21), and AGI is the South Sea Company's "extraordinary undertaking, nobody to know what it is" (39:57).
The Energy Case Is Structural, Not Geopolitical: The paper market trades roughly 60x physical (11:16) with retail crude shorts at record levels, so price is being set by positioning rather than barrels (11:43). Underneath, China has cut imports 4-5 million barrels a day (14:46) and the SPR is a one-trick pony already played (15:05), while the gold-to-oil ratio is stretched far enough that simple mean reversion implies oil near $250 — a gap that historically closes with oil rallying, not gold falling (15:28, 15:48). Energy is 3.5% of the S&P but 13% of free cash flow heading toward 20% as hyperscaler profits compress (29:36). George calls it the golden age of energy stocks (19:17) and invokes Chuck Clough's rule: buy where capital has been starved (19:24).
Money Illusion Is the Tax Nobody Reports: The clearest exhibit is a single asset priced three ways — the long bond in dollars, in Turkish lira, and in gold: same merchandise, three different stories, and only one of them honest (33:52, 34:12). George cites Greenspan's admission that the government can guarantee payment but not what the dollars will buy (35:46), and argues we are in a bear market in real money that boils the frog slowly (36:24, 36:37); Smithers' formulation is that in a bear market capital is returned to its rightful owners (36:57). The fiscal arithmetic backs it: Julien Garran of MacroStrategy Partnership scores today's misallocation at 17x the dot-com peak (58:55), corporate margins mirror deficits (1:01:09), and the debt actually rose $3.2 trillion in twelve months against a headline $1.9 trillion deficit — watch what they do (1:02:38).
SpaceX Was Built to Fail — and the Float Calendar Tells You When: George's estimate is $300-400 billion of value against roughly $1.7 trillion of market capitalization (47:49), with a float engineered to expand from 5% after the early-August earnings to 25%, then toward 100% within six to twelve months (48:27, 48:36) — a structure whose purpose is to supply exit liquidity to insiders and leave retail holding the last tranche (49:23). Tesla is the same shape: $38 billion of cumulative lifetime profit, much of it regulatory credits, against $1.6 trillion (50:17), an arrangement he compares to Ivar Kreuger (51:28) and calls a cash incinerator holding 1.7% of the nation's 401(k) assets (57:24). Read the unlock calendar as a supply schedule.
Investment Focus
The week's most complete descriptive-to-prescriptive arc, and the one that translates directly into a portfolio. The investment template: (1) replace cap-weighted exposure with equal weight, since the S&P is roughly 40-50% technology once reclassifications are counted, and RSP removes the concentration without leaving equities (29:29); (2) take energy from an index weight of 3.5% to a portfolio weight of 10-20%, underwritten by 13% of free cash flow heading toward 20% and a decade of underinvestment (29:36, 29:58, 30:05); (3) replace duration with debasement hedges — gold, copper and resources rather than bonds — because the unit of account is the position being taken (30:05, 33:52); (4) treat rallies as liquidity events to sell into rather than trends to chase, using the KOSPI's eight-standard-deviation break as the template for how they end (4:06, 5:52, 8:28); (5) trade the SpaceX float calendar as a supply catalyst, with the post-earnings unlock the first date that matters (48:27, 48:36), and note that George closes constructive rather than uniformly bearish — there is plenty to own, just not what everyone owns (1:02:04, 1:02:16).
EP 4 - Why Size Was Never the Edge
Wes Gray — Alpha Architect Founder & CEO
Wes Gray — Chicago finance PhD, former Marine, founder of Alpha Architect and the ETF Architect platform behind more than a hundred funds and $37 billion — treats the bubble question as an engineering problem rather than a narrative one, and reaches a conclusion that is equal parts useful and unsatisfying: spotting euphoria is easy, timing it is not, and the evidence says valuation-based timing models do not work. What he offers instead is a set of repairs to the tools most investors are using badly — why the size premium was never real, why book-to-market broke, and why earnings yield fixes both. He closes on Section 351 exchanges, the quiet structural development that lets low-basis portfolios move into an ETF wrapper without triggering the gain.
Actionable Bullet Points
You Cannot Time the Bubble — So Manage Your Exposure to It: Gray's position is that high prices do predict poor long-run returns on average, which means a 20-to-30-year horizon of mediocre index returns from here (4:48, 5:07) — but no valuation model reliably converts that into a timing signal (5:45), and only trend and momentum show any real efficacy (5:56). The prescription is unglamorous: grit, diversification and low fees (6:03). His sharpest point is that the cap-weighted index is not neutral — it is a concentrated bet on large-cap quality growth, a factor combination whose forward expected returns are weak (7:01, 7:33). Pull the bet down where you can rather than trying to call the top (7:56).
Size Was Never the Edge — Value Was Doing the Work: AQR's research finds no expected-return premium to size once it is isolated (10:43), and Jack Vogel's test makes it concrete: build an equal-weight portfolio of mid- and large-caps and compare it to small-caps — you get ten times the liquidity, the same value characteristics, and the same returns (11:06, 11:29, 11:41). The confusion comes from pulling cap-weighted "large value" off a database, which is systematically more expensive than small value and therefore never held value constant. Buy the 2x-earnings microcap because of the 2x, not because it is small (13:04).
Fix the Metric Before Declaring the Factor Dead: Between 20% and 30% of high book-to-market names are not profitable (8:41), because book-to-market is oddly negatively correlated with quality in smaller stocks; the income statement is where the information lives (9:00), and Google is the obvious illustration — enormous earnings, almost no book (9:39). Russell 2000 value is roughly half unprofitable (14:40), which is a quality problem misdiagnosed as a value problem. Earnings-to-price captures cheapness and quality in a single number (16:51) — Graham's 1972 study produced roughly 20% annually from 1920 to 1972, and Alpha Architect's out-of-sample replication produced nearly identical results (22:01, 22:30). Gray will bet it persists another thirty to forty years (22:51).
AI Repriced Growth Once — Which Tilts the Odds Back to Value: Gray's view has moved from AI being "kind of cool" five years ago to the most amazing technology that has ever existed, and he describes it as making him artificially smarter than he is (20:54). But his analytical read is a one-time repricing: growth was underpriced, prices are now set for the moon rather than Everest (19:13, 19:51), which argues the value effect reasserts out-of-sample (20:17). On alpha, short-horizon edges have been arbitraged toward the cost of capital (23:45), and nobody is playing the ten-year game because a Millennium seat does not survive a bad quarter (24:14) — so the remaining edge, if there is one, is at long horizons (25:16). The "God portfolio" paper is the proof: even with perfect five-year foresight the drawdowns get you fired (26:20, 27:02).
351 Exchanges: How Stuck Low-Basis Portfolios Get Unstuck: The mechanics are that contributing property into a new C-corp is tax-free under Section 351, and an ETF is structurally a C-corp that elects RIC treatment (34:41) — capital flows to whatever wrapper treats it best (35:57). The killer application is the dead tax-loss-harvesting portfolio: after a few years of gains it is an expensive index fund with tracking error and no losses left to harvest (37:26). The constraints are real — no position above 25%, top five below 50%, diversified going in (40:15) — and it only became workable after Rule 6c-11 permitted custom baskets in 2019 (42:06). Gray's firm does more than half these deals and turns away the schemers, because intent matters (47:40, 48:14).
Investment Focus
The week's antidote to narrative investing — a set of repairs rather than a call. The investment template: (1) plan for lower index returns and diversify away from the large-cap quality-growth concentration the cap-weighted benchmark imposes, accepting that trend is the only timing tool with evidence behind it (5:07, 5:56, 7:01); (2) express value through earnings yield rather than book-to-market, since E/P embeds the quality screen that book-to-market inverts (16:51, 8:41); (3) pay for cheapness, not for smallness — equal-weight mid- and large-caps deliver the same value exposure with ten times the liquidity (11:06, 11:41, 13:04); (4) use 351 exchanges to unstick low-basis and dead direct-indexing portfolios, respecting the 25%/50% diversification limits and the 6c-11 plumbing that makes it possible (34:41, 37:26, 40:15, 42:06); (5) underwrite the behavior, not just the strategy — underperformance is the price of the premium, and the God-portfolio result shows even perfect foresight does not spare you the drawdown (26:20, 27:02, 54:46, 55:15).
