Living Thesis
Each view below carries a probability: the odds we currently assign to it being right, synthesized from a tracked roster of investors and analysts. Disagreement among them is preserved, not averaged away — the named dissent is part of the view.
Named people are the sources behind each view; the "changes my mind" lines are the specific evidence that would force a rewrite.
Two forces dominate. First, an AI building boom that is real in demand but increasingly priced for perfection — with the speculative excess concentrated in commodity memory chips (DRAM/NAND), not the whole AI complex. Second, a live energy supply shock: the Strait of Hormuz, the channel that carries roughly a fifth of the world's oil, has been effectively closed for about three months, removing 11-13 million barrels per day. The world has not felt it yet because China halted imports, refineries cut their runs, and record amounts were drawn out of storage — but storage is now approaching its operational limits. Overlaying both: inflation has run above 3% structurally (63 straight months above the Fed's 2% target) while government debt sits at record levels relative to the economy, so the two jobs of a central bank — defending the currency (higher rates) and keeping the government's borrowing affordable (lower rates) — now pull in opposite directions. The result is what economists call financial repression: the Fed's policy rate, after subtracting inflation, has been negative in 61% of months since 2000, bottomed near −8.4% in the 2022 inflation shock, and sits just below zero today (charted below) — meaning savers quietly subsidize borrowers even when rates look high on paper. That persistent below-zero real rate is the engine under our hard-asset views (the cornered Fed, thesis 4; debt forcing money-printing, thesis 5). Meanwhile stock valuations sit at multi-decade extremes with the worst "breadth" on record — record index highs carried by a handful of giant companies while the average stock lags. Base case: tactical caution on the broad index, overweight real assets and energy, avoid the crowded memory-chip trade, and accept near-term chop in gold within a longer hard-asset bull market.
What skeptics called "circular" — AI companies funding and revaluing each other — is, for now, a powerful earnings engine: Anthropic's revaluation (from $380B in Q1 to ~$960B now) flows straight into Amazon's and Microsoft's reported profits as investment gains, contributing roughly 10 points of Big Tech's ~27% year-over-year earnings growth (Jim Bianco, market researcher), and could triple in Q2. Bubbles built on real earnings historically run for years, not months — the 1995-2000 internet boom is the working analogy. The strongest counter-case on record comes from Raoul Pal (macro investor, Real Vision founder) and Jordi Visser (40-year markets veteran, former hedge fund chief): this is a supercycle, not a bubble — demand compounds with network effects, price-to-earnings ratios have actually COMPRESSED because earnings kept pace with prices, data centers are only ~30% built versus announced plans, and fewer than 1% of large enterprises have meaningfully adopted AI; in their view supply bottlenecks concentrate capital rather than break the cycle. Whale Rock, a large tech-focused hedge fund, agrees and owns the whole memory-chip chain. Notably, even these bulls expect a 3-6 month digestion period as roughly $4 trillion of new stock offerings comes to market — the camps disagree about the aftermath, not the air pocket. The warning sign both camps share: if AI software becomes efficient enough that big tech cuts its hardware spending, every AI-adjacent trade — power, memory, networking — cracks at once. What changes our mind: a leading AI model company stalling, or prices rising faster than earnings (true mania behavior, which we are not yet seeing).
Three of our four AI-focused sources independently say short or avoid memory. Micron is up ~1,000% in 15 months; 2028 profit estimates for SK Hynix and Samsung run $200-250B — pricing them like the world's most profitable companies on what has always been a boom-and-bust commodity product, at ~10x book value, ten times the accounting value of the business (Anton Fritzell, semiconductor analyst). Visser sold his Micron at $600-650. Chinese manufacturers (YMTC, CXMT) bring new supply from 2027 — historically the thing that breaks every memory cycle. The dissent: Whale Rock owns the entire memory supply chain, arguing demand exceeds supply for four more years. The daily tell: spot prices for data-center memory, tracked by research firm TrendForce — when they turn down, the cycle is breaking.
The physical market is genuinely tight: Hormuz effectively closed ~3 months (11-13M barrels/day shut in), masked by China halting ~4.5M barrels/day of imports, ~9M barrels/day of refinery cuts, strategic-reserve releases, and record draws from storage; US diesel-type fuel inventories sit at ~23-year lows; the main US storage hub at Cushing, Oklahoma is draining toward its operational floor. If talks fail, the missile math says Iran could shut the region's energy exports with a few hundred missiles (Gordon Johnson, analyst), and losing 13-14M barrels/day would empty global storage by August-September. But markets are pricing peace: Trump has declared the war over with Iran "accepting" (Iran denies anything is final; the five stated red lines remain largely unmet), professional traders are overwhelmingly positioned for lower prices (>75%), and US crude (WTI) closed below the $89 support level tracked by our technical chart service, Northstar & Badcharts — opening their downside scenarios. The peace case gained a third independent voice with a coherent political mechanism (Forward Guidance, a macro research team): the administration is engineering de-escalation to protect the SpaceX stock offering and to avoid +80% gasoline prices going into the midterm elections, while demand destruction from high prices is underappreciated. The levels that resolve the debate: WTI back above $89 keeps the floor case honest; a close above $107 breaks the downtrend and targets roughly $150-180 (Matt Smith of the shipping-data firm Kpler argues $200 in a full rupture). The swing factor: a US-backed maritime insurance vehicle (DFC/Chubb) that could restart Gulf shipping, gated on Navy escort capacity. How we would express the view: the cash-generating end first (refiners, large producers); oil services as the under-owned angle (Larry McDonald, market strategist: SLB, WFRD); uranium structurally attractive but no rush. Changes our mind: Hormuz reopens or China releases inventory (the floor breaks); WTI reclaims $89 and then $107 (the spike revives).
May consumer prices ran the hottest since May 2023 (core inflation, which excludes food and energy, at 2.9%; producer prices, purchasing-manager surveys and shelter costs all pointing higher), and the bond market has repriced from expecting three rate cuts to expecting none: the 10-year yield's center of gravity moved up to 4.50%, the 30-year holds the 5% line, and the 2-year yield now sits ABOVE the Fed's overnight rate — a configuration that historically means the market is pricing the next move as a hike, not a cut (the chart below shows that spread at +0.51 points, in hike-pricing territory since April). Critically, this is "a rate story, not a credit story" (DoubleLine, the bond manager): the extra yield junk-rated borrowers pay is near record lows, signaling no recession fear in credit markets. Four distinct views on the next move, deliberately preserved: (1) a hike is coming — Bianco puts the odds at 100% by year-end and says the bond market already prices it; (2) the Fed CAN'T hike — McDonald argues that with interest on the national debt running ~$1.1 trillion a year, a hike is a "mirage," and he positions instead for the gap between 2-year and 30-year yields to widen; (3) hiking would be a policy error — Infranomics (research service) points to falling hiring, full-time employment down 79k, wages trailing inflation, and broad underemployment at 8.1%: raising rates into a supply shock deepens the downturn; (4) maximum hawkishness is already priced — Forward Guidance notes the ~2 hikes markets price through mid-2027 is the ceiling, falling inflation expectations against steady yields mean real borrowing costs are rising on their own (the economy tightening itself), and incoming Fed chair Warsh's preferred inflation measure strips out energy, giving him cover not to hike at his debut; their expression is a bet on falling short-term rates paired with gold — it wins on cuts OR a softening economy, and only needs the priced-in hikes to not materialize. Common ground across all four camps: inflation stays around 3% or higher and rate cuts are off the table. Changes our mind: job growth turning negative, or oil resolving lower.
The frame comes from Luke Gromen (macro analyst): record debt and AI-driven deflation cannot coexist at this debt-to-GDP — the endgame is parking the debt on the Fed's balance sheet, with the dollar as the release valve, likely disguised as technical plumbing ("fake QT": balance-sheet tightening in name, easing in effect). That is bullish for gold, Bitcoin, and stocks measured in dollars, and bearish for the dollar itself. The supporting evidence: gold now exceeds US Treasuries in global central-bank reserves (Lyn Alden, macro analyst), and the "Buffett indicator" — total stock-market value relative to GDP — is the highest in 80 years. The near-term gold stance is structural-but-not-urgent: the trend is intact, but after two outlier years the easy gains are gone and risk runs both ways. Tactically gold (~$4,209) sits below its ~$4,587 200-day moving average (the standard long-term trend line) in the support zone Northstar tracks (~$4,000); losing $4,000 opens $3,700-3,800 and then the low-$3,000s — and a dip into that zone is their long-term re-entry point, which is our roadmap for re-adding the gold sleeve. Forward Guidance pairs gold with their falling-rates bet: gold rallies if the priced-in hikes never come. Silver (~$66, down ~50% from its high, below its ~$73 trend line) has a probable mid-$50s downside target — but Visser adds a new structural demand leg: solid-state batteries need materially more silver, making it a scarce input to the AI-era buildout. Copper, the consensus metal, is now flagged as crowded (Pal: engineering routes around every shortage — Tesla's switch from 12-volt to 24-volt wiring cut its copper use 70%). The cheap expression of the whole hard-asset view is the gold MINERS (McDonald): sold down to multi-decade-low valuations — Agnico Eagle ~40% off its high at ~5.9x EV/EBITDA (a standard price-to-cash-earnings yardstick; under 6x is cheap) with $6-7B of annual free cash flow — structurally bullish, near-term careful because miners fall hardest in any shock. Changes our mind: a credible deficit-reduction path, or central banks turning from gold buyers to sellers.
Valuations are at extremes — the Buffett indicator is above both the 2000 and 2021 peaks — and leadership is the narrowest on record: index highs carried by a handful of mega-caps while the average stock lags (Gromen). The rotation has a mechanism, which McDonald calls "the migration": the cap-weighted S&P (where the giants dominate) is sharply underperforming the equal-weighted version (every company counted equally; the Magnificent 7 are roughly flat since October — the chart shows the cap-to-equal-weight ratio peaking near 122 in late 2025 and rolling over to ~117.7), as a ~$3.6 trillion wave of new stock offerings drains money from the crowded index toward smaller companies and value stocks — 1999-style insiders-selling-to-retail. DoubleLine independently confirms the picture: emerging markets +18% and small-caps +15% this year while defensive utilities are down 5%; their positioning notes add that mortgage- and asset-backed bonds look cheap and floating-rate loans win if rates stay high. The new supply is now LIVE: SpaceX priced at ~$2 trillion (4x oversubscribed, $70B+ of retail orders, fixed price, immediate NASDAQ-100 inclusion — on losses of $4.28B per quarter; Renaissance Capital, the IPO research house: the price needs 2028-2030 cash flows, "not an Nvidia"), with $200-250B of immediate supply and ~$3 trillion of insider-share unlocks behind it over 12-18 months. Forward Guidance quantifies the regime change: companies are now issuing more stock than they retire through buybacks for the first time in 10-15 years (SpaceX $75B + Google $80B + Oracle $40B + CoreWeave $3.5B) — and every dollar of newly issued shares must be bought with cash that would otherwise be supporting existing prices. Even the AI bulls (Pal/Visser, with their ~$4T pipeline estimate) expect 3-6 months of digestion: both camps agree on the near-term drain and disagree only about the aftermath. The firmest expression is healthcare: the most under-owned sector in the market (from ~16% of the S&P to ~8%; McDonald), surgical-robot maker Intuitive Surgical (ISRG) sitting exactly on the 200-WEEK moving average ($398.5) that McDonald calls his buy level after a 33% de-rating, options strategist Patrick Ceresna's defined-risk structure (own the XLV healthcare fund with an August 145/165 "collar" — protected below 145 in exchange for capping gains above 165), and XLV leading the market above its trend line while the index churns. Dissent on framing: Visser says this is rotation, not a bear market (credit spreads tight, earnings estimates still rising), and Northstar stays constructive as long as the S&P holds its breakout support — the clean technical line that pairs with the fundamental caution. Favor energy, financials, industrials, and healthcare over the cap-weighted tech-heavy index. Changes our mind: the SpaceX debut absorbs cleanly and leadership broadens (bullish resolution), or the S&P breaks its support (turns the call from rotation to bear market).
Bitcoin pierced its February lows — invalidating the earlier "bottom is in" call — and sits near $63,400: 19% below its 200-day moving average and roughly half of our model's fair value, with the model's composite at its buy trigger and crowd sentiment at extreme fear. That is an accumulation zone, but steady scheduled buying beats a lump sum until the low confirms. The timing consensus is the tightest in our corpus: Benjamin Cowen (quantitative crypto analyst) puts the most likely low at October 2026 based on blockchain ownership data, agreeing with InvestAnswers (crypto analytics channel; August-October) and Northstar's technician Wadsworth (Q3-Q4) — three independent methods pointing at one window. The timing dissent: Visser rejects calendar predictions outright — his re-entry requires the price to break back above its trend line PLUS visible evidence of capital rotating in ("this year or into next year"), and he is not yet positioned. A new watch item (Quinn of Forward Guidance): miners are selling their bitcoin to fund pivots into AI computing, and the professional arbitrage trades that normally absorb such supply are deteriorating — a structural seller that the usual cycle story does not account for. Downside magnitude is where the roster splits: Wadsworth's bear-case target is the low $30,000s (roughly 70% below the high) — which would be the first time a bitcoin cycle bottomed below the previous cycle's peak, a possible regime change, and far below everyone else's estimates. The leverage complex is the fragile link (per strategy.com, the company's own dashboard, and the framework of True North's Jeff Walton, our designated risk anchor for this name): MicroStrategy trades at 1.19x the value of its bitcoin holdings (a premium — the capital-raising flywheel still works), but the dollar reserve that pays the STRC preferred dividend is down to $1.0B, which covers only 7 months (down from ~2 years in February); debt plus preferred stock equals 42% of the bitcoin stack; and the stock's implied volatility is ~88% (options markets expect violent moves). STRC trades near $96, mildly below its $100 par value, yielding ~11.9%. Risk-adjusted ordering: the plain bitcoin fund BITB (clean ownership, no leverage) > STRC (income, mildly below par) > MSTR stock (the most leverage and volatility, plus the reserve dependency). A low-$30k bitcoin print would severely stress that structure — the strongest argument for scheduled buying, dry powder, and avoiding the leveraged versions. Changes our mind: bitcoin reclaims its 200-day average, or Strategy rebuilds the reserve toward its 2-3 year guidance.
This is the demand thesis under the family's dealership, held to the same standard as everything else. The household cushion is gone: the personal savings rate is 2.6% — the bottom 1% of the past decade — while credit-card delinquencies at 2.92% sit in the 82nd percentile and rising, consumer sentiment (University of Michigan survey) is at a decade low of 49.8, and gasoline at $4.15 sits in the 76th percentile of its decade range. Yet spending on durable goods still grows +3.3% a year — households are financing big purchases by draining savings and borrowing at a 6.75% prime rate, and that arithmetic has a time limit. For a boat dealership this is the whole demand picture: discretionary, financed, fuel-sensitive purchases. The first green shoot appeared this week — June's preliminary sentiment reading ticked UP as gas prices eased — but one month is noise, not a trend. Implication for the family book: the dealership's harvest-don't-feed stance holds, used and consignment beat new inventory, and the marine Industry watch is this thesis's daily feed. What changes our mind: the savings rate rebuilding above 4% (the trigger board watches it), credit-card delinquencies rolling over, sentiment recovering for three consecutive months, or gas sustainably under $3.50 — any two of those and the dealership demand trough is in, which would also raise the business's blue-sky value.
The second family-specific thesis, for the investment real estate. Nominal home prices are up just 0.67% over the past year (Case-Shiller index, 5th percentile of its decade range) while inflation runs above 4% — meaning real, purchasing-power prices are falling roughly 3.5% a year without a headline crash. Sellers don't capitulate because the sticker number holds; buyers quietly gain ground every month. Meanwhile the 30-year mortgage at 6.52% (76th percentile) against residential capitalization rates of roughly 5-6% produces negative leverage: borrowing costs more than the property yields, so debt subtracts from returns instead of adding. Rent growth has decelerated to 3.16% (19th percentile), and bank commercial-real-estate delinquencies at 1.56% sit at the 98th percentile of the decade — credit stress is already visible on lenders' books. Implication for the family book: keep the existing rentals (judged on their own yields once the mortgage schedule arrives), but fill the 30% real estate target through recovery and time, not new purchases — a new levered buy today starts underwater on carry. What changes our mind: the 30-year mortgage sustained below 5.5% (on the trigger board — that is where the financing math flips), real prices stabilizing (Case-Shiller year-over-year rising back above inflation), or distress wide enough that purchase prices clear the negative leverage on day one.
Overweight: energy (Exxon, Chevron, refiners), real-asset and infrastructure plays (the PAVE infrastructure fund, electrical equipment), gold and gold miners (structural; accept near-term chop), selected AI compute and applications (Eli Lilly, Nvidia, TSMC, ASML, Celestica per Whale Rock; Anthropic privately), Bitcoin (accumulate on a schedule below its 200-day average), and elevated cash/T-bills as optionality — dry powder for the washout the roster still expects.
Underweight / avoid: memory chips (Micron, SK Hynix, Samsung), legacy enterprise software (Salesforce-type incumbents), the giant cloud platforms as a relative underweight (Visser), long-maturity Treasury bonds (Bianco), and the broad cap-weighted index.
Family balance sheet (theses 8 and 9 applied): the dealership releases capital rather than absorbing it until the consumer thesis breaks; no new levered residential purchases until the housing thesis breaks; freed capital routes to the liquid sleeves per the Family page targets.
- Hormuz reopening, or the DFC/Chubb shipping-insurance vehicle going live → oil relief; unwinds the energy overweight, broadly risk-on.
- Spot prices for data-center memory turning down (TrendForce) → the memory cycle breaking.
- Job growth turning negative → recession overtakes inflation as the story and kills the rate-hike case.
- Gold reclaiming its 200-day average (~$4,587) versus losing the ~$4,000 support → confirms or denies the hard-asset leg.
- A leading AI model company (Anthropic/OpenAI) stalling → breaks the AI earnings loop that powers thesis 1.
- Corporate borrowing spreads widening from record lows → Visser's earliest warning that the macro backdrop is breaking.
- The 10-year yield near 5% AND oil above $100 at the same time → the specific combination Visser says breaks the AI trade.
- Big tech cutting AI hardware spending → the shared falsifier: ends every downstream trade (power, memory, networking) regardless of which camp is right about the supercycle.
- Bitcoin miners selling to fund AI pivots → structural selling pressure outside the usual cycle math.
- Strategy's (MSTR) dollar reserve → 7 months of dividend coverage and falling; further erosion or a deeper STRC discount threatens the income leg.
- Bitcoin holding versus losing its February lows → confirms or denies the bottoming-zone call.