Daily scan · Edition 001
Two directions dominate the window: out of US equities and government bonds, and into gold and non-US markets. Both are unusually well-documented.
Gold ETFs took in $18B and 121 tonnes in August — the second-largest month on record. Total holdings hit 4,189 tonnes, an all-time high, with AUM at $615B. The genuinely odd part: Europe out-bought North America, at +$7.9B against +$7.7B, its largest month ever. UK +$4.4B, France +$1.5B, both records.
The $2.3T Norwegian fund formally advised cutting government bonds from 70% to 50% of the fixed-income benchmark, rotating into agency MBS and corporate credit. Dollar weighting stays just above 50% — this is a credit decision, not a currency one. It also reopened the case for private markets, noting peer funds average roughly 37% unlisted assets against its own near-zero. A ministry decision follows an expert review due January 2027.
US equity funds −$22.3B. European equity +$7.9B. Asian equity +$4.8B. Gold funds +$4.2B, a six-month high. Global equity funds saw their first outflow in thirteen weeks. Money market funds −$19.7B — cash is being spent, not hoarded.
Alphabet became a top-three holding: +48.1M shares (+83%) to roughly $37B, about $17B added in the quarter. Around 60% of that came through a private placement rather than open-market buying — the structure is the more interesting detail than the size. Bank of America cut for the eighth consecutive quarter. Portfolio $299.3B across 27 holdings; Apple still 22% of it. This is Greg Abel deploying, not Buffett.
Ackman put over $3B into payment and financial-data rails in one quarter — Visa ~$1.12B, Mastercard ~$1.09B, S&P Global ~$1.06B, plus Netflix at $934M — and exited Alphabet entirely, the position Berkshire was building at the same time. Portfolio $19.47B across 15 names.
SpaceX priced at $135 on 11 June and raised $75B at roughly a $1.75T valuation — more than twice Saudi Aramco's $29.4B record. Cerebras listed 14 May, raising $5.5B. Consequently Coatue, Tiger Global, Altimeter, Third Point and Nvidia itself all show SpaceX or Cerebras as "new positions." Much of that is primary allocation, not accumulation. Reading it as a crowded conviction trade is the single most common error in this quarter's 13F commentary.
A directional reversal worth noting. PIF, with roughly $900B, held Lazard-arranged meetings in New York with Apollo, Blackstone, Brookfield, Carlyle, KKR and Stonepeak — raising capital rather than deploying it. Its AI arm HUMAIN is raising a $2.5B data-centre fund and targeting a 2029 dual listing. An adviser's framing: "PIF has been explicit that portfolio companies must reduce their draw on the sovereign." The Gulf's role in global capital markets is changing shape.
33,575 unsold PE portfolio companies as of June, against roughly 15,923 a decade ago. Q2 exits were about $100B — half of Q1. Sponsor-to-sponsor sales fell ~40% quarter-on-quarter, the weakest in at least ten years. Meanwhile institutional secondaries set records and retail-facing private credit funds took redemptions: Blackstone's BCRED fielded $4.3B of Q3 repurchase requests against roughly $750M of new subscriptions.
This is where the conventional story is weakest — and where two 2026 findings quietly demolish the usual advice from opposite directions.
Smith, Yagan, Zidar and Zwick used the death and retirement of business owners as a natural experiment and found that roughly 75% of pass-through profit flowing to top earners is attributable to owner human capital. Firm profits fall sharply when the owner stops working. Fewer than 13% of the top 0.1% live primarily on interest and rents; over 70% are under 60 and running mid-sized firms — professional services, auto dealerships, beverage distribution.
Current composition confirms it. Fed data puts the top 0.1% at 49% public equities, 20% private business, 9% real estate. The bottom 50% is 49% real estate and 5% equities. The asset that separates them is the private company, and it is not passive. contested The Fed's noncorporate-business series for the top 0.1% has not been updated past Q3 2024, so any 2026 figure for that share is an estimate.
Ebrahimian and Ljungqvist (Swedish House of Finance, February 2026) built patent-value measures for public and private firms from 1975 to 2015. Private firms' share of innovation value roughly doubled, from 16.5% to about 40%, tracking the top-1% wealth share from 23.2% to 35.8%. Their model attributes about two-thirds of the top-1% wealth-share increase to unequal access to private innovation.
That reframes the concentration argument entirely. The mechanism is not that wealthy people make braver bets — it is that the highest-return assets are structurally unavailable to public-market investors. Diversifying into public markets is not timidity; for most people it is the only door that opens. single source Working paper, not yet peer-reviewed, but methodologically serious.
The leverage is at records — FINRA margin debt hit roughly $1.4T in July, up 38.6% year-on-year. But the Yale Budget Lab, citing Liscow and Fox, finds borrowing is only about 1% of income for the top 0.1% by net worth. Total borrowing by the ultra-wealthy runs around $260B, of which perhaps $140B is attributable to unrealised gains. About 35,000 households would be covered by a borrowing tax — and roughly half the revenue comes from about 400 people.
The part doing the work is step-up in basis at death, not the loan. Anyone selling this as a replicable strategy at a $5–50M net worth is selling something.
US estate exemption is now $15M per person, $30M per couple, permanent, effective 1 January 2026. Removing the sunset killed the deadline-driven gifting industry and raised the threshold at which complex estate structuring pays for itself.
QSBS was rewritten for stock acquired on or after 4 July 2025: per-issuer cap $10M → $15M, company asset ceiling $50M → $75M, and the holding period is no longer all-or-nothing — three years now gives a 50% exclusion, four years 75%, five years 100%. Shares issued before and after that date are now different assets with different optimal hold periods.
Private placement life insurance is under direct legislative attack for the first time. An 18-month Senate Finance minority investigation found at least $40B sheltered through PPLI, which is 0.003% of outstanding life policies, with no IRS reporting requirement at all. The Wyden bill introduced 13 April is unlikely to pass this Congress; the reporting gap can close administratively without it, and probably will first.
The capital gains inclusion rate is 50%. It was cancelled, not deferred — by the Prime Minister on 21 March 2025, confirmed in Budget 2025. A large number of accounting-firm pages still say "deferred to 2026." They are stale.
The Canadian Entrepreneurs' Incentive was cancelled in Budget 2025 and never became law. What survived and matters: the Lifetime Capital Gains Exemption at $1.25M on qualified small business corporation shares, retroactive to 25 June 2024. With the inclusion rate holding at 50% and the CEI gone, the LCGE is now the single largest tax lever available to a Canadian owner.
2026 registered limits: TFSA $7,000 (cumulative room $109,000 if never contributed since 2009); RRSP dollar limit $33,810; FHSA $8,000 per year.
Bare trust reporting is finally live. Bill C-15 received Royal Assent 26 March 2026, ending three years of relief. Obligations begin with the 2026 tax year and first returns are due 31 March 2027. Nominee corporations holding real estate, unrelated-party joint ventures, and "in trust for" accounts above thresholds are caught. Penalties run $25/day, and gross negligence draws the greater of $2,500 or 5% of the highest property value.
Context first, because it colours everything: PCE inflation is 3.7%, Kevin Warsh chairs the Fed, and markets went into the 16 September meeting pricing a coin-flip chance of a rate hike. National debt has crossed $40T. This is a stagflation-scare tape, not a soft-landing one.
Ray Dalio (21 Aug) projects $7.5T of outlays against $5.5T of revenue and a sovereign debt crisis within roughly three years. Druckenmiller's WSJ op-ed (24 Aug) puts the deficit above $2T and debt above $40T. Dimon (7 Aug) points to infrastructure, global deficits and remilitarisation all pushing long yields up. There is no meaningful dissent from this among named principals.
Across 307 family offices averaging $2.7B of family net worth, 65% expect confidence in the dollar as a reserve currency to weaken; only 6% expect improvement. Euro and Swiss franc named as preferred alternatives. Separately, 60% plan to change strategic asset allocation within twelve months — the highest reading UBS has ever recorded.
Even the bears concede the technology. Howard Marks: AI is "very real, capable of doing a lot of work that heretofore has been done by knowledge workers." Jeremy Grantham agrees it is transformative — and says that universal recognition is precisely what produced "dangerous overinvestment." Nobody credible is arguing the technology doesn't work. Everybody is arguing about what it's worth.
More informative than the consensus, and currently sharper than it has been in years.
A genuine public rupture between a mentor and his protégé. After Treasury doubled long-dated buybacks from $2B to $4B per operation on 19 August, Druckenmiller went to print.
"This wasn't liquidity management, it was price management… Every basis point of artificial yield suppression is a subsidy to procrastination. If the 30-year must trade at 5.5% to clear, that isn't a crisis. It is an invoice." Stanley Druckenmiller · WSJ · 24 Aug 2026
Jon Hilsenrath's summary: "Those are two diametrically opposed views of the world." Note that Bessent is grading debt he issues.
Warsh moved this from a ~30% view to the base case in a single Jackson Hole speech on 28 August: "The responsibility for 65 months of sustained, elevated inflation sits squarely with the central bank… Otherwise, we have work to do." Market-implied odds roughly doubled overnight. On the other side, David Einhorn's Q2 letter (14 Aug) states Greenlight is positioned for no hike in 2026 — real money against the consensus.
Burry's is the harder and more falsifiable claim: not that AI is overvalued, but that hyperscaler earnings are overstated. It landed hard enough that Nvidia emailed sell-side analysts a memo rebutting him — an issuer lobbying analysts against a short seller is itself a tell.
Howard Marks occupies the only interesting middle ground, with a distinction the bears skip: "More money is going into inference capex these days than training capex… inference capex is taking place in response to actual demand." His advice — "no one should go all-in without acknowledging risk of ruin" — is the most defensible position in the argument.
Telling detail: when Huang declared AGI on 9 September, Nvidia fell 2%. Gary Marcus asked publicly whether Huang "had a financial incentive to declare AGI." The market appears to agree with Marcus.
Dalio and Cathie Wood are exact opposites. Dalio (21 Aug) wants 10–15% in gold, a small bitcoin allocation and reduced fixed income. Wood forecast 5% GDP growth with deflation, argued gold repeats its 1980 fate — "over the next five years it dropped 67%" — and expects the bitcoin-to-gold ratio to resume its uptrend.
Scoreboard as of today: PCE 3.7%, gold +35% year-on-year, bitcoin around $78K and falling. Wood's call is losing on both legs. Einhorn, meanwhile, lost money on gold and rates in Q2 — so the crowded trade has hurt believers too.
A large share of "what wealthy people do" content is marketing. These are the specific things encountered in this scan that should not be believed.
This is the largest finding in the section. Henley & Partners — the origin of nearly every "X thousand millionaires left Britain" headline for three years — abandoned its millionaire headcount in the 2026 report and replaced it with a score out of 100. The Financial Times found its numbers failed standard forensic accounting tests and were "highly likely to be fabricated"; Dan Neidle's analysis showed digits ending in 0 and 5 wildly over-represented and UK millionaire figures overstated by roughly 100% against ONS and HMRC data. The datapoint that replaced the UK figure is applications to Henley's own migration services. Henley sells residency-by-investment. Meanwhile actual HMRC statistics published 30 July show the non-dom tax take rose 9% to £13.6bn and departures fell.
Scion Asset Management deregistered with the SEC on 13 November 2025 and has filed no 13F since Q3 2025. Any content describing new Burry positions from a 2026 filing is fabricated, and at least one site is publishing it. His writing at Cassandra Unchained is real and current; the portfolio disclosures are not.
A widely-syndicated page headlined as a September fund manager survey carries survey dates of 5–11 September 2025 and numbers that directly contradict the genuine August 2026 survey — cash 3.9% vs 3.5%, US 14% underweight vs 27% overweight, "Long Mag 7" most crowded vs "long semiconductors." A live trap for anyone searching by month name.
Soros's "sevenfold move into an AI chip stock" is a Micron position of $25.9M inside an $8.14B portfolio — about 0.3%. Multiple outlets ran it as conviction. And "institutions are piling into bitcoin" rests on August's +$3.52B while US spot bitcoin ETFs remain net negative $1.77B year-to-date, with September opening on a $236M outflow day.
Searches for Anthropic IPO access return almost entirely affiliate and broker content selling "pre-IPO" shares. On 2–3 September, US prosecutors charged two former Linqto CEOs over a $450M pre-IPO fraud scheme; one pleaded guilty. The alleged mechanism was price inflation on pre-IPO shares — precisely the product currently being advertised.
Sam Altman, 3 September: warns of "unsustainable silliness" in new neoclouds and says "I am worried about the world's compute buildout plans" — immediately followed by "I'm not worried about our compute buildout plans." OpenAI targets roughly $600B of compute by 2030. That is a competitive attack dressed as candour. Add Larry Fink's "own assets" letter arriving attached to BlackRock's infrastructure product suite, and Einhorn calling a top while down 4.3% in a quarter the S&P gained 15.2%.
Knight Frank's index tracks auction results for objects that came to auction. The headline "art +11%, Old Masters +68.7%" reflects what sold, not what collectors hold. The honest number in the same report is ten-year growth of +38.6% — roughly 3.3% a year before insurance, storage and ~25% auction friction. That is a real-terms loss, published by a brokerage selling to the same clientele.
The evidence in this edition points at three levers and away from most of the rest: own a concentrated productive asset, get the gain out at the lowest available rate, and refuse to pay rent for "access" that isn't access.
Before holdcos, trusts or private deals: TFSA room for 2026 is $7,000, with cumulative room of $109,000 if never contributed since 2009. RRSP dollar limit is $33,810. Nothing in the structural section beats tax-free compounding on a risk-adjusted basis at this size, and most people are wrong about their own numbers.
With the inclusion rate settled at 50% and the Canadian Entrepreneurs' Incentive cancelled, the $1.25M Lifetime Capital Gains Exemption on QSBC shares is the largest single tax lever available to a Canadian business owner. It is also conditional: the company must meet active-business asset tests, and a company carrying too much passive investment or redundant cash fails them. Purification takes time, which means the answer matters years before a sale, not months.
The research says the excess return came from access to private assets, not from the asset class itself. What is being retailed — through the DOL's 401(k) proposal, evergreen funds, interval funds, non-traded BDCs — is fee-bearing exposure, arriving at the exact moment institutions hold 33,575 unsold portfolio companies and retail credit funds are meeting redemption caps. Exposure is not access. The OFR named the risks in March: liquidity mismatch, valuation opacity, covenant erosion.
Bare trust reporting obligations begin with the 2026 tax year, with first returns due 31 March 2027. Caught: nominee corporations holding real estate, "in trust for" accounts above thresholds, joint arrangements between unrelated parties. Penalties are $25/day, and gross negligence draws the greater of $2,500 or 5% of the highest property value. Three years of administrative relief made this easy to forget; it ended in March.
The data is unambiguous in both directions and people usually only quote one half. Large fortunes are made through concentrated ownership of a productive asset the owner controls — and the owner-labour finding means that asset's value is partly a claim on continued personal effort, which is a risk most owners never price. Large fortunes are kept through diversification: the top 0.1% sit at 49% public equities. These are two different jobs done in sequence, not a debate to pick a side in.