High Risk

Recent Performance — Last 5 Sessions

Today = vs our Oct 26 mark ($1,342.23). That mark was published to the website on Monday but never emailed or texted, and Tuesday, October 27 was not published at all — no letter, no page, nothing. So measured against Friday’s $1,383.67 — the last figure that actually reached you — the book is +4.7%. Live marks (Wed, 7:34 MDT, four minutes after the open, 9:34 Eastern); the daily move is measured against our Oct 26 (Monday) published mark, which is the most recent mark we published anywhere. Two consecutive sessions failed to reach you, and they failed differently. Monday, October 26 reached the website but was never emailed or texted. Tuesday, October 27 was not published at all — no letter, no page, nothing. So the last figure that actually reached you is still Friday’s — three bases today, and all three print. The direction matters, so read it before the numbers: our last published mark was taken on Monday at 10:05 Eastern, and the tape kept rising across the rest of Monday and through Tuesday, so Tuesday’s official close landed above that mark. That window is a session and a half, not an afternoon. It makes the reading against our own mark the more flattering of the two — your own screen measures from Tuesday’s official close and will look worse than it. Both readings are printed, for the book and for every position. October MTD is measured from Wednesday Sep 30’s official month-end close. October MTD vs Sep 30 close ($1,733.06). Tuesday’s official close was $1,440.89 — 7.35% above our mark — so on the close basis today is +0.5%. Sessions before Jul 14 reflect the prior book; the current holdings were established at the Jul 14 open.

Baseline note — Three numbers today, not two, and the reason is our failure rather than the market’s. Two consecutive sessions failed to reach you, and they failed differently. Monday, October 26 reached the website but was never emailed or texted. Tuesday, October 27 was not published at all — no letter, no page, nothing. So the last figure that actually reached you is still Friday’s — three bases today, and all three print. Worst first, ordered by the numbers themselves. Measured from Tuesday’s official close of $1,440.89 — the basis your own brokerage screen uses it is +0.5%. Measured against the $1,383.67 you were last actually sent it is +4.7%. And measured from our own Oct 26 published mark of $1,342.23 it is +7.9%. The 7.39% between those two is the distance the book travelled between our Oct 26 mark and Tuesday’s closing bell. The reading that matters most is the one against $1,383.67 — the figure you were last actually sent, on Friday Oct 23 — and that is +4.7%. Your brokerage screen will show you the +0.5% reading, measured from Tuesday’s official close; we print all three. October is -16.4% and the year -11.5%. October MTD is measured against the Sep 30 official month-end close of $1,733 for the current whole-share model. September is locked closed in the monthly table at +19.50%.

Position Scoreboard — Sorted by Portfolio Weight

Today = vs our Oct 26 mark, with the vs-Tuesday’s-close reading beneath it wherever the two differ. Since Fill = vs SOXL $179.62 / TQQQ $75.19 / MSTR $94.78 / BE $205.15 / NBIS $172.19. Weight = % of unit. Live marks (Wed, 7:34 MDT, four minutes after the open, 9:34 Eastern); the daily move is measured against our Oct 26 (Monday) published mark, which is the most recent mark we published anywhere. Two consecutive sessions failed to reach you, and they failed differently. Monday, October 26 reached the website but was never emailed or texted. Tuesday, October 27 was not published at all — no letter, no page, nothing. So the last figure that actually reached you is still Friday’s — three bases today, and all three print. The direction matters, so read it before the numbers: our last published mark was taken on Monday at 10:05 Eastern, and the tape kept rising across the rest of Monday and through Tuesday, so Tuesday’s official close landed above that mark. That window is a session and a half, not an afternoon. It makes the reading against our own mark the more flattering of the two — your own screen measures from Tuesday’s official close and will look worse than it. Both readings are printed, for the book and for every position. October MTD is measured from Wednesday Sep 30’s official month-end close.

Portfolio Allocation by Category

Live Track Record — Since January 22, 2026 (Actual)

▲ Current month through today; July, August and September 2026 are locked closed; return measured from the Jan 22, 2026 entry. * Jan–Mar 2026 are estimated partial-period returns carried from our May 2026 updates; they predate our daily archive (begins 15 April 2026) and are shown as estimates, not as-published daily marks. Apr–Sep cells are locked closed from official month-end closes; close-basis restatement pending the anchor review. No backtests, no hypotheticals. Units are price-return; dividends are not accrued.

Calculate Your Position Size

High Risk is marked at $1,449 on the current whole-share model, and before any of it: this is the third letter running that you should have had and did not. Oct 26 reached the website and no further — no email, no text. Oct 27 reached nothing; it is the first session all month that produced no page at all. The last numbers you actually hold are Friday Oct 23’s, $1,383.67 a unit. So there are three comparisons below and they are three different stretches of time. Against Tuesday’s official close of $1,440.89 — one ordinary session, and the basis your screen uses — the book is +0.5%. Against our Oct 26 published mark it is +7.9%; against the Oct 23 figure you received it is +4.7%. We lead on the smallest, and on a day when all three are positive that rule matters more, not less. And three green readings do not change the standing: October is -16.4% month to date and the year is negative at -11.5%.

The +7.9% is the number we most want you to distrust, and it is ours. It is not a day. It measures our Oct 26 morning mark against this morning, and 7.35% of it landed before today’s opening bell — during Oct 27, the session we never published. Today’s own contribution is +0.5%. The same distortion runs through every line in the table: SOXL reads +21.9% against our Oct 26 mark and +2.8% against Tuesday’s close. Both are arithmetically right. Only the second one is a day. A triple-leveraged fund printing +21.9% in a column headed with a single date is exactly the kind of number that would flatter us if we left it unexplained, so we are explaining it instead.

Why the chips moved: memory, and the numbers are genuinely extraordinary. SK Hynix reported before our mark and it is the cleanest read yet on the AI memory shortage. Third-quarter revenue 104.6 trillion won, up 327.9%, and operating profit 83.3 trillion won, up 632.1% — a 79.7% operating margin, a sixth consecutive record quarter, and management saying DRAM and high-bandwidth memory are effectively sold out through 2029, with signed agreements already covering more than half of 2027 and 2028 capacity. It pledged roughly 100 trillion won of dividends and buybacks. Micron is +5.02% on it, Lam Research +2.29%, Broadcom +2.19%, AMD +1.44%, and the Philadelphia semiconductor index +0.86%. One line of that release deserves your scepticism, though: net income rose only 215.7%, to 39.8 trillion won, because 32.6 trillion won of non-operating expense sits between the operating line and the bottom line. The 632% is real and it is also the flattering cut of it.

And why Oct 27 moved, which is most of what you are looking at. Chips rose a third straight session, the semiconductor index +3.60% to 11,482.50. Two things did it. Anthropic filed its S-1 to list on Nasdaq under ANT, and Broadcom — with Bank of America, Citigroup, Morgan Stanley and Blackstone — finalised a $63bn off-balance-sheet vehicle to buy AI chips and networking gear and lease it to Anthropic, covering 1.7 of the 3.5 gigawatts in their compute agreement, deliveries from 2027. That is an answer to the question this letter has been asking since October 15: who actually funds the build-out. The answer is a syndicate of banks and a private-equity firm, through a leasing vehicle that does not sit on Broadcom’s balance sheet. An answer is not the same as an absence of risk, and we would rather you held both thoughts than either one alone.

The prospectus itself is the most important document the AI trade has produced, and it cuts both ways. Anthropic’s S-1 shows first-half revenue of $16.3bn against about $1.2bn a year earlier — more than thirteen-fold — and third-quarter revenue of $16.8bn, which confirms the 45% sequential figure we printed to you on October 15. It also shows a net loss of $80.7bn, and that number has to be taken apart rather than repeated. About $72.4bn of it is non-cash — a fair-value remeasurement of the convertible notes, SAFEs and derivative liabilities used to raise money, which grows precisely because the company is being valued higher. Take it out and the operating loss for the half is $693m, against $2.1bn a year earlier; on the company’s own adjusted measure that becomes $1.6bn of operating income, and the half was free-cash-flow positive. So the honest reading of this filing is not a company running out of money, and we are not going to sell you one. The figure we would actually underline is the forward obligation: roughly $417bn of aggregate future minimum payments for compute and infrastructure, which reported tallies put nearer $518bn once unfinalised Google TPU capacity is counted — about $161bn to Broadcom, $111bn to Google and $110bn to Amazon — with roughly 80% of it non-cancelable. Set that against $16.3bn of half-year revenue. That, and not the loss line, is the shape of the risk in this book: not that demand is fake, but that the commitments are enormous, long-dated and mostly payable whether or not the capacity is used. And one thing we will not overstate, because it would be easy to: customer concentration here is low — no single customer was more than 10% of revenue through September, and the five largest together were about 25%. A correction we owe you in the same breath: on October 15 we printed that Anthropic’s expected IPO valuation had ‘fallen below $1.5 trillion’. Reported discussions are now $1.5–2.0 trillion. That is our own number moving against us and we would rather say so than let it quietly disappear.

Position by position, both bases. Four of the five are higher against Tuesday’s close. SOXL is +21.9% against our Monday mark, +2.8% against Tuesday’s close and TQQQ is +7.2% against our Monday mark, +1.1% against Tuesday’s close — the two leveraged chip funds that did most of this month’s damage are now doing most of the repair. BE is +12.9% against our Monday mark, +2.2% against Tuesday’s close, still working back from the post-earnings fall we wrote about on October 23, when a record quarter was met with a 12% drop. NBIS is +13.6% against our Monday mark, +0.5% against Tuesday’s close, +0.06 points of the book. MSTR is +0.7% against our Monday mark, -1.2% against Tuesday’s close and is the only position lower against the close, costing the book -0.47 percentage points; hold it flat and High Risk would read +1.0% instead of +0.5%. One position flips sign between the two bases today. MSTR is +0.7% against our Monday mark, -1.2% against Tuesday’s close. Where the readings differ, the close-basis column is the one your statement is measured from.

Strategy reports tomorrow and it is 39% of this book, which makes it the single largest thing that can happen to you this week. Results come after the close on Thursday October 29, with a webinar at 5:00 p.m. Eastern — that is from Strategy’s own announcement of October 8, not a data vendor, and one vendor carrying ‘Friday October 30’ is simply wrong. Going in: the company has paused bitcoin purchases, repurchased $204.9m of its STRC preferred, and holds 848,517 bitcoin at an average cost near $75,446. Bitcoin is about $80,612 at our mark, essentially flat on the session at -0.12% and holding above $80,000 throughout, which puts the holding roughly 7% above cost — a thinner cushion than this position has had for most of the year. On MSCI, nothing has changed and the mitigant stays printed next to the risk: the affected-company list publishes November 11 and any exclusion takes effect after the close on November 30, but an existing index member is only removed if it fails the screen in each of its two most recent annual filing periods. On Nebius, we still have no company-confirmed reporting date. Vendors estimate November 10 while stating the company has not confirmed it. That has been our position since October 19 and it is a caution, not a date.

Bottom line: High Risk is at $1,449 a unit, +0.5% against Tuesday’s official close, +7.9% against our Oct 26 mark across two sessions, and +4.7% against the Oct 23 figure you last received. October is -16.4% and the year is negative at -11.5%. Both of those need saying straight. On our published monthly table this is the worst October this book has recorded — the previous worst was -12% in 2025 — and we checked it against every October in the table rather than asserting it. But it is not the worst month we have had: 6 months in the record are worse, including February 2025 at -23% and July this year at -22%. It is also better than it was three sessions ago, and we are not going to pretend otherwise in either direction. Nothing was traded and nothing is proposed; every position is held unchanged and no trade was made at any point. What we are watching: the Federal Reserve at 2:00 this afternoon, which lands after this mark and is in none of these numbers; Strategy’s results after tomorrow’s close, the largest single event for this book; whether Nebius confirms a date; the MSCI list on November 11; and the midterm elections on November 3.

Portfolio Builder | Market Update | Wednesday, October 28, 2026

Trade Ledger — Entry & Exit History

Every position we open and close, newest first. Realized P/L on closes is marked against the original entry price.

Entry prices verified against the Jan 22, 2026 cost basis. “Open” = still held today. An entry marked “Closed <date>” was exited in that rotation — its realized result is shown on the CLOSE row for that date. This ledger builds our public track record going forward.