Strong chip demand does not settle the AI-return question

Strong chip demand does not settle the AI-return question

Research cutoff: October 15, 2026, 10:15 UTC / 06:15 Toronto. This review covers developments since 04:12 UTC. The research horizon remains February 13, 2027; individual contracts have earlier exit dates. All probabilities and price outcomes below are ATG judgments, not market-implied probabilities or promises.

Decisions in brief

Hold one December micro Nasdaq, one March micro copper and one March SOFR. Add none. Exit the April micro gold. Enter three March micro soybean-oil/meal pairs. These are simulated decisions, not real orders.

The useful distinction this morning is between strong demand for a supplier’s products and attractive returns on its customers’ investment. TSMC’s outlook supports retaining the small Nasdaq recovery position. It does not remove the risk that expensive AI infrastructure earns disappointing cash flows. Copper likewise retains a physical-tightness argument without requiring an outright global shortage. Gold’s rebound, by contrast, has consumed almost all the remaining modeled upside. Oil versus meal has become cheaper enough to warrant a small new position without waiting mechanically for the next release.

What is new—and what was already known

TSMC’s October 15 release guides fourth-quarter revenue to $52.3–54.3 billion and reports a 66.4% third-quarter gross margin. The quarterly revenue total had substantially been disclosed on October 8; today’s useful information is profitability and the forward outlook, not a second revenue surprise. We interpret this as evidence against an immediate collapse in chip orders, not proof that AI customers will monetize their spending. TSMC results and guidance.

Europe is not delivering a uniformly dovish signal. August euro-area industrial output fell 0.5% month-on-month, with July revised to −0.3%, but capital-goods production rose 0.4%. France’s final September harmonized inflation rate remained 3.4%; national CPI was revised up from 3.0% to 3.1%, and services inflation reached 2.4%. Weak output and sticky inflation pull the euro thesis in different directions. These releases concern August/September activity, not October measurements. Eurostat, INSEE final, preliminary estimate.

A genuinely new energy risk is the renewed attack on Russia’s Ryazan refinery. Reuters’ October 15 report, first published at 06:03 UTC, cites the regional governor’s confirmation of a fire. Damage and lost output remain unquantified. This weakens confidence in smooth product-supply normalization; it does not establish a full shutdown or a particular loss of diesel exports. Reuters.

The earlier shipping-sanctions proposal, API inventory estimate and October 14 Gulf-production recovery remain context, not fresh catalysts. Proposed sanctions relief is not an observed insured cargo. A refinery outage can tighten refined products while reducing crude intake, so “buy crude” is not automatically the right expression.

Dated price map

These are native-contract last prices returned by IBKR, not presumed settlements. All dates are October 15, 2026; times are UTC. Retrieved bid/ask studies below were collected at 10:04:19 but do not have independent exchange timestamps. Legs are not claimed to be synchronized.

Source: Interactive Brokers, native contract observations retained with the research. Standard energy and oilseed prices are explicitly identified where used as micro-contract proxies. EUR/USD 1.1217 is separate spot context in the WSJ’s 09:17 UTC report, not a December micro-euro quote.

Re-underwriting the existing book

The following table measures remaining net payoff from each current mark. Entry costs already incurred are not charged twice. Remaining friction includes one adverse exit tick and $3.50 commission per contract. The gold row describes the decision immediately before closing it.

Nasdaq outcomes apply to the November 20 exit, not to an unrolled December contract held into February. Copper, SOFR and the former gold study use February 12 exits. Stops are management instructions, not guaranteed fills at those prices.

Nasdaq: improve the recovery probability, keep the financing tail

We move five probability points from the ordinary adverse branch to partial recovery, changing 30/30/30/10 to 30/35/25/10. The favorable outcome and gap probability are unchanged. Better supplier evidence merits this modest change; it does not justify raising every target or removing the financing tail.

One micro is $2 per index point. At the 29,842 ask, a new unit’s $8 full friction gives $1,048 EV and $613 ordinary buffered risk to 29,600. Moving ten favorable points to the 28,000 gap branch leaves $318. However, combining that probability shift with a 500-point reduction in both favorable and base endpoints produces −$232. A single unit’s buffered gap exposure is about $4,613, close to the roughly $5,013 total-thesis ceiling. An additional unit would breach that gap-aware budget and compound copper’s growth exposure. Hold; add zero. Invalidate on observed 29,600 or a material deterioration in AI demand/financing, rather than mechanically waiting for the price stop.

Copper: available metal is not the same as annual production

Antofagasta’s October 15 report gives Q3 output of 151.9kt, up 7% sequentially, partly from processing previously accumulated Los Pelambres inventory. Year-to-date output is down 8.3%; 2026 guidance stays 625–655kt and 2027 guidance is 630–690kt. The Centinela strike remains a constraint. This is mixed evidence, not a sudden mine-supply boom. Company report.

The latest verified London observation remains October 14: 228,000 tonnes of stocks and a $116/tonne cash premium over three-month copper. It has not been relabeled as today’s physical data. Westmetall.

ICSG’s October 14 forecast, newly incorporated here rather than treated as post-cutoff news, points to refined surpluses of 104kt in 2026 and 368kt in 2027. April forecasts were already 96kt and 377kt. The revisions are small; a global annual surplus still need not place immediately deliverable metal outside the United States. October reporting, ICSG April release.

One micro is 2,500 lb. The existing position remains worth retaining, but a new unit at 6.6815 has $261.75 EV after $9.50 friction. Five favorable-to-adverse points reduce that to $36.75; a further 5¢ reduction in the base endpoint makes it −$19.50. March also trades about 10.7¢ above December. That premium can erode; London backwardation is not free carry on COMEX March. Keep 6.35 invalidation, 7.00 review, 7.60 target and February 12 exit. A fall in physical premia with rebuilding available inventories or weaker consumption can invalidate sooner. Review physical evidence on October 16.

Gold: exit because forward value has narrowed

The weighted February exit assessment remains $4,250/oz. At $4,245.50, only $40.50 expected net value remains for a ten-ounce micro, against $2,448.25 buffered ordinary risk to 4,050. A five-point favorable-to-adverse shift removes $340. There is no verified new structural investment-flow evidence that warrants raising the forecast simply because gold recovered.

The decision is Exit, not a claim that 4,050 traded or that gold’s structural role disappeared. The one-unit position cannot be meaningfully trimmed. Closing it also reduces monetary-tightening exposure shared with SOFR. Re-entry requires a new price/evidence assessment, not regret over a subsequent rally.

SOFR: fragile Hold with a defined policy test

March’s 95.485 implies 4.515% compounded funding, while December’s 95.730 implies 4.270% for its different reference quarter. The 24.5 bp difference is a curve observation, not a literal count of meetings or guaranteed carry. Repo basis and risk premia matter. CME specifications.

Why retain SOFR while closing gold? SOFR’s remaining EV is about 6.9% of ordinary buffered risk, versus gold’s 1.65%, and the modeled adverse tail is smaller. More importantly, SOFR directly expresses a limited future funding peak with an imminent, observable policy-and-data test. It is not simply a second inflation-hedge narrative. Still, five favorable-to-gap probability points make its $84 remaining EV negative. A new unit at 95.490 has only $55.50 full-cost EV. Retain one; add none; review after today’s U.S. data and Warsh’s discussion. Strong underlying inflation plus a materially higher funding outlook can justify exit before the unchanged 95.100 stop. Exit by February 12, before the March–June reference quarter.

New starter: oil demand versus meal supply

Enter +3 MZLH27 and −3 MZMH27, a small relative-value exposure rather than an outright bet on soybean prices. More crushing creates both oil and meal. The thesis requires oil disappearance—particularly fuel-related demand—to absorb oil more readily than feed demand absorbs the meal byproduct. The EPA proposal remains a proposal; it is not booked physical demand. Greater competing vegetable-oil availability and disappointing fuel use are counterarguments.

The fresh price changes the decision. In a conservative book-based study, round the standard oil ask 66.97 up to the micro’s valid 66.98-cent tick and use meal bid $361.80. Each micro pair’s dollar-weighted price is 60 × oil − 10 × meal = $400.80. This is a calculation, not a quoted exchange spread. A standard pair has ten times the exposure and exceeds the usual starter budget at the same meaningful stress; three micro pairs fit.

Full round-trip friction is $61.20 for all three pairs: $42 commissions plus $19.20 adverse-tick allowance. EV is $324.60; conservative buffered study risk is $1,363.20. Ten favorable-to-adverse probability points leave $79.68; doubling costs as well leaves $18.48. A combined ten-point shift and 0.50¢ reduction in every oil endpoint makes EV −$10.32. This is a modest, fragile edge, not an invitation to scale immediately.

NOPA consensus already anticipates about 203.7 million bushels of crush and 1.116 billion lb of oil stocks. Low absolute stocks alone therefore need not surprise. Low stocks despite strong crush would support disappearance; weak crush can tighten meal as well and undermine this relative trade. Review the actual joint release at 16:00 UTC and exit if the mechanism fails. Reuters agricultural preview.

The protective price-ratio threshold is 100 × oil/meal ≤ 16.900524, corresponding to the 64.56/382 joint stress; the favorable ratio is 20.982143. These ratio triggers are not identical to a fixed dollar spread: another combination of prices can produce a different loss, and gaps can exceed the budget. Both micro contracts are cash-settled and last trade February 19 under CME’s calendar rule; exit February 12, with no automatic roll. The $2,500 reserve per micro, $15,000 total, is a deliberately conservative paper assumption, exceeding twice published maintenance of $205 oil/$150 meal, not an exchange initial-margin assertion. CME micro specifications, RHB margins effective October 6.

Challengers across the liquid futures board

These studies include full costs, not sunk-cost Hold accounting. One whole contract or package is tested before ranking. Standard references used for micros assume compatible same-month pricing; none is represented as an observed micro execution.

For wheat/corn the retained scenario weights are 20/45/35 at 230/195/150 cents. Harvest progress, previously announced Chinese buying and October 14 USDA balance commentary are context, not new overnight demand. CONAB and delayed export data are the next tests. The old March MES setup remains superseded by the December comparison. The INR policy idea remains a watch without a defined superior exchange-listed payoff; the earlier offshore response is not a new current quote. Historical cocoa, sugar, nickel, OAT/Bund, calendar-spread and rotation ideas remain retired/context, not silently reopened orders.

There is no newly verified CFTC positioning release at this cutoff. Existing positioning evidence is lagged, not proof of a fresh liquidation or squeeze. Friday’s report will describe Tuesday’s snapshot, before Wednesday’s CPI. Liquidity is assumed for the paper fills; observed depth and matched-leg execution are not claimed.

Next 72 hours: what can change the decisions

Calendars: CONAB, BLS, Census, CPC, EIA gas, EIA petroleum, NOPA, Federal Reserve, CFTC.

Assumed execution and simulated book

At 10:17:01 UTC on October 15, the paper engine closed one MGCJ27 at 4,245.40, using the exact 4,245.50 observation from 09:47:30 minus one adverse tick. Entry had been 4,167.80. Gross gain was $776; net after $7 round-trip commissions was $769. No historical stop or fill path was invented.

It also entered three long MZLH27 at 67.00¢/lb and three short MZMH27 at $361.80/ton. References were standard ZLH27 66.97 at 10:03:20, rounded up to 66.98 for micro tick compatibility, and ZMH27 362 at 10:02:39. One adverse micro tick per side was then applied. Equal same-month settlement basis is assumed; these were not observed micro-contract prints. Full-cost EV using those last-reference inputs is $330.60, slightly above the conservative book study's $324.60. The ledger's entry risk is $1,358.10; the more conservative comparison-table risk is not substituted into accounting.

The book now holds four positions: one SOFR, one micro copper, one micro Nasdaq and the three-pair oil/meal position. There are no pending orders. Simulated cash is $1,002,555.375, open gross P&L $57.15, and NAV $1,002,612.525 at the completion time. Cumulative closed-trade net P&L is $2,586.875 across eight closed positions. Margin reserves total $38,000; ledger open risk is $4,151.1625, about 0.414% of NAV. These measures use dated/proxy marks, not executable liquidation prices. Margin is reserved, not spent; cash earns zero in this simulation.

The economic decisions are distinct from those assumed transactions and from subsequent simulated performance. The central outlook is not a guarantee: correlated growth shocks, refinery disruption, policy surprises and relative commodity basis can produce losses beyond modeled stops. Small sizing is the response to that uncertainty, not evidence that the uncertainty has disappeared.