AI’s split market: buyers, builders and three wheel decisions

TUESDAY · SEPTEMBER 15, 2026 · Before the U.S. open · September 14 closing data · USD throughout. The Fed meeting begins today; its decision is tomorrow. [10]

The morning brief

• The AI trade split. Alphabet and Meta gained while Nvidia, Micron and Marvell fell. The chart measures the move; it does not prove why investors traded. [13,15,17,21–22]
• Control becomes a product question. Microsoft published a draft code for future MAI models. Existing models are not yet trained on it; distinguish a proposed standard from delivered capability. [1–2]
• Premium needs context. GOOGL is conditional; META needs a cooler entry; MU is a pass before earnings. Monday’s options are dated studies, not executable Tuesday bids.

The lead · Buyers and suppliers face different AI risks

Monday: Alphabet +3.22%, Meta +2.71%, Micron −5.25%, Marvell −7.32%, Microsoft +1.97%, Amazon −1.26%. These individual moves do not prove that all cloud buyers benefit from hardware suppliers’ difficulties. [13,15,17,19–21]

Our interpretation: slower frontier development could reduce urgency to buy capacity, while established advertising, cloud and software businesses retain useful applications to sell. Suppliers depend more directly on orders and capacity utilization. Can buyers earn more per dollar spent, and suppliers sustain margins if purchasing becomes less urgent?

Better efficiency could expand total AI demand by making more workloads affordable. Commitments, replacement cycles and inference needs may support orders despite sentiment. One session cannot resolve these paths or isolate price drivers. Follow workload growth, pricing and cash generation. A wheel candidate must remain acceptable to own if the favorable scenario fails; premium cannot replace that judgment.

Eddie’s watchlist radar

• GOOGL · Compute efficiency. Monday’s Dataflow release adds batch Pause/Resume and Blackwell G4 support: a product milestone, not disclosed revenue. Next: adoption and cost per completed workload. [3,12]
• MSFT · Rules before implementation. The September 14 consultation targets a year-end code revision and model development from 2027. Await evaluations and enforceable behavior before inferring safer deployed systems. [1–2]
• META · Monetization. Latest-quarter ad impressions rose 14%, average ad price 12%. These company metrics do not isolate AI’s contribution. Next: cash left after expense growth and capital spending. [6]
• MU · Calendar risk. Fiscal Q4 call: September 30, 16:30 Toronto, inside the October study. Watch memory prices and customer commitments; a stock decline does not prove expectations have reset enough. [8]

AI buildout & revolution

Recover work, do not repeat it. Google says Dataflow can preserve completed batch processing during a pause and resume later. Only qualifying pipelines are eligible; archived data remains billable. Our lens: less repeated work may lower cost per useful result without a bigger model. Compare identical workloads; no universal savings percentage is supplied. [3,12]

Factory agents need boundaries. Google’s September 14 manufacturing blueprint covers maintenance, procurement and security with defined permissions and human oversight. This is vendor guidance, not an audited deployment study. Our lens: integration, monitoring and reliable data access create opportunities, but plausible recommendations are insufficient when errors can interrupt production. [4]

The market in one picture

The strongest-to-weakest spread was 10.54 percentage points: one session’s dispersion, not a diversified portfolio or proof of permanent winners. Seek confirmation in orders and earnings. Shared AI exposure can produce different cash needs and drawdowns.

3 Wheel Trade Ideas

How to read the cards: conditional, standard 100-share cash-secured-put studies, all expiring October 16: 31 calendar days from today. Credits use the saved bid, with no fill assumed; amounts exclude fees and taxes. Confirm the actual contract deliverable and reprice at the September 15 open. Cboe snapshots omit exact bid/ask timestamps and open-interest reference dates; last-trade times below are not quote times. [11,14,16,18]

Technicals use 500 daily adjusted Yahoo closes through September 14: Wilder RSI(14), SMA20/50/200 and session volume divided by the prior 20-session average. RSI is momentum, not rebound certainty; delta is sensitivity, not assignment probability. IV rank/percentile are unavailable. The Fed decision precedes every expiry. Early assignment is possible, and a technical level cannot guarantee an exit. [10–18]

GOOGL · Alphabet

Conditional · wait for confirmation · Dated study: sell the $330 put, 2026-10-16 expiry; one contract obliges purchase of 100 shares at $330 if assigned. Standard deliverable assumed; confirm with broker.

• Why own the shares: Cloud demand supports the business, but the $325.55 study entry values June-share equity at 74.7× trailing free cash flow. June cash/securities were $242.47bn versus $98.17bn long-term debt. Capital spending and dilution still matter. [5]
• Trend and levels: Monday $349.39; RSI 55.3; SMA20/50/200 $341.08/$346.77/$336.54; volume 1.56×. Above all three averages with stronger participation; $347 is a reference to hold and Monday’s $350 high a resistance area. [13]
• Contract quality: 330 put: bid/ask $4.45/$4.55; spread $0.10; volume/OI 1,732/8,473; delta −0.2376; IV 30.29%. Last trade September 14, 15:59:52, provider timezone unspecified. [14]

META · Meta Platforms

Watch only · momentum is stretched · Dated study: sell the $600 put, 2026-10-16 expiry; one contract obliges purchase of 100 shares at $600 if assigned. Standard deliverable assumed; confirm with broker.

• Why own the shares: Advertising supports ownership, but Q2 free cash flow was only $0.784bn. Cash/securities $90.26bn versus $83.66bn long-term debt. Entry $593.05 is 24.0× four times Q2 GAAP EPS: a rough screen distorted by legal/severance costs, not forward P/E. [6]
• Trend and levels: Monday $665.60; RSI 70.8; SMA20/50/200 $590.25/$602.56/$622.85; volume 1.08×. RSI is elevated, not oversold. The $623 long average is a distant reference; chasing Monday’s $669 high offers little technical patience. [15]
• Contract quality: 600 put: bid/ask $6.95/$7.20; spread $0.25; volume/OI 1,201/5,877; delta −0.1656; IV 39.22%. Last trade September 14, 15:57:51, provider timezone unspecified. [16]

MU · Micron

Pass · earnings inside the contract · Dated study: sell the $800 put, 2026-10-16 expiry; one contract obliges purchase of 100 shares at $800 if assigned. Standard deliverable assumed; confirm with broker.

• Why own the shares: May-quarter operating cash was $25.39bn; cash/investments/restricted cash $30.2bn versus $5.722bn current plus long-term debt. Entry $783.95 is 7.9× four times quarterly GAAP EPS, not forward P/E: exceptional memory margins can normalize sharply. [7]
• Trend and levels: Monday $924.03; RSI 46.4; SMA20/50/200 $960.43/$927.58/$625.43; volume 1.04×. Below the short averages, with no oversold reading. A recovery above $928 then $960 would improve the picture; neither level guarantees support. [17]
• Contract quality: 800 put: bid/ask $16.05/$16.95; spread $0.90; volume/OI 1,226/6,558; delta −0.1769; IV 58.90%. Last trade September 14, 15:57:38, provider timezone unspecified. High IV reflects risk. [18]

Next on the radar

• September 15 · 09:30 Toronto: U.S. open. Revalidate all studies against fresh prices; distinguish an opening rebound from a confirmed trend.
• September 16 · 14:00 Toronto: Fed decision; press conference at 14:30. Watch financing conditions and valuation assumptions, without presuming the outcome. [10]
• September 30 · 16:30 Toronto: Micron fiscal Q4 call, before October 16 expiry. Compare guidance with the exceptional May-quarter base. [7–8]

From YieldCove

Use the wheel calculator to compare cash, credit and downside. Three AI-linked names do not guarantee diversification. Record why you would own each before choosing a premium; no existing positions are implied.

Photo: Federal Reserve building, Washington, May 31, 2024. Tony Webster / Wikimedia Commons, CC BY 2.0. Resized; historical context for this week’s meeting, not a photograph of today’s event. [23]