AI capacity is live. Now comes the payback test.

TUESDAY · SEPTEMBER 22, 2026 Before the U.S. open. Market studies use Monday, September 21; all dollars are USD.

The morning brief

• Capacity becomes real. Microsoft’s Hyderabad region is live; the investment question moves toward utilization and cash generation. [1]
• Distribution meets adoption. Bedrock adds model choice; Amazon Quick targets finance workflows. Product availability alone does not establish profitable demand. [2–4]
• Three wheel decisions. Microsoft and Amazon are conditional; Micron is a pass before earnings. Premium never eliminates the ownership obligation.

The lead · Hyderabad opens; the payback clock starts

Microsoft’s September 21 announcement identifies Hyderabad’s India South Central region as live, with three availability zones and further Azure and AI services arriving progressively. Its mechanical-cooling claim concerns effectively zero water use for cooling, not the entire facility’s water footprint. Those boundaries matter when judging what the infrastructure can deliver today. [1]

Our investment reading: local capacity can make deployment easier for customers whose workloads need nearby computing and appropriate data controls. The economic chain still has several links: a usable service, a production workload, repeated consumption and cash collected. A region opening establishes the first opportunity; it does not disclose the utilization rate or prove the eventual return on the equipment.

For Microsoft, the attraction is the possibility of earning recurring cloud revenue from infrastructure that already exists. The counterargument is that depreciation, power and staffing arrive before every customer scales. A slower service rollout or price competition could stretch payback even while usage grows. Watch for three milestones in subsequent disclosures: which services become available locally, whether customers expand beyond pilots, and whether cash generation keeps pace with investment. That is a more demanding test than counting announcements—and a useful discipline after a strong AI-led market session.

Eddie’s watchlist radar

• MSFT · Deployment next. The new region moves the story from construction toward customer use. Next question: service availability and paid workload expansion, not another capacity headline. [1]
• AMZN · Check the date. Yesterday’s Grok 4.6 Bedrock guide explicitly dates its launch to August 18. Treat it as a distribution update, not a September 21 launch. [2]
• MU · The next evidence arrives September 30. Micron confirms its earnings call at 16:30 Toronto. Guidance and cash conversion can reset expectations before October options expire. [9]
• META / INTC · Price is not an order book. Monday’s roughly 11% / 12% gains reflected renewed AI-agent enthusiasm. Look next for paid adoption and actual chip demand; a rally alone confirms neither. [21]

AI buildout & revolution

Agents enter the finance department. AWS’s September 21 X demonstration promotes Quick for reconciliation and month-end reporting; its product page describes connected finance workflows. These are vendor claims, not audited time savings. The investable test is repeated use with fewer corrections, controlled access and a cost per completed task that beats the old process. [3–4]

Productivity is not automatically a rate cut. A September Chicago Fed paper models how expected future productivity can pull spending forward and raise the natural interest rate before the gains arrive. This is a conditional model, not a forecast or proof that AI caused current inflation. For capital-intensive AI businesses, financing assumptions still deserve scrutiny. [5]

The comparison · Three very different earnings engines

Margins measure operating profit per revenue dollar, not cash left after equipment spending. Amazon includes retail; Micron’s memory economics are cyclical. The highest bar therefore does not identify the safest wheel stock. Ownership also needs a sustainable cash-flow case and an acceptable entry valuation. [6–8]

3 Wheel Trade Ideas

Studies, not orders. These October 16 puts have 24 calendar days remaining. Monday options require repricing at today’s 09:30 Toronto open. Exact bid/ask timestamp, provider timezone/delay, OI date and IV rank are unavailable; last-trade times below are not quote timestamps. Verify standard 100-share deliverables and events before considering entry. [14]

Daily Wilder RSI14 and SMA20/50/200 use adjusted completed closes; relative volume compares Monday with the prior 20 sessions. None is oversold. Delta is not assignment probability. The Fed’s 3.75–4.00% range raises the opportunity cost of reserved cash. [11]

MSFT · Microsoft

Conditional · confirm support and events · The setup: sell the $475 put, expiring October 16, 2026. One contract means agreeing to buy 100 shares at that price if assigned. A study of owning the business on a pullback.

• Why own the shares: Recurring cloud demand helps, but rates discipline valuation. June quarter: $19.64bn operating cash less equipment; $76.84bn cash/short investments versus $40.29bn debt, excluding leases. Entry is 26.3× FY2026 GAAP EPS; investment gains flatter that denominator. [6]
• Trend and levels: Close $501.61; RSI 57.3; SMA20/50/200 $498.43/$466.54/$430.15; volume 1.32×. Above all averages; watch $498 first and $467 beneath it. [15]
• Contract quality: Bid/ask $3.80/$4.15; spread $0.35; delta −0.1969; IV 26.85%; volume 550/OI 2,461. Last trade September 21, 15:59:29, provider clock. [16]

AMZN · Amazon

Conditional · cash flow must recover · The setup: sell the $245 put, expiring October 16, 2026. One contract means agreeing to buy 100 shares at that price if assigned. The entry discount does not solve capital intensity.

• Why own the shares: AI distribution offers growth, but investment consumes cash. June TTM FCF −$7.60bn; $122.99bn cash/securities versus $128.89bn long-term debt, excluding leases. Entry is 3.37× TTM sales per period-end share: not a low-multiple guarantee. [7]
• Trend and levels: Close $258.45; RSI 52.7; SMA20/50/200 $256.12/$256.10/$240.42; volume 1.30×. Just above short averages; require stability around $256. [17]
• Contract quality: Bid/ask $2.84/$2.93; spread $0.09; delta −0.2315; IV 30.73%; volume 1,831/OI 12,376. Last trade September 21, 15:59:16, provider clock. [18]

MU · Micron

Pass · wait until after September 30 earnings · The setup: sell the $900 put, expiring October 16, 2026. One contract means agreeing to buy 100 shares at that price if assigned. The larger premium comes with a larger event and cash burden.

• Why own the shares: AI memory demand supports earnings, but supply cycles threaten durability. May quarter adjusted FCF $18.3bn; $30.13bn cash/investments versus $5.72bn debt, excluding leases. Entry is 9.0× annualized quarterly GAAP EPS—not a forecast or a normalized valuation. [8]
• Trend and levels: Close $1,043.96; RSI 61.2; SMA20/50/200 $963.46/$928.55/$644.04; volume 1.13×. Above all averages, not oversold; $963/$929 are reference levels. [19]
• Contract quality: Bid/ask $15.35/$16.00; spread $0.65; delta −0.1613; IV 62.84%; volume 2,163/OI 6,161. Last trade September 21, 15:59:57, provider clock. [20]

Next on the radar

• Today · 09:30 Toronto: refresh option prices, bid depth and overnight news.
• September 30 · 16:30 Toronto: Micron’s confirmed earnings call. [9]
• October 14 · 08:30 Toronto: September U.S. CPI, two days before expiry. [12]

From YieldCove

Use the wheel calculator to compare cash committed with downside. Covered calls after assignment can cap recovery; ownership quality comes first.

Photo: Hyderabad Financial District, August 17, 2026. iMahesh / Wikimedia Commons, CC BY-SA 4.0. Geographic context, not Microsoft’s facility; 1,280px Wikimedia rendition, no crop. [13]

AI-assisted research and translation, separately counter-verified by Codex. Independent Grok review required before subscriber delivery.