Public Markets Digest
TL;DR
The proof points arrived all at once: Starship went orbital and delivered its first revenue mission, Micron printed a $54B quarter with $150B of take-or-pay backlog, and Nvidia answered the bubble debate with a record $150B buyback. Yet the market's verdict on all three was a shrug — the 29K payrolls print priced out the October hike, but the 10-year still holds 5.24%, and the high-beta chip names are quietly in correction territory while megacaps make records. Dispersion, not collapse.
Micron settled the "is AI demand real" debate with contracts — and the stock still can't rally
The strongest fundamental confirmation of the AI thesis anywhere in the complex landed Wednesday, and the equity fell anyway.
"Even at floor prices, we expect margins meaningfully above any prior cycle peak margins." — Micron FY26 Supercycle
+1 (from prepared remarks PDF)
The numbers are absurd by any historical standard: FQ4 revenue of $54.23B (+379% YoY), 87.0% gross margin, EPS of $33.42 against a $31.72 estimate. The load-bearing detail is contractual — 26 take-or-pay strategic customer agreements, ~$150B of RPO, and $32B of customer cash deposits extending into 2031. CEO Sanjay Mehrotra went further: "We do not have line of sight to when supply and demand will return to balance." This directly validates the component-cost-inflation leg of the capex story from last cycle's digest — memory pricing is pushing hyperscaler spend up, not demand weakness pulling it down (Micron FY26 Supercycle+1).
The market's response is the tell: the stock fell ~1% the day after the print and sits -11.4% off its high at a $1.21T market cap and just 14.45x earnings. At 14x for a company growing EPS 811%, the debate is purely about cycle durability — and management just answered it with contracts, not adjectives. One embedded warning: server content growth is now expected at a "modestly lower rate" amid tight memory supply, the first hard evidence that scarcity is rationing AI hardware builds.
What to watch: Buyback speculation as the next catalyst — management committed to returning 100% of excess cash, starting from a $73.5B cash pile and net cash of $68.3B.
The payrolls print took the October hike off the table — but the 10-year didn't get the memo
The labor market cracked, and with it the second leg of the Fed's hiking cycle moved from likely to doubtful.
"US job growth slowed more than expected in September and the nonfarm payrolls count for the prior two months was revised sharply lower, almost taking another interest rate hike from the Federal Reserve this month off the table." — Fed Week II
The +29,000 print (private payrolls +46K vs. 85K consensus, prior months revised down 60K) unwound the 66% October-hike pricing from two weeks ago. Yet the 10-year sits at 5.24% — still well above the policy rate — because CPI keeps climbing. Equities read weak data as good news: the S&P rose 0.7% Friday to 7,722.72, within 1% of its all-time high (Fed Week II). The other overhang resolved quietly: no shutdown — H.R. 6500 funds the government through December 11, 2026, which is why payrolls arrived on time.
What to watch: The late-October FOMC — if the Fed hikes anyway into a cracking labor market, higher-for-longer hardens; if it pauses, the 10-year's 5%+ level becomes the sole inflation-discipline mechanism.
Nvidia answered the bubble debate with the largest buyback authorization ever
With Burry's short still hanging over the complex, the strongest free-cash-flow franchise in the market put a floor under its own equity.
"Nvidia's board added a record $150 billion to its share buyback authorization — the largest single authorization ever, surpassing Apple's $110B from 2024." — Nvidia $150B Buyback
(from Business Times)
Total remaining authorization now stands at $235B through fiscal 2028, and the stock briefly touched a record $237 Friday. But under the megacap surface, dispersion is the real story: Micron -11.4% off its high, Intel -15.3%, Qualcomm -25.7% — already in correction territory — while the Dow snapped a five-month winning streak in September (Nvidia $150B Buyback). The Burry watch trigger (SOXX -20% or a 200-day break) hasn't fired at the index level. If the AI trade's marginal buyer shifts from the index fund to the issuer itself, the debate moves from "is the growth real" — Micron just settled that — to "what is the equity worth when growth normalizes."
What to watch: Whether the megacap/high-beta divergence widens — single-name corrections leading the index is how tops have historically formed, or how they haven't.
Starship went orbital and delivered the payload — the SpaceX proof point landed
The event the SPCX thesis was waiting for succeeded on September 28.
"Most importantly, Starship made it to orbit... We did our first-ever orbital insertion burn. We got on orbit." — Starship Orbital
+1 (from Ars Technica)
All 26 Starlink V3 satellites deployed with contact confirmed — the first revenue-generating mission, resolving the thread from last cycle's digest. The flight wasn't clean: early Raptor shutdowns forced a cautious go-for-orbit call, controllers cut the mission to ~2 orbits instead of 6, and the vehicle tipped over and burned on splashdown (Starship Orbital+1). But the commercial logic changed: the rocket can now do the one thing the valuation's launch-revenue layer requires. The bear case shifts from "can Starship work" to "can it work repeatedly and cheaply."
What to watch: Booster tower catch and V3 launch cadence — that's where the Starlink capacity story either compounds or stalls.
Qualcomm entered its binary event with the stock already round-tripped
The trial this market has tracked for months is underway — daily jury sessions in Delaware October 5–9 — and QCOM has already given back most of its re-rating.
"Qualcomm and Huawei announced a multi-year, broad patent cross-license agreement covering 5G, compute, AI, and networking." — Qualcomm Trial Week
The Huawei cross-license lifted shares 1.5% to $184.87, but the stock is -8.5% over five days and -25.7% below its 52-week high. The trial outcome puts a judicial price tag on Arm-licensing economics that underpin every Arm-based data-center entrant — a Qualcomm loss hands Arm pricing power and clouds the data-center roadmap (Qualcomm Trial Week). Meanwhile Nvidia's N1X ships this fall to six PC makers — Microsoft, Dell, HP, ASUS, Lenovo, MSI — converting client-compute share risk from headline to shelf-level, with the June announcement having knocked Intel and AMD down ~4.2% each in premarket.
What to watch: Trial verdict and early N1X sell-through — both hit the same client/data-center franchise from different directions.
What surprised us
- Micron couldn't rally on the best print in semiconductor history. A beat-and-raise with $150B of contracted backlog, 87% gross margins, and 14.45x earnings — and the stock fell. The market has fully pre-priced the supercycle and is now trading only the terminal question: when does it end? (Micron FY26 Supercycle
- The Fed's hiking cycle may be dying of natural causes. Two weeks ago a 66% October hike was the consensus; one 29K payrolls print later it's "off the table" — while the 10-year holds 5.24% anyway. The bond market has replaced the Fed as the discipline mechanism (Fed Week II
- Nvidia's record buyback is a defensive move, not a victory lap. It arrived the same month AMD and Intel outran it on the chart and Micron's print flagged memory scarcity squeezing Nvidia's own server content growth — the endgame move of a cash machine, read however you like (Nvidia $150B Buyback
- The unlock-overhang thesis went 0-for-3. SpaceX absorbed the unlock, then delivered the orbital proof point days later — execution milestones, not share flow, are setting this stock's path.