Buyback Misses, Meta Rips Into Resistance
US Stocks · Options · News · Views
Yields hit back. Russell is on the red line again
The index decline on September 9 in New York looked modest. Internals were not. SPX had 404 decliners and 99 advancers. Only energy finished higher among eleven sectors. Oil and Treasury yields rose together, and duration-sensitive growth and small caps were hit first. Russell gave back a week's gains in one session and is back on the 288 line. IWM closed near 291, low 290. 288 has not broken. Two sessions ago small caps and SOX were squeeze fuel. Today the fuel left Russell first as yields jumped.
September and October volatility is rising and spare buying power is thin. There is no continuity: up one day, down the next, shoot and move. Lose 288 and small caps go from "gave back a week" to another leg down. Hold it and risk has not gotten worse. That is not a new bid. Hedge funds were still near the 4th percentile a session ago. Nobody is paid to push the index through the prior high. Treat 288 as the switch. Do not read energy's lone green as risk appetite coming back.
A $6 billion Treasury buyback, and the 10-year still jumped
Treasury set Thursday's long-end buyback cap at $6 billion in 10–20 year notes. That is above the earlier "at least $4 billion" floor and triple the old ~$2 billion print. The tape had heard about $10 billion. BNP Paribas had said anything under $7 billion would trigger bond selling. $6 billion landed below the surprise line. After the release, the 10-year yield jumped to 4.84%–4.85%, closer to 5%. The 20-year and 30-year lifted toward 5.3%. Some voices say real damage needs 5.5% or 6%. 5% is still the psychological line. Break it and rating warnings plus speculative flows can chain in ways a model does not print. 4.85% to 5% is not far, but the buyback already showed the size was too small. The intervention failed for now.
Buybacks are not a standing tool. Occasional help is fine. Daily use does more harm than good to the bond market. Treasury held back, which means they are wary of emptying the clip. For growth and small-cap holders, watch the 10-year first, then the stock. Higher yields rewrite DCF discount rates. A drop back under 4.8% is what would give 288 room to breathe. Thursday's actual operation and the long-end auctions test whether 4.85% sticks. The size miss is already priced. A weak execution would bring 5% in faster.
SOXX closed 532. One more day for confirmation
AMD is still stronger than NVDA. After the volume break of 498, it is parked against 528. The close was about 521, the high about 527, with no clear stall. Clear 528 and new highs are open. Fail and it stays a hotspot, not a trend. Watch whether the next session is a wide-range up bar on volume or a spike that gets sold. That is a live tell. Today's tape cannot settle it. If the index confirms tomorrow, the leader clears more easily. If SOX is rejected, 528 becomes a chew-up cap.
SOXX closed exactly 532, after a high near 535. The falling trendline lives here, with a point or two of drawing error. Closing on the line is not confirmation. A clear gap up tomorrow, or a close above today, plus rising volume, would confirm the break. Today is still a touch short. The group has been strong. The tape has not been stable. Buying a full book on the breakout day gets punished when flows rotate overnight. Wait one more session. Trade semis as a hotspot. Wait for the close before calling a trend. NVDA is still funding higher-beta names. It is not today's line.
Meta launched Muse. Google and Amazon get diverted
META printed huge volume, cleared 632, and opened at 648 into right-side resistance. It closed near 654, high near 658. 680 is the right-side break. It rallied. It did not break. The product matters more. Muse, a personal AI agent, is live in the US. It connects Meta's own apps and third parties: Google Workspace, Ticketmaster, OpenTable, Spotify, Apple Health. It sends mail, books flights, fills forms, compares prices, pays, and runs 24/7 in the background against long-term goals. Light users get about 100 million tokens a week for free. Monetization is a take-rate on completed transactions plus a subscription for heavy users. The hurdle is lowered first. Pay for outcomes.
This is existing profit buying share. Bull case: the agent covers token cost quickly, contributes profit, and the subsidy can stop. Bear case: it runs at a loss for a long time and drags ad profits and the balance sheet. After Q2, META's AI monetization lagged GOOG, AMZN, and MSFT among the four. None of those three is a finished success. This is a catch-up move and a price war. Startups charge monthly. Meta lifted the free tier. Smaller firms are now competing on fundraising. Token prices are being cut across the industry. Grab DAU and MAU first, monetize later. If the pie grows, everyone can take more. If the pie is fixed, late entrants are fighting over a stock of demand.
For GOOG: Workspace is being borrowed. Search and shopping ads can be skipped by an independent browser. Gemini's native agent use cases get diverted. That is not an instant collapse of the search moat. ChatGPT did not steal search. The response is still a stronger first-party agent so users do not leave. For AMZN: Muse can browse, compare, add to cart, and pay. The default shopping destination weakens. If a lot of those orders still land at Amazon, near-term GMV can rise. Longer term Meta wants commerce of its own, so the competition is real. AWS remains the profit engine. Shopping diversion hurts less than cloud. MSFT overlaps Muse less. It was the least affected name today.
GOOG closed near 328. The 2026 forward band is 286–430, median 358; 2027 is 297–446. AMZN closed near 252. The 2026 band is 212–285, median 249; 2027 is 241–324, median 282. Versus those medians the shares are not expensive. The long-term slots still work. The weak tape versus META is relative diversion pricing, not a valuation break. META's near-term line is 680. The fundamental line is how long the subsidy lasts. Fail 680 and today's volume bar is just an emotion wave inside resistance.
Apple's foldable did not reprice the stock. Two prints tomorrow
AAPL launched a foldable iPhone. The stock whipped around and closed near the open, about 315. Loud process, flat result. Whether the new devices drive upgrades is a sales question, not today's close. Apple sat out the AI capex race, kept the thickest cash flow, and took the least damage among the four. A leader often copies the chase rather than running out in front. On valuation, the FY27 September upper band is about 288. After the Q4 print, FY28's upper band may clear 300. Spot is already against that cap. Sales data has to do the next work. On the chart, 303–318 is nearby support. Lose it and the middle is thin, with 281 in view near the 200-day. That is a range floor, not a one-way decline.
Tomorrow after the close: ORCL and ADBE. One trades the OpenAI compute story. One trades software being rewritten by AI. Tonight's map: 5% on the 10-year, IWM 288, SOXX a close above 532, META 680, and how long Muse can be subsidized. If both hit — yields push again and SOX fails — cut the rotation. If only one hits, then you can talk about a semi trend and META looking higher.