Nasdaq prints a high; don't misread Amazon
US Stocks · Options · News · Views
Nasdaq closed at a high. SPX did not pick a side
Tuesday's tape split. The Nasdaq Composite closed 27244, up 0.5%, after tagging 27289. That was a closing high. The Nasdaq 100 closed 30732 and traded 30770, also a closing high. QQQ closed 747, up 0.8%. The one-year high is about 749; the session high was 748. SPX closed 7765, basically unchanged. The Dow closed 51864, down 0.4%. The Russell 2000 closed 2890, up 0.5%. IWM closed 287. 288 is still out of reach.
Monday had all four indexes green, with tech carrying the tape. Tuesday left Nasdaq making the high while the S&P sat still and financials sold off first. Energy and financials were the weakest groups. XLE fell 1.1%. XLF fell 1.9%, closing 54.80 after a 54.55 low, with volume jumping. Tech can still lift the index. Financials took the legs out. SPX is still swinging in that 7000 to 8500 range. The upper end opened a bit last week. That is not a side chosen. New highs with a top-heavy interior is still the structure.
Oil moved lower. The hike path is not clean
Crude kept falling. WTI settled 94.59, extending the slide. U.S.-Iran contacts were described as productive, with another meeting still to be scheduled. No deal is done. Oil already priced in a slice of that optimism. Cheaper crude eased inflation expectations a notch. The 10-year yield closed 4.97, little changed. Money markets still leave room for another 25 basis points in 2026.
Boston Fed's Collins backed last week's hike, arguing a somewhat tighter rate helps inflation return to target. Richmond's Barkin said last week's hike helps restore price stability and left further hikes as a wait-and-see. The path is not clean. Soft energy is not the same as the hike trade being over.
Models got cheaper. Texas just jammed the power
The talk is whether to slow frontier AI. The product cadence did not slow. Monday, xAI shipped a new model. Tuesday around midday, Anthropic launched Opus 5.5. In the afternoon OpenAI launched GPT-6 Sol and GPT-6 Luna and cut token prices in half. Lighter models and a price war to take share is the live tape. Rivals are not pausing. Supervisors are not forcing a pause. Whoever blinks first loses.
Data-center construction is slowing. Texas tightened the freeze: environmental regulators pause all new data-center permits until the grid audit is done. Audits can run fast or slow. This one can slip toward year-end. Texas sits at the top of planned capacity. That pause pushes the compute-build calendar out. From April to June, about 45 projects totaling $68 billion of data centers were blocked or delayed by local opposition. Power, water bills, electricity bills, and the November midterms all point the same way: states will not loosen approvals from here.
Demand is still there. Permitting is slower. Cheaper models are a workaround when watts cannot keep up, not proof that demand died. What matters next is whether software and lighter models extract more tokens per kilowatt-hour, not whether the old cabinet-fill plan still holds.
Meta and Amazon are fighting over the door. Shopify got paid
META's Muse tried to check out on Amazon. Amazon blocked the door. Users now get a popup that the agent violates the conditions of use. Checkout and discovery are being deepened with SHOP. The two giants collide. Shopify puts on weight.
Do not let that fight knock AMZN off its tape. Last Friday it broke its downtrend on the same day as the semiconductor ETF. Monday closed 258. Tuesday pulled back to 255, high 259, low 254, down 1.3%. That is a breakout pullback. 244 to 256 is the strong support band; 217 to 256 still sits underneath. As long as price stays above that broken downtrend, quiet down days remain entries, not shorts. Around 250 maps to this year's forward median. Next year's median sits near 282, with 323 at the upper end. The structure is intact. What is missing is a catalyst. Sparring with META can slow the grind. A run at the prior high is still the high-probability path. Consumer agents will not be a monopoly. Google, Microsoft, and Amazon will each have a door. The question is whose story gets loud first.
SHOP rose 7.3% Monday and closed 138. Tuesday added 7.1%: open 144, high 150, close 148. Two sessions are almost 15%. The 123 floor never broke. It went straight up. The stock is not cheap. Growth names get a premium for a few months a year. That is volatility, not a permanent hall pass. A new-high trend needs 140 to 158 to clear. Price is still inside 123 to 158. 147 is already in the supply zone, not a market-order chase. Either wait for a hold at support, or wait for the right side of 158. The worst trade is treating the news spike at 147 as a breakout.
META itself closed 737, down 0.6%, after a 757 high. Monday's move through 730 is still the structure. The short-term repair is unchanged: extend then pull back, or pull back first then grind higher. Hold above 638 and treat it as a right-side tape. 690 to 730 is turning into the new support band. Connect runs the 23rd through the 24th. Usage numbers are still not on the table.
SOXX is at 573. 577 is not done
The Philadelphia Semiconductor Index rose 2.1%. SOXX opened 554, high 574, closed 573, up 2.4%. Last Friday it left the 520 box on volume and closed 533. Monday closed 559. Three sessions have already walked the 550 to 577 supply zone to its lower edge. 577 is a few points away. The 655 cycle high is still about 14% overhead. The ETF just touched overbought, and resistance has not cleared. Chasing 573 on the close is a poor payoff. The comfortable pullback is Friday's body, 522 to 533. A strong pullback only gives back Monday's body, 545 to 559. If 577 goes, FOMO toward the old high will speed up. Until it goes, it has not gone.
The internals are top-heavy too. AMD closed 624, up 1.4%, high 625, tagging the prior high. NVDA closed 229, up 0.7%. TSM closed 452, up 1.5%. Those three sit closer to their own highs than the index does. Plenty of members are still sitting, some a long way from 52-week highs. The money is in the leaders. If you stay in semis, stay with the crowded strength. Laggards do not have flow evidence of a catch-up bid. Cheap is not the same as due to squeeze.
Financials dropped 2%. Watch 54 first
XLF fell nearly 2% in one session, which is unusual for a low-beta group. The low was 54.55. It closed 54.80. Strong support starts at 52.6 to 54.1, and the tape just tagged the top of that band. There is not much of a shelf overhead.
One read is that AI agents will strip friction out of financials, telecom, and travel booking. Comparison shopping, auto-renewed insurance, phone plans, airline tickets: those businesses used to live on hassle. An agent that compares, cancels, and rebooks weakens inertia. BKNG fell 2.6%. EXPE was basically flat. That story can explain a slice. It is not enough evidence to pin Tuesday's drop on one cause.
Slower data-center permits also blur the calendar for large AI-related IPOs, which cools the underwriting and trading-revenue outlook at the banks. GS fell 1.1%. MS fell 2.9%. JPM fell 3.4%. The harder layer is the fundamental tape: financials' expected EPS growth this year and next sits below the S&P, in the single-digit bucket. Technically the drop is getting close to support. Fundamentally the group is not a leadership sector. If you touch it, watch for a large IPO to land, and for underwriting exposure at Goldman and Morgan Stanley. Do not short the whole financials complex as if AI had already eaten it.