NOW Breaks Out, NVDA Prefers a Slow Grind; Data-Center Risk Is a Fast Retreat
US Stocks · Options · News · Views
Friday closed higher, with the Nasdaq leading, but volume shrank for a third session. About two-thirds of SPX members rose, and 9 of 11 sectors advanced. The tape crawled higher on reluctant selling. The real driver was payrolls: soft jobs and softer wages pulled September hike odds from about 55% to 40%–44%. Stocks, Treasuries, and precious metals caught the dovish repricing; the dollar softened.
Payrolls: Cold Jobs, Lower Unemployment, Softer Hike Odds
July nonfarm payrolls fell 23,000 versus an expected +80,000. The prior two months were revised down by a net 103,000. Government jobs fell 53,000, leisure and hospitality fell 40,000, and private payrolls rose only 30,000 versus 78,000 expected. Unemployment eased to 4.1%, more consistent with a 0.1-point drop in participation to 61.4% than with sudden labor strength.
Wages were softer too: average hourly earnings +0.1% month over month and +3.2% year over year, both below forecast. September may not need a hike, but inflation prints still matter. For now, the market has marked the price of money one notch easier.
Data Centers: Size Is Large; Speed of Retreat Is the Risk
The AI / data-center capex wave is nearly three years old since 2024. Hyperscale spending is projected near 3% of GDP in 2027–2029, versus 0.3% in 2019 and about 1.4% in 2025.
History helps size it. Telecom / fiber peaked near 1.2% of GDP in 2000 before collapsing into a mild recession and the internet bust. Housing investment peaked near 6.6% in 2005. Today’s data-center share sits between those poles — more than twice the telecom peak, less than half the housing peak.
Speed matters more. A rise from 0.6% in 2023 to 3.1% in 2027 is about 2.5 points; from 1.4% in 2025 to 3.1% in 2027 is 1.7 points in two years, or roughly 0.85 points a year. Housing’s fastest stretch averaged about 0.5; telecom about 0.15. AI capex is climbing near twice as fast as housing’s hottest phase.
A slow decelerations can be digested; a sharp retreat cannot. Capex can keep rising only if revenue catches up. If revenue disappoints, spending gets cut. The market still prices 2027–2029 boom conditions — a cut into the 2028 election year would be ugly. For data centers, mega-cap tech, and semis, watch the peak, the boom, the fade, and the slope — not just whether growth exists.
NOW: Volume Break above 120; Next Band 147–159
NOW gap-opened near 121, broke the 120 right-side trigger on volume, and held highs all day. Near-term 125–140 can be ignored by longer holders. The first swing observation band is 147–159. The larger fill zone is 180–193; only a clean pierce opens the path toward the old high near 239. On a rich 30–40x frame, the 2027 ceiling is about 200 — without roughly 20% more earnings lift, 193 is hard to smash quickly.
From ~124, a stop under 110 is about 14 points versus ~23 points up to 147, but support continues below 110, so a break there is not automatic trend failure. The cleaner fundamental buy zone remains 100–120; a dip toward 110 looks more like a reset than a collapse.
NVDA and SOXX: Intact, but Prefer Slow
NVDA looks clean technically; another push on the daily chart edges into overbought. Overbought warns; it does not force a sell. With no divergence and a strong trend, that condition can last. The break level is 216 — only then does structure argue for highs above 236. With thin index volume, a sprint higher is more likely to print a divergence and force a technical washout. Weekly is fine; monthly RSI has crossed up. After a long consolidation, chopping the position without a deep pullback is a good way to never get back in.
SOXX still presses ~580, and software IGV has not confirmed the SPX / Nasdaq breakout pace. Semis have real earnings under them — the debate is growth speed, not disorderly collapse. Take risk down when the tape runs too hot; buy back when price returns to fundamental value. Breakouts may take time; reflexive shorts are the worse habit.
UNH and SHOP: Time versus 158
UNH tagged about 461 after the April rebound, then drifted lower. Forward value is only modestly higher; holding 440–484 and pushing toward 500 likely needs another ~10% lift in 2028 forwards to support the mid-540s. Without a real earnings acceleration, a sustained rich breakout is a stretch. Long-term holders in 347–357 or 317–342 can give it time — this is not a melt-up name, and the first rebound already ran hot.
SHOP on a 60x growth benchmark has a 2027 forward ceiling near 135 and still looks rich. The right-side gate is 158; only above that do new highs become the base case. Post-earnings gap, high consolidation, and lighter volume still need confirmation. Near support sits around 123–128, with a firmer shelf at 136–141.
Payrolls eased the rate path, NOW converted that into a software breakout, and NVDA still wants a slow grind. The medium-term risk in data centers is not whether spending exists — it is whether the retreat arrives too fast.