NOW Breaks Out, NVDA Still Favours a Slow Grind; Data-Centre Capex Faces a Speed Risk
US Stocks · Options · News · Views
Stocks finished Friday higher, led by the Nasdaq, although trading volume fell for a third straight session. Roughly two-thirds of SPX constituents advanced, and 9 of 11 sectors closed in the green. The market kept grinding higher because sellers never really showed up. The main catalyst was payrolls: weaker job growth and softer wages pushed the market-implied odds of a September hike from roughly 55% to around 40%–44%. Stocks, Treasuries, and precious metals all benefited from the dovish repricing, while the dollar weakened.
Payrolls weaken, and rate-hike odds move lower
July nonfarm payrolls fell by 23,000, well below expectations for an 80,000 increase. The previous two months were revised down by a combined 103,000. Government employment dropped by 53,000, leisure and hospitality lost 40,000 jobs, and private payrolls rose by only 30,000 versus 78,000 expected. The unemployment rate slipped to 4.1%, but that looks more consistent with the 0.1-point decline in labour-force participation to 61.4% than with a sudden improvement in labour demand.
Wage growth softened as well. Average hourly earnings rose 0.1% month over month and 3.2% year over year, both below expectations. A September hike may no longer be necessary, although the inflation data still has the final say. For now, markets have repriced the cost of money slightly lower.
Data centres: the bigger risk is how fast spending eventually falls
The AI and data-centre capex cycle is now nearly three years old, dating back to 2024. Hyperscale spending is projected to approach 3% of GDP between 2027 and 2029, compared with 0.3% in 2019 and roughly 1.4% in 2025.
History gives us some useful context. Telecom and fibre investment peaked at roughly 1.2% of GDP in 2000 before collapsing during the dot-com bust and a mild recession. Housing investment peaked near 6.6% in 2005. Today’s projected data-centre share sits between those two extremes: more than twice the telecom peak, but less than half the housing peak.
The speed of the buildout matters even more than its absolute size. A rise from 0.6% of GDP in 2023 to 3.1% in 2027 is a 2.5-point increase. From 1.4% in 2025 to 3.1% in 2027, the increase is 1.7 points in only two years, or roughly 0.85 points annually. Housing’s fastest stretch averaged about 0.5 points a year, while telecom managed roughly 0.15. AI capex is therefore ramping at close to twice the pace of housing’s hottest phase.
A gradual slowdown can be absorbed. A sudden pullback is much harder. Capex can keep climbing only if revenue eventually catches up. If monetization disappoints, spending will be cut. The market is still pricing boom conditions through 2027–2029, so a sharp retrenchment heading into the 2028 election year could be painful. For data centres, mega-cap tech, and semiconductors, I am watching not only the peak in spending, but also the slope on the way up and the speed of the eventual slowdown.
NOW: breakout above 120 puts 147–159 in view
NOW gapped open near 121, cleared the 120 right-side trigger on strong volume, and held near its highs throughout the session. Longer-term holders do not need to react to every move between 125 and 140. The first meaningful swing target is 147–159. Above that, the larger gap-fill zone sits around 180–193. Only a clean break through that area would open a path back toward the old high near 239. On a 30–40x valuation framework, my 2027 ceiling is roughly 200. Without another 20% or so of earnings upside, breaking decisively through 193 could take time.
From roughly 124, a stop below 110 risks about 14 points against roughly 23 points of upside to 147. But support extends below 110, so a break there would not automatically invalidate the broader trend. The cleaner fundamental accumulation zone remains 100–120. A pullback toward 110 would look more like a reset than a collapse.
NVDA and SOXX: trend intact, but slower is healthier
NVDA still looks technically clean. Another push higher on the daily chart would move it deeper into overbought territory, but overbought is a warning, not an automatic sell signal. Without bearish divergence and with the trend still strong, that condition can persist. The key breakout level is 216; holding above it keeps the structure pointed toward highs above 236. With index volume still thin, however, a rapid sprint higher would be more likely to create divergence and set up a technical shakeout. The weekly chart remains healthy, and monthly RSI has crossed higher. After such a long consolidation, trimming aggressively without a meaningful pullback creates a different risk: never getting a good entry back.
SOXX is still pressing around 580, while software ETF IGV has not matched the breakout pace of the SPX and Nasdaq. Semiconductor earnings remain solid. The debate is about the pace of growth, not an imminent collapse. I would reduce risk when the tape gets overheated and add it back when prices return toward fundamental value. Breakouts can take time; reflexively shorting strength is usually the worse habit.
UNH and SHOP: give UNH time; watch 158 on SHOP
UNH reached roughly 461 after its April rebound, then started drifting lower. Forward value is only modestly higher from here. Holding the 440–484 area and pushing toward 500 would likely require another roughly 10% increase in 2028 forward estimates to justify valuations in the mid-540s. Without a genuine acceleration in earnings, a sustained breakout at a richer multiple looks difficult. Long-term holders with entries around 347–357 or 317–342 can afford to give the thesis more time. This is not a melt-up stock, and the first rebound already ran hot.
SHOP, using a 60x growth multiple, has a 2027 forward valuation ceiling near 135 and still looks expensive. The key right-side trigger is 158. Only a move above that level would make new highs the base case. The post-earnings gap, high-level consolidation, and lighter volume still need confirmation. Near-term support sits around 123–128, with a stronger shelf at 136–141.
Payrolls eased the expected rate path, NOW turned that tailwind into a software breakout, and NVDA still looks healthier grinding higher than sprinting. The medium-term risk in data centres is not whether spending is still growing. It is how quickly that spending eventually comes back down.