The AI Spending Model Is Changing — and a Crack in the Semiconductor Faith
US Stocks · Options · News · Views
On the first trading day of July, the market stopped following first-half inertia. The Nasdaq fell 1.54%, semis rolled back over, and the Philadelphia Semiconductor Index dropped nearly 6% in a single session, dragging the back-and-forth action since mid-June back to the lows. Among the Magnificent Seven, every heavyweight except NVDA — which was dragged down by semis — actually rose, with META the standout, surging about 8%.
This is not ordinary mega-cap rotation. The single most important line in the AI trade is starting to shift: capex can no longer be judged by "more is better" — it has to be measured by whether each dollar deployed converts into higher-quality revenue and free cash flow. June ADP private payrolls printed 98K, below May and below expectations; Challenger layoffs fell to 45.8K, but tech remains the most concentrated area of cuts, with AI-driven restructuring still reshaping the labor mix. The macro signal isn't extreme, but it's enough to confirm that companies are prioritizing efficiency.
META's plan to package excess AI compute and sell it as a cloud service may not rewrite the fundamentals immediately, but the signal is sharp: if frontier-model spending doesn't pay off fast enough, the hyperscalers will pivot from piling up more compute to monetizing the compute they already have. The market gave META positive feedback — and at its core, that's a reward for financial discipline, not for more aggressive AI spending.
The strongest pricing logic for semis has been persistent compute scarcity — supply perpetually tight, prices and margins sustained at altitude. As long as that premise holds, GPUs, memory, optical modules and the data-center chain can all keep reaching for valuation off distant forward growth. But the META print forces a counter-question: if the hyperscalers already have surplus compute to sell, just how scarce is compute?
That question lands directly on semis and the new cloud providers. If high-performance compute rental rates keep falling, the compute-shortage narrative weakens, and the companies with the most aggressive 2027 revenue and profit expectations get discounted first. NBIS and CRWV — the new cloud names — sold off hard; MU, SNDK and other memory/hardware names that had been propped up on high expectations came under pressure too. The fact that nearly the entire Philly semi index fell tells you this isn't one or two stocks getting sold — it's a re-rating of the forward assumptions across the entire hardware chain.
This is not yet a "capex recession" call — it's a "capex growth may downshift" call. The distinction matters: 2026 spend can still hold at high levels, and 2027 doesn't have to shrink outright. But if expectations were for spend to go from $100B to $120B and it only reaches $110B, the market will still cut the multiples of companies priced for the higher growth rate. NVDA also takes a confidence hit, but its sheer size, market share and the fact that it hasn't been swept up in the small-cap speculation mean its price elasticity and downside damage may both be smaller than the most aggressively bid niche hardware names.
The shakier the semi side gets, the more pronounced the seesaw into software. IGV rose about 3% today on expanded volume — a sign that software isn't just passive drift, but is starting to attract active buyers. That said, this isn't yet a full-blown breakout. The key line sits around 94.5; price tagged it intraday but couldn't hold above it into the close, so for now treat it only as an enhanced bounce.
What's actually worth watching is the names that rally on volume — not a blanket read on software. ORCL also expanded volume, but in the down direction, because it's more tightly bound to data-center buildout, unfulfilled contract obligations and balance-sheet stress. If cloud and compute-demand expectations decelerate, ORCL — which pre-funded infrastructure expansion with debt — is exactly the kind of name the market will question hardest.
The volume-up leaders worth tracking are PLTR, APP and NOW. PLTR is coiling in the 121–130 zone; a breakout opens a path toward 151. APP has been lifting on light volume for three straight sessions; above 573 the next stops are 617 and then the prior high at 660. NOW's volume expansion is gentler — a decisive break above 120, once the bottom is in, opens a path toward 143. For short-term trading, the names without volume are mostly just being carried by quant and index flows; the initiative isn't there, and they belong lower on the priority list.
META has lifted back toward 611 in the near term, with 559 now the line of defense. As long as 559 isn't decisively broken, the prior leg down can be treated as temporarily arrested; but real overhead remains, and the bounce needs the next earnings print to deliver a clearer capex and revenue signal before it can become a more durable uptrend structure.
MSFT is still inside its range. Below 375 has offered dip-buying opportunities; overhead, 392–435 is the key resistance band, and only a break above 435 opens the right-side setup. GOOGL and AMZN are both consolidating on contracting volume — few fresh signals for now. TSLA remains a high-volatility structure, with 425 marking near-term resistance and 465 the level that would matter for a real breakout. These stocks go up fast and come down fast — don't mistake short-term elasticity for long-term certainty.
The most important thing ahead is Q2 earnings season. If the four mega-caps simultaneously hint at slower capex growth in 2027, the heavyweights may keep getting positive feedback on free-cash-flow repair, while the semi names with the richest forward expectations, fastest runs and smallest bases get de-rated first. In terms of trading, there's no rush to pronounce the death of the AI theme — but the narrative is shifting from "who spends the most" to "who spends the most efficiently." Semis need consecutive down days to break the positive feedback loop created by leveraged ETFs; a fast snap-back the next session still leaves it inside the chop. The real direction comes only when continuity and earnings signals confirm together.
Press and hold the QR code to follow the public account for daily U.S. equity insights and commentary.
Disclaimer: This article reflects personal market observation and trading review only. It is not investment advice. Markets carry risk — trade with caution.