On July 28, 2026, the US AI computing sector faced a new shock. The Nasdaq fell 1.2% that day, with AI chip giants like NVIDIA, AMD, and Broadcom all under pressure, dropping over 3%. Market sentiment shifted from early-month euphoria to caution, driven by a "double whammy" of lowered earnings expectations and persistently high capex.
Earnings season stress test: Early signs of AI growth slowdown
This week marks the peak of US Q2 earnings season. By July 27, over 60% of S&P 500 companies reported results. AI computing-related firms showed clear divergence: downstream players like Microsoft and Google saw cloud AI revenue growth slow to under 15% QoQ, while infrastructure suppliers like NVIDIA posted 45% YoY data center revenue growth, missing the >50% expected.
More concerning, several cloud giants lowered their H2 2026 AI business growth guidance in earnings calls. Amazon AWS CFO stated, "Enterprise customers are becoming more pragmatic in AI deployment, shifting from large-scale experiments to selective implementation, leading to slower near-term order pace." This directly dampened beliefs in explosive AI computing demand.
Capex 'arms race' shows no signs of slowing
Despite slowing revenue growth, tech giants' capex on computing infrastructure did not shrink. Microsoft, Google, and Amazon combined spent $62 billion in Q2, up 38% YoY, with over 70% on AI data centers, GPU clusters, and network gear. Meta raised its full-year capex guidance to $45 billion, mainly for next-gen AI chips.
This pattern of "slowing revenue but accelerating spending" has raised Wall Street concerns about returns. Morgan Stanley analysts noted in a recent report: "If AI applications fail to achieve large-scale commercialization within 2-3 years, current investment levels will cause significant capital efficiency decline, and some firms may face asset impairment risks." The firm downgraded AI computing hardware from 'overweight' to 'equal-weight'.
Market tug-of-war: Valuation reversion or short-term correction?
From a valuation perspective, the AI computing sector's median P/E remains at 35x, well above the S&P 500 average of 21x. Although YTD gains have narrowed, absolute valuations remain undigested. Fund flows show AI-themed ETFs saw net redemptions of about $4.5 billion in the last week of July, the largest weekly outflow in nearly a year.
However, bulls remain. Optimists view the current slowdown as a natural transition from 'infrastructure boom' to 'application penetration'. NVIDIA CEO Jensen Huang stressed in a recent speech, "Computing demand will see a second wave of exponential growth in 2027, as next-gen model training requires 10x current compute." Chipmaker TSMC also expects continued shortages for advanced nodes at 3nm and below.
Additionally, President Joe Biden's signing of the AI Infrastructure Act on July 26 injected policy certainty. The act plans $50 billion over five years for national AI computing centers and domestic chip manufacturing, seen as crucial support for US AI long-term competitiveness.
Sector divergence: Who is the true 'safe haven'?
Amid the broader pullback, AI computing sector shows clear divergence. Cloud giants with in-house chip design (e.g., Google TPU, Amazon Trainium) held up relatively better, while pure hardware suppliers (NVIDIA, AMD) fell more. Institutions generally believe vertically integrated firms are more resilient to demand fluctuations.
Another divergence theme is 'earnings visibility'. Companies with long-term contracts and recurring revenue (software subscriptions, cloud services) command market premiums. For instance, AI cloud providers CrowdStrike and Palantir rose 5% and 3% post-earnings, as over 80% of their revenue comes from recurring contracts.
Outlook: Three key variables to watch
Looking ahead 1-2 months, the market will focus on these variables:
- Application breakthrough signals: Whether Apple's developer conference in early Aug and Google Cloud Next in Sep deliver killer AI apps will be key catalysts for demand expectations.
- Fed monetary policy: If the July 30 FOMC meeting delivers a rate cut of less than 25bps, it may further dampen risk appetite for high-valuation tech stocks.
- US-China tech rivalry progress: Rumors that the US Department of Commerce will expand AI chip export controls to China, if enacted, would directly hit NVIDIA and AMD's China revenue.
Overall, the US AI computing sector is in a transition from 'storytelling' to 'performance watching'. Short-term volatility is inevitable, but long-term industry trends remain intact. Investors should beware of valuation bubbles while focusing on core names with real earnings delivery. For A-share equivalent plays (optical modules, servers, etc.), watch for sentiment spillover from US stocks and manage positions prudently.
