On Thursday, U.S. equities deepened their losses into the close, ending at session lows. The S&P 500 finished down roughly 0.60%, the Nasdaq 100 weakened in step, and the Philadelphia Semiconductor Index tumbled nearly 5% on the day—becoming the dominant short-side driver in the tape.
What you’re watching isn’t a systemic market meltdown; it’s a classic high-level position redistribution. The AI trade that dominated most of the year was hit with concentrated profit-taking, as capital rotated swiftly into defensive sectors. The extreme bullish expectations that had been priced to perfection are now being soberly corrected. The market is starting to reassess the delivery cadence of the AI supply chain and whether the valuations still make sense.
Market Overview: Semiconductors Lead the Selloff While Defensive Sectors Rise
The day’s structural divergence was stark:
-
Growth names under broad pressure: The chip sector was the epicenter of the damage—MICRON, NVIDIA, and BROADCOM all weakened significantly, and SANDISK is on track for its worst weekly performance since last April. The entire semiconductor group has now erased all the gains of the past two months; at its peak, the Philadelphia Semiconductor Index had surged 140%–150% year-over-year, and with the pullback the year-to-date gain has narrowed to 67%, taking the index back to levels seen one-and-a-half to two months ago. AMAZON saw a sharp, news‑free plunge into the close, adding to growth‑stock volatility.
-
Defensive pockets outperformed: The consumer staples sector posted its best two‑day rally in more than a month; financials and healthcare also beat the broader market, serving as a haven for capital. Major bank earnings were broadly solid, and both ABBOTT LABS and UNITEDHEALTH raised their guidance. UNITEDHEALTH has climbed 60% since late March, making it the second‑best non‑tech component in the S&P 500 over that span.
-
Fixed income and commodities: The 10‑year U.S. Treasury yield rose 2 basis points—bonds did not act as a hedge on the day and instead saw mild selling. Brent crude, still elevated compared with the prior few weeks, edged lower by 0.80%.
The core worry is clear: the high valuations across the AI supply chain rest on the assumption of perfect execution and consistently better‑than‑expected earnings, leaving virtually no margin for error. The moment the market starts to question downstream customers’ return on investment, or begins to fear a slowdown in the pace of capex, the crowded positioning at the top begins to crack.
The AI Market Cooling-Off: Expectations Fulfilled
Over the past 48 hours, ASML and TSMC have released their financial results and guidance. The numbers themselves are not bad. TSMC's sales outlook even exceeded market expectations. Yet the market’s reaction was the exact opposite—capital chose to sell on the news and lock in gains.
The logic behind this is straightforward: as downstream customers continue burning cash on massive expansion, the base of capital expenditure keeps growing while the return on investment declines marginally. The cost-effectiveness of pure AI themes gets reassessed. Investors are shifting toward a more cautious positioning strategy: prioritize companies with durable long-term fundamentals and avoid stocks that are fully betting on an unimpeded AI business, where the share price has already priced in perfection.
Semiconductors: A Pullback Amid Rotation — Digestion, Not Reversal
According to Bernstein's semiconductor industry analysis, this pullback is essentially a healthy digestion of the earlier excessive rally, and the long-term industry logic remains intact:
-
The supply-chain rotation pattern persists: Over the past 6-12 months, the AI rally has rotated sequentially along supply-chain bottlenecks—from GPUs to memory chips, then to semiconductor equipment, optical modules, networking gear, power components, CPUs, and recently spreading to discrete devices. AI has grown so large in scale that it can now pull the entire supply chain forward; only the pacing of upswings differs across segments.
-
Core names show greater resilience: This also lies at the heart of the most common question in the market right now—why computing-power pillars and foundational supply-chain plays like NVIDIA and Broadcom are trading sideways while capital crowds into niche bottleneck trades. Actual data shows that core names are relatively resilient during pullbacks: some memory stocks have plunged sharply from their peaks, and although NVIDIA has corrected, the decline is far smaller than that of other high-beta names. Once sentiment stabilizes, core assets tend to recover first.
-
Overseas volatility amplifies sentiment swings: In the Korean market, SAMSUNG and HYNEK'S together account for more than 60% of the broad index, and a large number of retail investors are leveraged with borrowed money, making the turbulence far more violent during corrections—this has, to a certain extent, pushed up volatility for the global semiconductor sector. In contrast, the U.S. market has lower position concentration, so the impact is relatively contained. Still, single-day individual stock swings of 6%-7% are becoming normalized, a phenomenon worth watching.
The Next Layer of AI Investment: From Standalone Hardware to Full Supply Chain Opportunity
Morgan Stanley Investment Management's view emphasizes the cross-industry, pervasive nature of AI, arguing that this industrial trend cannot be assessed through a single lens:
-
AI has already become a cross-asset, cross-industry foundational opportunity, much like the internet, and cannot be isolated in investment analysis. Supply bottlenecks at the infrastructure layer keep iterating, with new shortage areas emerging every few months; the software layer is currently priced by the market for collective pessimism, but companies possessing data, industry expertise, and distribution advantages are well-positioned to capture a windfall in the agent economy—mispriced opportunities lurk here.
-
The long-term trajectory of computing demand is highly certain: from reactive AI (Q&A-style) to generative AI, and onward to future autonomous agents, computing demand is set to leap by a thousandfold or even millionfold. Capital expenditure and commercialization continue to materialize, as already validated by the earnings of hyperscale cloud providers. Enterprises may delay their investments, but they will not cancel them.
-
Within specific segments, memory, as the foundational layer of the token economy, has a clear supply-demand picture over the next 6-8 quarters and is still supported by solid fundamentals; the software sector, after a significant pullback, now presents compelling allocation value.
The market's core confusion over the software sector lies in the restructuring of valuation logic: in the past, enterprise software could charge per user. In the agent era, a single user may correspond to multiple agents, and it remains unclear how software companies’ revenue models will evolve. Coupled with high token costs and the initially opaque profitability models, the market has broadly suppressed software sector valuations. However, quality companies in cybersecurity, industrial software, and healthcare software—leveraging their data moats, industry know-how, and distribution advantages—are likely to emerge as winners, making the current pullback a window for positioning.
Earnings Season Spotlight: Netflix at the Crossroads of Valuation and Sentiment
One of the worst-performing S&P 500 stocks lately, Netflix is drawing intense scrutiny the night before its earnings release. The drawdown from its peak is approaching the plunge triggered by the unexpected subscriber loss in 2022 — the company’s first subscriber decline in a decade.
Unlike 2022, the company no longer discloses subscriber figures. The market’s attention is now concentrated on three core questions: whether user engagement is topping out, whether margins will contract, and whether the company will pursue M&A — and the motivation behind it. The market once viewed M&A as merely “nice to have” for Netflix; the prevailing view has now shifted to “it may become a must-have.”
On the valuation side, Netflix’s stock has fallen roughly 42% from its all-time high in June 2025, with multiples retreating to historically low ranges. After the sharp sell-off that followed the last earnings report, this release is seen as a critical sentiment inflection point. Bloomberg consensus estimates call for revenue to grow about 14% year-over-year to $12.6B, with GAAP earnings up about 10%. Bloomberg will also host noted bull Ross Gerber, former Netflix executive and television producer Evan Shapiro, for an in-depth post-earnings discussion after the close.
WEDBUSH SECURITIES maintains a Buy rating and a high price target on the stock. Their core thesis centers on these points:
-
Engagement data matters more than earnings: The market has largely priced in earnings expectations; the user engagement data accompanying the release is the real driver for the stock. Guidance, margin trajectory, and the contribution from the advertising business are also key areas to watch.
-
The ad business is significantly undervalued: Declining per-user engagement is a given fact — the company needs to keep ramping content investment just to keep total engagement stable. The tightening of ad-tier controls also has some impact on engagement, but the market substantially underestimates the growth potential of the advertising business.
-
The company’s ad load is the lowest in the industry, so there is plenty of room to increase it.
-
Ad targeting capabilities are currently only at a moderate level, leaving significant room for future optimization.
-
Advertisers are not leaving — demand for Netflix’s ad inventory remains solid.
-
Structurally, CPMs for general content are declining, but CPMs for premium ad slots, such as sports events, continue to rise, and the share of high-value content keeps growing. Monetizing “high-efficiency engagement hours” is the core logic.
The company employs a differentiated placement strategy — fewer ads during hit content, regular placements during series — while also using live sports like the MLB Home Run Derby to unlock high-end ad value. The effectiveness of this long-term strategy will unfold gradually. The market’s earlier debate over the ad strategy during the MLB All-Star Game does not alter the overall long-term logic.
-
-
Multiple growth levers remain untapped: The recurring “second-quarter viewership softness” in the U.S. market largely stems from the company spending less on promoting second seasons compared to first seasons. This is an exceptionally easy lever to pull, and the company is expected to increase investment in response to market attention. Additionally, the expanding portfolio of gaming, podcasts, lifestyle channels, and live TV will enrich the user experience and become a second growth curve for both engagement and monetization.
-
The risk threshold is clear: Only a cliff-like drop in user engagement would reverse their positive stance on the company. Under the current trend of a gradual slowdown in engagement, steadily improving margins will gradually digest market concerns.
Strategy Outlook: Mildly Bullish with Short-Term Disturbances
The core tension in the market right now is extremely clear: the earnings fundamentals remain strong, and AI-driven capex continues to be revised upward, which supports the earnings resilience of the tech sector. However, the macro environment is becoming less friendly – the Fed leadership transition, recurring Middle East tensions, and already elevated valuations are all capping the upside for the indices.
Regarding the market’s focus on the capex problem at the tech giants: their revenues are still growing at a high rate of around 30%, but massive capital expenditure continues to eat into free cash flow. Going forward, it’s expected that companies will funnel 90% of operating cash flow into capex and may even need to tap the credit market for financing. This capital-intensive nature weighs on valuation multiples. That said, current valuations already partially reflect this risk – we’re in a digestion phase. It will take years for the AI narrative to fully materialize and for a clear margin profile to emerge, but the earnings resilience itself is certain.
When it comes to the market rhythm ahead, the historical pattern around midterm elections is worth noting: historically, U.S. equity performance in midterm election years tends to be weaker than in non-election years, with August through September often being the worst stretch of the year. But once the election uncertainty is cleared, the market typically stages a rebound, and the tech sector is usually the biggest beneficiary – even when it sits at the epicenter of uncertainty right now, it tends to be the first to repair after the vote.
On balance, Thursday’s pullback looks more like a period of cooling off after the AI-fueled sprint, rather than a trend reversal. The fundamental earnings support is still in place, and the long-term industrial logic hasn’t broken. But overstretched valuations, overly optimistic expectations, and layered macro and policy uncertainties have made market sentiment fragile. The upcoming earnings season will be the critical proving ground: whether it’s the pace of semiconductor capex or the monetization efficiency of internet platforms, both will determine the depth of this correction and the market’s next direction.