The Rotation Tell: Chip Rout and $100 Oil Split the Market in Two

The Friday snapshot: a market splitting in two

The S&P 500 eked out a 0.1% gain on Friday to close at $738.93 via the SPY ETF, but the headline masked a sharp divergence underneath^[1]. The Nasdaq 100 (QQQ) fell 1.1% to $684.23 as chip stocks and AI-exposed megacaps sold off, while the Dow Jones (DIA) rose 0.5% to $518.76, lifted by defense and industrial names^[1]. The Technology Select Sector SPDR (XLK) dropped 1.4%, the worst-performing major sector ETF, while Financials (XLF) gained 0.9% and Energy (XLE) added 0.4%^[1].

For the week, the S&P 500 finished down 0.6% — its second straight weekly decline, the longest losing streak since late March^[2]. The Nasdaq shed 2.1%^[3], while the Dow added roughly 108 points^[3]. The divergence is not noise. It is the market repricing two competing stories at once: one about whether AI infrastructure spending will ever pay for itself, and another about whether $100 oil and a rearming Pentagon shift the profit center of gravity away from Silicon Valley.

AI capex anxiety: when record revenue isn’t enough

The catalyst for the tech rout was the earnings season’s first real inflection point. Tesla reported Q2 revenue of $28.2 billion, up 26% year over year, on record quarterly deliveries of 480,126 vehicles^[4]. By the old playbook, that is a blowout. But operating margins collapsed to 1.4%, free cash flow turned negative, and management signaled a further ramp in spending on AI, robotaxi development, and robotics^[4]. The stock crashed roughly 14.5% on Thursday^[4] and slid another 2.1% on Friday to close at $313.03^[5], wiping out approximately $140 billion in market value over two sessions^[4].

Alphabet told a parallel story — cloud revenue surged 82% year over year, but the company also flagged higher AI capex^[4]. The message investors heard: even when the AI investment is generating visible revenue, the spending trajectory is outrunning the return. With 27% of S&P 500 companies having reported, 86% have beaten EPS estimates and 80% have exceeded revenue expectations^[3] — yet the market is punishing the spenders.

The semiconductor sector bore the brunt. The PHLX Semiconductor Index (SOX) fell roughly 3% on Friday^[6], with memory-chip names hit hardest: SanDisk dropped 10.3%^[7] and Micron and Intel also declined sharply^[6]. AMD fell 3.3% to $521.95^[5], and even Nvidia slipped 0.9% to $206.84^[5]. Oracle tumbled to a 52-week low, closing at $114.99, down 4.2% on the day^[5][6]. The logic connecting these moves: if the biggest AI buyers are signaling that spending will keep climbing before returns materialize, the chip suppliers feeding that capex cycle become a proxy for the same anxiety.

Defense and energy: the bid rotates

If capital fled tech, it found a home in the sectors most directly tied to geopolitical escalation. Lockheed Martin surged 13.3% to $582.65 after a beat-and-raise quarter, with sales up 11% to $20.1 billion and a record backlog of $230 billion^[5][8]. RTX gained 9.3% to $212.79 after its own beat-and-raise, with sales up 14% to $24.7 billion and a raised 2026 outlook^[5][8]. Both companies cited the Pentagon’s push to restock weapons amid the wave of global conflicts^[8].

Energy joined the bid. Brent crude topped $100 a barrel for the first time since May after Houthi attacks on Saudi ships in the Red Sea^[9], and President Trump threatened Iran with “major military punishment”^[9]. Oil eased on Friday amid hopes that Pakistan could mediate talks^[3], but energy stocks held their ground. Schlumberger (SLB) jumped 11% to $52.42^[5], while ExxonMobil and Chevron were essentially flat on the day at $156.94 and $194.79, respectively^[5] — suggesting the energy bid is being driven more by oilfield services and exploration than by integrated majors at current prices.

The macro backdrop: a 2006 echo

The macro snapshot provides the frame for next week’s Fed decision. Headline CPI is running at 3.46% year over year, well above the 2% target^[10]. The Fed funds rate sits at 3.63%^[10], and the 10-year Treasury yield has climbed to 4.71% — its highest level of Trump’s second term^[2][11]. The yield curve has steepened to +36 basis points^[10], a shift from inversion that historically marks the late-cycle transition. Unemployment is 4.2%^[10], industrial production is growing at just 1.14% year over year^[10], and consumer sentiment has cratered to 44.8 — down 14% year over year^[10].

The FRED analog search flags mid-2006 as the closest historical match^[10], when unemployment was 4.6–4.7%, CPI was around 3.9–4.2%, and the Fed was holding at 5.25% after a long hiking cycle. That period did not immediately lead to recession — the economy muddled through for another year before the cracks appeared in 2007. The parallel is imperfect: in 2006, the Fed was holding above neutral; today the Fed funds rate at 3.63% is already below the estimated neutral rate, and $100 oil is pushing inflation back up rather than cooling down.

The FOMC meets July 29–30. The consensus expects a hold^[11], but rising oil has shifted the odds: CME’s FedWatch tool now puts the probability of a 25-basis-point hike at 46.5%^[11]. The base case is still a pause, but it is no longer the overwhelming consensus it was a month ago.

What would have to be true

There are two ways to read this rotation, and both have honest supporters.

The bear case says the AI capex cycle has hit its “show me” moment. Tesla’s negative free cash flow and collapsing margins are the canary: the largest AI spenders are burning cash faster than returns are materializing, and the market is repricing the entire supply chain — from chipmakers to data-center landlords — for a slower payback. If oil stays above $100 and the Fed is forced to hike, the combination of higher rates and cost-push inflation would compress margins further, and the consumer-sentiment reading of 44.8 suggests the demand side is already cracking.

The bull case says this is a healthy rotation, not a top. Earnings season is running at an 86% beat rate^[3], real GDP is growing at 2.66%^[10], and the defense and energy bids reflect real, contracted demand — Lockheed’s $230 billion backlog is not a sentiment indicator, it is a signed order book. The tech sell-off is concentrated in names with company-specific execution problems (Tesla’s robotaxi spending, Oracle’s cloud transition stumbles) rather than broad-based deterioration. The VIX at 18.7^[10] and HY credit spreads at 2.77%^[10] — both well within normal ranges — confirm that the system is not pricing systemic stress. The 2006 analog that the FRED model flags as most similar did not tip into recession for over a year.

For the bear case to be right, AI spending returns need to disappoint systematically across multiple capex-heavy names, not just Tesla. For the bull case to be right, the Fed needs to hold, oil needs to stabilize, and the earnings beat rate needs to hold above 80% as the season continues.

What to watch next

FN2 Research provides financial research and education, not personalized investment advice.