Semis Bear, Oil Surge: A Two-Front Week for Markets

Semis are the cleanest tell in this week’s tape. The Philadelphia Semiconductor Index entered a bear market — down more than 20% from its June 21 peak of 14,655 — erasing roughly $2.1 trillion in sector market value{{cite:67c68857ad8d}}. The SMH ETF closed Friday at $556.53, off 2.18% on the session{{cite:fd8a8b69bdb9}}. Meanwhile, energy was the only sector ETF to close positive, with XLE finishing at $57.68, up 1.16%{{cite:fd8a8b69bdb9}}. The contrast captures the week: capital rotating away from the AI infrastructure thesis and toward the geopolitical risk trade, all against a backdrop of cooler inflation and a new Federal Reserve chairman finding his footing.

The S&P 500 (SPY) closed at $743.29, down about 1% on the day and posting its first losing week in three{{cite:fd8a8b69bdb9}}{{cite:f7e5d85efbe9}}. The Nasdaq Composite fell 2.9% for the week{{cite:f7e5d85efbe9}}. The Dow (DIA) held comparatively steady at $520.81, down 0.77%{{cite:fd8a8b69bdb9}}, while small caps (IWM) showed relative resilience at $294.04, off just 0.52%{{cite:fd8a8b69bdb9}}.

The Semiconductor Bear: Pricing Power, Not Earnings

The surface narrative is “AI hype is over.” That is too lazy.

The trigger was Broadcom. On June 4, the company reported $22.19 billion in revenue and beat earnings estimates, but revenue fell 7% quarter-over-quarter and management did not raise its 2026 AI semiconductor sales forecast{{cite:67c68857ad8d}}. Broadcom sank 12% that day; the contagion spread to Micron, Intel, and AMD within a week.

The mechanism was hedge-fund positioning, not institutional repricing. Four consecutive weeks of net hedge-fund selling flowed through chip stocks through early July{{cite:67c68857ad8d}}. After the SOX index’s near-90% gain in Q2 — its best quarter on record — funds that chased the rally late were forced to exit. Order-flow data shows net outflows across every size bucket for NVIDIA: block orders, large orders, medium orders, and retail{{cite:67c68857ad8d}}.

But the structural signal beneath both layers is pricing power. NVIDIA’s B200 AI chip rental price on cloud compute platforms fell from $6.11 per GPU-hour on May 30 to $4.22 by June 21 — a 31% drop in roughly three weeks{{cite:67c68857ad8d}}. Rental rates are the marginal price signal for AI compute demand. A decline that sharp does not mean AI demand is dead; it means the pricing power that justified the sector’s 300% run-up is eroding as compute supply catches up.

The fundamentals have not collapsed. NVIDIA still grows revenue at roughly 71% year-over-year with a 74% gross margin and a 47% free cash flow margin{{cite:67c68857ad8d}}. The selloff is about whether those margins are sustainable when supply scales faster than demand — and the rental market is where that question gets answered in real time.

Additional pressure arrived from China. Moonshot AI, the startup behind last year’s disruptive DeepSeek model, unveiled its “Kimi K3” open-source model, which reportedly delivers performance approaching leading US models at far lower cost{{cite:dd6b0d644f16}}. The announcement rattled US chip and tech stocks, adding to the narrative that the AI compute stack is commoditizing faster than hyperscaler capex budgets can sustain premium pricing.

The Oil Front: Iran and the Strait of Hormuz

A short-lived US-Iran ceasefire collapsed in mid-July, plunging the Strait of Hormuz back into crisis{{cite:cfe597159a8d}}. Brent crude surged more than 4% on Monday as Washington reinstated a naval blockade of Iranian ports and both sides resumed military strikes{{cite:cfe597159a8d}}. At least nine ships have been attacked since July 6, and a maritime risk CEO described the strait as approaching a “worst-case scenario” with shipping traffic slowing sharply{{cite:cfe597159a8d}}.

The energy sector’s divergence from the rest of the market was stark. XLE’s 1.16% gain{{cite:fd8a8b69bdb9}} was the lone green close among major sector ETFs on Friday. The Arkansas Democrat-Gazette reported that the AI sell-off “yanked stock markets lower worldwide” while “oil prices continued to jump because of the war with Iran”{{cite:f7e5d85efbe9}}.

What makes this two-front week unusual is that the geopolitical risk bid for oil is colliding with a disinflationary impulse at home. June CPI slowed to 3.5% year-over-year{{cite:9a915387446b}}, and the FRED macro snapshot puts the Fed funds rate at 3.63% with the 10-year Treasury at 4.55%{{cite:85821adbb238}}. If oil stays elevated, the pass-through to gasoline prices could complicate the disinflation story — a tension Fed officials are explicitly navigating.

Fed Chair Warsh’s First Testimony

New Federal Reserve Chairman Kevin Warsh delivered his first semiannual Monetary Policy Report to Congress on July 14-15, testifying before the House Financial Services Committee and the Senate Banking Committee{{cite:1755dec64922}}{{cite:47a3827cc41c}}.

Warsh reaffirmed the Fed’s commitment to bringing inflation under control but offered few concrete signals on the timing of the next rate move{{cite:47a3827cc41c}}. The FOMC held rates steady at its June meeting — the fourth consecutive hold{{cite:1755dec64922}}. Top Fed officials, including NY Fed President John Williams, said inflation has peaked and backed holding rates steady, though some market participants still see the possibility of further action by September{{cite:9a915387446b}}.

Warsh characterized AI as “the most striking feature of the economy right now,” saying the AI infrastructure buildout would soon be called simply “investment”{{cite:1755dec64922}}. He announced five internal task forces to review communications, balance sheet policy, economic data, productivity and jobs, and inflation frameworks, with findings to be presented first to the FOMC and then to the public{{cite:1755dec64922}}. He pledged that any balance sheet policy changes would be telegraphed in advance{{cite:1755dec64922}}.

On independence, Warsh told lawmakers “we’re honored to be independent” and committed to following “the law and the data”{{cite:1755dec64922}} — a pointed exchange given the political pressure the Fed has faced on rates.

The Macro Backdrop: Calm Indicators, Anxious Consumers

The FRED macro snapshot as of June 2026 paints a mixed picture:

VIX at 16.73 and high-yield credit spreads at 2.71% suggest the systemic plumbing is not stressed{{cite:85821adbb238}}. But consumer sentiment at 44.8 — down 14% year-over-year and 10% month-over-month — is a genuine yellow flag{{cite:85821adbb238}}. The FRED analog search returns 2006-07 as the most similar period, when unemployment sat near 4.6-4.7%, CPI ran around 3.6-4.1%, and the economy was twelve to eighteen months from recession{{cite:85821adbb238}}.

That does not mean a recession is imminent. It means the current macro configuration — moderate inflation, low unemployment, solid GDP, but deeply negative consumer sentiment — has historical precedent in periods that turned out to be late-cycle. The question is whether the AI investment boom and solid growth can override the consumer pessimism signal, or whether that signal is the early tell.

What Each Side Would Need to Be Right

The semis bear case requires compute supply to keep outpacing demand normalization, rental rates to keep falling, and hyperscaler capex to eventually flatten as returns on AI infrastructure come under scrutiny. If B200 rental rates stabilize around the $4.22 level and stop declining, the repricing may be closer to over than beginning{{cite:67c68857ad8d}}. If they keep falling, the margin story gets worse.

The geopolitical escalation case requires the Strait of Hormuz disruption to persist or deepen, pushing oil higher and feeding through to gasoline and transport costs. A ceasefire or de-escalation would reverse the energy bid quickly. The fact that the first ceasefire collapsed does not guarantee the next attempt fails, but it lowers the prior probability of a quick resolution.

The disinflation-and-soft-landing case requires the CPI cooldown to continue despite oil’s geopolitical premium, the labor market to hold at 4.2% unemployment, and the consumer sentiment slump to remain a lagging — not leading — indicator. Real GDP at 2.66% supports this read{{cite:85821adbb238}}, but the 2006-07 analog reminds us that GDP can stay positive well into a late-cycle environment.

What to Watch Next

• B200 rental rate trajectory. If $4.22 per GPU-hour is the floor and rates stabilize, the worst of the semiconductor repricing may be over. Further declines would signal deeper margin compression{{cite:67c68857ad8d}}.
• NVIDIA’s next earnings call. Watch for whether management signals pricing discipline or leans into volume growth to offset margin pressure. Broadcom set the template; NVIDIA’s response will set the tone{{cite:67c68857ad8d}}.
• Strait of Hormuz shipping traffic. Any de-escalation or ceasefire would reverse the oil bid. Further attacks or a widening blockade would compound the energy shock{{cite:cfe597159a8d}}.
• Hedge-fund positioning normalization. Four consecutive weeks of net selling is a meaningful reset, but the flush needs to show a bottom. Watch for when block net outflow on NVIDIA turns positive — that is the capital-flow signal that distribution is done{{cite:67c68857ad8d}}.
• Consumer sentiment and July CPI. The June sentiment reading at 44.8 is deeply negative{{cite:85821adbb238}}. If July data confirms a sustained deterioration, the late-cycle analog from 2006-07 gains weight. The next CPI print will determine whether the disinflation trend holds despite oil’s geopolitical premium.
• Fed task force timeline. Warsh said findings go to the FOMC first, then the public{{cite:1755dec64922}}. Any signal on balance sheet policy changes could move rates markets well before implementation.

FN2 Research provides market commentary and education, not personalized investment advice. Past performance does not guarantee future results.