ReportGem ReportGem 中文

REAL-TIME GLOBAL RESEARCH

JPM|Marissa’s Macro Musings: Cracks in the Facade: When Distillates, DeepSeek Disciples, and Data Centers Collide

Published: 2026-07-20Institution: JPMorganPages: 8Original language: EnglishEvidence page: 1

Research evidence excerpt

JPM|Marissa’s Macro Musings: Cracks in the Facade: When Distillates, DeepSeek Disciples, and Data Centers Collide

Specialist Sales

US Specialist Sales J P M O R G A N

20 July 2026

JPM | Marissa’s Macro Musings - Cracks in the Facade: When

Distillates, DeepSeek Disciples, and Data Centers Collide

Marissa Gitler

+1 212 622 2934

marissa.gitler@jpmorgan.com

US Thematics focuses on key macro views, market debates, and favored investment themes.

Market Dynamics of Note:

Oil & Products Risk Is Underpriced Relative to Equity Complacency: Despite a confluence of supply-side shocks, crude

is barely moving while equities remain buoyant on backward-looking CPI/PPI relief. Distillate crack spreads surging toward

record highs signals the shock is morphing into a refining bottleneck story, which has more persistent, broad-based

inflationary pass-through than a crude spike alone.

AI Capex Cycle Is the Dominant Macro Factor, But Concentration Risk Is Bifurcated: The Kimi K3/Qwen disruption

fear was quickly absorbed in the US due to compute dependency on NVIDIA’s ecosystem, but the Korea margin call episode

(1.2M accounts, 340K wiped out) is a stark reminder of what concentrated AI exposure looks like without rotation flexibility.

In the US, momentum factor drawdown (~25% off June highs for JPPQMO) is occurring within a resilient index as capital

rotates rather than exits. This is a factor-level risk, not a systemic one, yet. GOOGL earnings Wednesday are critical:

hyperscaler CAPEX guidance ($325-350B investor expectation vs. ~$250B Street for 2027) will set the tone for whether the

AI infrastructure trade re-accelerates or faces a valuation reset.

AI ROI Dependency as a Systemic Equity Risk(?): The financials earnings read-through indicates that AI has graduated

The English excerpt is extracted automatically from the cited source page and may contain layout or recognition errors. It is never batch translated.

Open report viewer