Dollar holds gains ahead of Jackson Hole; won rises on BOK rate hike

Nvidia reported Q2 results that “blow past” estimates and guided for ~70% annual revenue growth, a clear earnings momentum positive for the stock/AI complex. At the same time, hotter-than-expected U.S. PCE inflation (PCE +3.7% y/y in July vs 3.6% forecast) kept alive expectations for further Fed tightening, with markets pricing about a 60% chance of a rate hike next month. FX markets remained sensitive as the dollar held near a one-week high (DXY ~99.14) alongside tighter-rate expectations.

Analysis

The market should read this as a confirmation that AI capex is still in an acceleration phase, not a digestion phase. That matters because in a high-rate backdrop, investors are increasingly paying for earnings visibility rather than distant TAM narratives; NVDA’s cleanest path to multiple support is that its revenue base is translating into near-term cash flow faster than most mega-cap software or semiconductor peers. The immediate loser is any “AI spend is peaking” thesis: if demand is still outpacing supply, then the bottleneck shifts from willingness to spend to physical deployment, which extends the cycle for infrastructure names upstream and downstream.

Second-order winners are the picks-and-shovels around deployment intensity, not just chip design. TSM, ASML, AMAT, KLAC, VRT, and ETN benefit if the backlog turns into sustained orders for advanced packaging, lithography, power, and cooling; the risk is less about one quarter of earnings and more about whether customers pre-commit multi-quarter capex. Conversely, AMD and other AI challengers face a harder valuation setup because they need either better product cadence or a lower-cost narrative to compete against a still-dominant incumbent.

The macro overlay is important: hotter inflation keeps real yields elevated, which typically compresses duration-heavy growth multiples. That creates a subtle tension—NVDA can still outperform on estimate revisions, but the broader semi complex may not get the same multiple lift unless Jackson Hole softens policy expectations. Over 1-3 months, the key falsifier is whether hyperscaler capex commentary or order timing slows; over 6-18 months, the thesis breaks if export controls, supply normalization, or customer concentration start capping incremental revision upside.

Key Decisions for Investors

• Long NVDA vs short QQQ into the next 2-6 weeks: use earnings strength to express relative earnings-revision momentum while hedging broad multiple risk from sticky inflation; thesis fails if NVDA gaps down below its post-earnings support and underperforms QQQ on follow-through volume.
• Pair trade: long NVDA / short AMD over 1-3 months. NVDA has the cleaner demand visibility and pricing power; AMD needs a valuation reset or stronger AI revenue proof. Cover if AMD guides AI revenue materially above consensus or the spread narrows sharply on better-than-expected hyperscaler commentary.
• Long infrastructure beneficiaries (VRT, ETN) vs short a basket of low-quality AI software names over 1-3 months. The mechanism is physical deployment intensity: power, cooling, and data-center buildout should see faster order conversion than story stocks if capex remains aggressive.
• Buy call spreads on SMH or NVDA for the next earnings cycle rather than outright calls. The setup is supportive but macro-sensitive; defined-risk upside captures continued revision momentum while limiting damage if higher yields re-rate the complex lower.