Morning Bid: Yield to worst

The 10-year U.S. Treasury yield reached a 24-year high of 5.34% after climbing more than 80bps in Q3, while French yields neared 5% and the France-Germany spread widened above 140bps, its widest since 2012. Softer U.S. data, including 3.4% August PCE inflation, pushed the implied probability of an October Fed hike below 50% from about 70%, but persistent above-target inflation and resilient growth leave further bond-market pressure possible. Geopolitical risks lifted Brent crude more than 4%, diesel shortages intensified after Chinese refiners suspended October exports, and U.S.-China tariff cuts covering $60B of goods provided a limited offset.

Analysis

The relevant cross-asset mechanism is not simply higher rates, but a renewed rise in the equity risk premium for long-duration growth: NVDA’s earnings revisions can offset part of that pressure, while AAPL’s more mature growth profile and consumer-finance exposure leave its multiple more vulnerable. NVDA’s enlarged repurchase capacity creates a mechanical bid during blackout-free windows, but it does not protect against a valuation reset if real yields remain elevated; the more consequential near-term variable is whether payrolls and FOMC minutes reprice the terminal-rate path. A soft labor print could produce a days-to-weeks duration rally led by semiconductors, whereas resilient data would likely extend the compression in high-multiple AI beneficiaries.

Energy inflation is a second-order headwind for AAPL: it taxes discretionary consumer budgets and raises freight/input costs, while a stronger dollar translated from rate differentials compounds international-revenue pressure. For NVDA, the direct energy effect is smaller, but higher power prices and utility constraints increase the all-in cost of AI deployment, potentially slowing marginal data-center capacity decisions over the next 6-18 months. The consensus is likely too focused on GPU demand visibility and too dismissive of the financing/power constraint on customers; capex commitments can remain high while deployment schedules, and therefore revenue recognition, slip.

A tactical AI long remains justified only on a disinflationary-rate reversal, not as an unconditional momentum trade. The falsification point is a sustained rise in real yields alongside any evidence that hyperscalers are stretching data-center construction timelines; conversely, falling yields without downward earnings revisions would support renewed multiple expansion.

Key Decisions for Investors

• Maintain NVDA as a tactical long only if the 10-year Treasury yield retreats after payrolls/FOMC minutes; use a 1-3 month horizon and define risk with a break above the recent yield high or evidence of delayed hyperscaler capacity deployments. Prefer call spreads over outright calls to limit exposure to implied-volatility repricing.
• Initiate a 1-3 month relative-value pair: long NVDA / short AAPL in equal beta-adjusted dollars if rates remain range-bound or ease modestly. NVDA retains stronger earnings-revision support, while AAPL faces greater consumer, FX and duration-multiple sensitivity; exit if NVDA data-center commentary weakens or AAPL shows improving China/upgrade-cycle demand.
• If the 10-year yield decisively exceeds its recent peak following strong payrolls, reduce gross exposure to both names rather than adding on weakness; the likely initial move is multiple compression, not an earnings-driven buying opportunity. Reassess after the next earnings cycle provides customer-capex and gross-margin evidence.
• Watch diesel and power-price indicators as a 6-18 month AI-capex risk alert. A sustained rise in data-center power costs or reported utility interconnection delays would favor trimming NVDA and rotating technology exposure toward lower-capex software rather than treating it as a broad semiconductor demand signal.