Add back the 0.014 percentage points the 10-year shed on the session and the previous close comes to 4.796%, which makes that 4.818% intraday print, correctly reported as the highest since November 2023, a move of 2.2 basis points onto ground the market had already stood on [3][1]. The two-year closed lower as well [7]. From 4.782% the distance to the 5% everyone is talking around is 21.8 basis points [2], which this market can travel in an afternoon without any help from semiconductors.
What relief-for-investment does is take a number out of the customs schedule and put it into a capital budget: a duty you can avoid by building is a duty quoted in fabs rather than in percentage points, and the account we have names the two targets and the condition [1] without naming a rate or a spending threshold, which leaves anyone contracting for memory next quarter pricing a direction instead of a magnitude [2].
The counter-thesis is in the same day's tape. Japan's 10-year reached 3% on the 1st [12] behind budget requests of 143 trillion yen for the fiscal year beginning April 2027, a record for the fourth year running [13]; Britain's 30-year gilt printed 5.919%, the most since May 1998, and its 10-year the most since August 2007 [14]; German and French 10-years hit their highest since 2011 and 2008 [15]. None of those issuers buys DRAM from Korea. What they share is supply, and the American version of it is a fiscal 2026 deficit expected above 6% of GDP against 2.1% growth over the past year [17][18], a gap running close to three times the growth rate meant to carry it [7]; the Wall Street Journal read the G20 meeting in Asheville as a failing grade [16].
So my read, which is probably wrong in one specific way I will name: the tariff matters more as an allocation instrument than as an inflation input, because the path from memory contracts to consumer prices is slow and arguable while a capex condition binds on a board calendar, and the incentive to keep talking about it runs to November [21]. The price that would change my mind is the one already misbehaving, since futures held 62.3% odds of a hike at the September 15-16 meeting, down from 67.2% [11], on a day when private hiring missed consensus by 9,000, or 19% [5], and the bond market did not flinch [10]. If the official payrolls report on the 4th [22] comes in soft and yields rise anyway, the market is telling you it fears supply and prices rather than chips; if memory contracts jump while Samsung's and SK hynix's American spending plans do not move, then the allocation story is the wrong one. That second test is the cheaper one to run.
