The setup is less about single-name earnings and more about whether the market can keep paying up for long-duration growth while real yields reprice. NVDA is the hinge for the entire AI complex: if the guide is merely solid, crowded positioning can still unwind because the market needs evidence of accelerating capex, not just continued demand. That makes SMH/QQQ vulnerable to a multiple reset even on an earnings beat, with second-order pressure on TSM, AVGO, AMAT and other AI supply-chain names.

PCE matters because a hot print would amplify the move in rates and make any tech upside harder to monetize. If inflation is sticky, the Fed stays higher-for-longer, which hurts leveraged cyclicals and small caps first, but also keeps pressure on any equity valued on distant cash flows. The Canada trade angle is a slower-burn inflation impulse: retaliation raises input costs and keeps goods disinflation from flowing through, which is a margin headwind for autos, industrials and cross-border logistics over the next 1-3 months.

The contrarian risk is that consensus is treating this as a binary 'NVDA beats, risk-on' event when the more important variable is yield response. A soft PCE print combined with a confident NVDA capex commentary would force a sharp squeeze in duration shorts and could re-rate semis higher for weeks. Falsifier: core PCE at/below expectations and NVDA raising forward demand/capex guidance enough to pull 10Y yields lower; that would invalidate the bearish duration/semis view.