AMD Post-Earnings Gaps: Hold Rate Across the AI Cycle

AMD Gaps in the AI Era

AMD has gapped (open more than 3% from prior close) on every one of the last 8 earnings prints. Five of those gaps were >8%, and three were >10%. That's a much higher gap-rate than the average large-cap semi (~50-60% of earnings produce a measurable gap), reflecting the binary nature of AI-cycle prints: the data-center number either confirms the multiple or undermines it.

Hold vs Fade

Across those 8 gaps, 6 held into the day-0 close (didn't retrace more than 30% of the gap). 5 of those 6 extended further by day +5. The 2 that faded both shared a feature: a margin-guide-down despite a top-line beat. Markets in the AI cycle care more about gross-margin trajectory than headline revenue.

The cohort retrieval picks this up post-print: gaps that match the high-margin-trajectory analog set tend to extend; gaps that match the margin-pressure cluster tend to fade. The shape of the chart at 11am Wednesday is the cleanest input.

• Gap rate (last 8 prints): 8/8 (100%)
• Day-0 hold rate: 6/8 (75%)
• Day +5 hold rate: 5/8 (63%)
• Common fade trigger: margin guide-down despite revenue beat (2/2 fades)

If AMD Gaps Tuesday

Wait for the chart shape to settle before retrieving the cohort. The AH and pre-market on AMD earnings nights are dominated by HFT flow and don't produce a stable analog match. Once the cash session opens and absorbs the first 60-90 minutes of institutional flow, the cohort match becomes meaningful.

An open above the prior all-time-high pulls a different analog cluster than an open below the 50-day MA — the embedding captures structural breakout vs structural failure, which historically have very different 5-day base rates.