Gross margin expanded 160bps, opex leverage improved 330bps, and EBITDA still fell year-over-year.
Thesis: Diodes is a fabless-ish analog/power supplier running a 50/50 hybrid model, so they hold real pricing leverage while bottlenecks sit in power product supply across the sector. Beat-and-raise six quarters running with channel inventory below normal is demand pulling product, not stuffing. The EBITDA dip is depreciation on capacity that is already being utilized, and free cash flow stayed positive at $34.8M on $33.6M capex. The bull case holds as long as the AI server and automotive content ramps materialize as new products launch over the next few quarters.
Verdict: LONG — Conviction: MEDIUM
Catalyst: Q3 print of $510M at 35% gross margin, plus disclosure of specific AI server design wins behind the $267 per-platform content claim as new products launch over coming quarters.
Key Risk: AI content claim ($267/platform vs $109 today) has no named customer, platform, or design win. If AI server cycle decelerates, six quarters of double-digit growth plus an above-range guide becomes Unsustainable Trend Confidence with channel inventory normalizing against it.
The Tell: Gary Yu went beyond guidance unprompted: 'with the market we're seeing, we feel like it should be stronger than usual' for 2027, after Emily confirmed Q4 momentum verbally. Volunteered forward optimism on a call that otherwise sticks to one-quarter guidance shows he believes capacity build converts to revenue, not a setup for a guide-down.
Friction Level: MODERATE_FRICTION — Both sides agree on the operational data. Disagreement is whether EBITDA compression from capex-fueled depreciation is the cost of a structural growth engine or evidence that the 'margin expansion engine' narrative oversells unit economics. Second fault line: the $267 AI content claim has no attached socket.
2026Q2 · Aug 5, 2026
3 of 5 readings agreed3 LONG · 2 HOLDRead the full call
Revenue Up 22%, EBITDA Down: Capex Is Eating The Operating Leverage
Diodes is a fabless-ish analog/power supplier running a 50/50 hybrid model, so they hold real pricing leverage while bottlenecks sit in power product supply across the sector. Beat-and-raise six quarters running with channel inventory below normal is demand pulling product, not stuffing. The EBITDA dip is depreciation on capacity that is already being utilized, and free cash flow stayed positive at $34.8M on $33.6M capex. The bull case holds as long as the AI server and automotive content ramps materialize as new products launch over the next few quarters.
Gary Yu went beyond guidance unprompted: 'with the market we're seeing, we feel like it should be stronger than usual' for 2027, after Emily confirmed Q4 momentum verbally. Volunteered forward optimism on a call that otherwise sticks to one-quarter guidance shows he believes capacity build converts to revenue, not a setup for a guide-down.
Sixth consecutive quarter of double-digit YoY growth. Q2 revenue $445.5M beat guidance midpoint by 10% sequential; Q3 guide of $510M raises the bar again with 30% YoY. Non-GAAP EPS guide of $1.05 vs $0.70 current, a 2.8x YoY lift.
Gross margin up 160bps YoY to 33.1%, guiding another 190bps to 35% on utilization gains plus cost initiatives from the downturn. Opex improved from 28.9% to 25.6% of revenue.
Capex of $33.6M, 7.5% of revenue, within 5-9% target. Cash from ops $68.5M covers it with $34.8M free cash flow left. 6-to-8 inch migration, external capacity in Korea/Taiwan, internal assembly at 75% all backed by spending.
$267 per-platform content opportunity vs $109 today is a 145% expansion claim with no named customer, no socket, no hyperscaler commitment. AI also attached to smart glasses, IoT, smart home, wearables in the same remarks.
Six straight double-digit growth quarters plus an above-range Q3 guide builds confidence on a trend. Communication revenue already down 7% QoQ; Asia is 79% of revenue; channel inventory below normal creates a comparison that reverses on normalization.
Management touts utilization gains as margin driver while acknowledging 'pockets' of supply constraint. 8-inch migration and assembly additions are underway but not yet producing at scale, so capacity is the near-term ceiling on upside.
+72
CONFIDENT"Gary Yu volunteered specifics unprompted: 50/50 internal-external split, 6-to-8 inch migration, named external geographies (Korea, Taiwan), selective DFN/CSP capacity additions, and a 2027 call of 'stronger than usual' beyond guidance range. Emily gave quantified channel inventory (11-14 week range, now below)."
Rich unit economics: inventory days, channel weeks, POS, capex ratio, tax rate, EBITDA reconciliation all disclosed. Gap: SPFAB P&L refused, AI content opportunity has no customer or design win attached, supply constraints described only as 'pockets' without quantification.
Q3 print of $510M at 35% gross margin, plus disclosure of specific AI server design wins behind the $267 per-platform content claim as new products launch over coming quarters.
AI content claim ($267/platform vs $109 today) has no named customer, platform, or design win. If AI server cycle decelerates, six quarters of double-digit growth plus an above-range guide becomes Unsustainable Trend Confidence with channel inventory normalizing against it.
The verdict is public. The case behind it is Pro: the forensic reading of guidance and language, the bull and bear cases, the red flags, and the play. It opens to everyone on Nov 3, 2026.