Monolithic Power Systems

▲ Bull Case

• ◆NVIDIA Blackwell/Rubin GPU families consume 1,200–2,000W, requiring MPS multi-phase VRM designs that expand dollar content per board 3–5x, driving enterprise data segment to 50%+ of revenue and sustaining 30%+ YoY growth through 2027.
• ◆Automotive segment inflects above $500M annualized run-rate by mid-2026, demonstrating successful diversification beyond AI and reducing cyclical revenue risk profile.
• ◆Gross margins expand durably to 57–59% as AI-specific products command premium ASPs and no meaningful competitive pricing pressure emerges from Texas Instruments or Analog Devices in AI data center applications.

▼ Bear Case

• ◆BIS expands export controls to cover advanced multi-phase VRM ICs for AI accelerators, forcing MPS to exit 20–30% of China revenue representing mainland AI data center build-outs, triggering $400–600M annualized revenue reset.
• ◆AI hyperscaler capex pauses in 2026 as hyperscalers reassess ROI on LLM infrastructure following slower-than-expected enterprise AI adoption, causing enterprise data segment to decelerate from 40% to sub-10% YoY growth and triggering 30–40% earnings cut.
• ◆Texas Instruments successfully penetrates AI data center VRM market using its 300mm manufacturing cost advantage, winning 20–25% market share from MPS and compressing gross margin below 54% within 18–24 months.

“The consensus debate is not whether MPS is high-quality—it is. Rather, whether the ~31x GAAP forward P/E and ~9x EV/Sales already prices in the AI bull case. Bulls (~70% of sell-side) argue sustained 20%+ revenue growth justifies premium multiples through 2027+. Bears argue: (a) China concentration is a binary geopolitical tail risk justifying 20–30% multiple compression, (b) 53–56% historical gross margin band suggests 58–60% bull-case level is structurally unreachable, (c) Texas Instruments' 300mm fab cost advantage will compress pricing within 3–5 years. Probability-weighted expected value (~$520) is nearly identical to current price—the market appropriately prices dispersion, but skew is mildly negative as bear-case downside (-50%) is sharper than bull-case upside (+50%).”

• ◆NVIDIA Blackwell production ramp confirmation; MPS primary VRM supplier status confirmed (+15–25% re-rating potential, 0–6M horizon)
• ◆Enterprise data revenue beats >30% YoY for 2–3 consecutive quarters (+5–10% per quarter, 0–6M horizon)
• ◆Hyperscaler capex guidance upside from MSFT, GOOGL, AMZN, META (+5–10% sympathy moves, 0–6M horizon)
• ◆Automotive segment crosses $500M annualized run-rate with clear path to $1B (diversification re-rating, 6–18M horizon)
• ◆Sustained gross margin >57% for 3+ consecutive quarters (+5–10% EPS revisions, 6–18M horizon)
• ◆Hyperscaler custom-ASIC power win publicly confirmed (Google TPU, Amazon Trainium, Microsoft Maia; +10–20% per win, 6–18M horizon)
• ◆48V rack architecture industry-wide adoption by OCP Foundation (+30–50% re-rating; TAM expansion unlock, 18M+ horizon)

• ◆China revenue concentration / BIS export controls targeting advanced multi-phase VRMs for AI (Med-High probability, Severe impact -40% to -60%); $250–300M annual revenue at risk; only mitigations are EAR99 status and broad civilian applications.
• ◆AI hyperscaler capex cycle deceleration if ROI on LLM infrastructure questioned (Medium probability, High impact -25% to -40%); partially mitigated by automotive diversification and multi-customer hyperscaler exposure.
• ◆Multiple compression from current ~31x P/E if near-term growth assumptions reset (High probability, High impact -25% to -40%); offset by AI secular thesis and 22% dividend CAGR track record.
• ◆Texas Instruments competitive penetration in AI VRM using 300mm fab cost advantage (Medium probability, Med-High impact -15% to -30%); mitigated by BCD process moat and NVIDIA design depth relationships.
• ◆TSMC supply disruption or Taiwan Strait escalation (Low probability, Severe impact -40% to -70%); mitigated by TSMC resilience, inventory buffers, and limited foundry alternatives.