Stifel reiterates Buy on Monolithic Power stock on capacity growth

Monolithic Power Systems expanded annual revenue capacity targets from its prior $4 billion level to more than $6 billion, with a longer-term goal of $8 billion, supported by a long-term GlobalFoundries Singapore manufacturing agreement. The company reported Q2 2026 adjusted EPS of $6.50 and revenue of $980.6 million, exceeding expectations, while revenue rose 48% year over year on enterprise data-center demand. Analysts forecast 48% revenue growth this year, and multiple firms raised price targets as MPWR expands AI data-center, automotive and robotics power-management capacity while diversifying supply chains away from China exposure.

Analysis

MPWR’s capacity plan matters less as a near-term volume unlock than as evidence that it expects AI power content to remain structurally tight through the next design cycle. The key economic question is utilization: adding internalized/dedicated 300mm supply can protect gross margin and lead times, but a move from capacity reservation to underutilization would be dilutive given the stock’s premium growth multiple. The market has likely discounted the latest earnings beat and upward revisions; the incremental rerating case requires enterprise-data revenue to continue outgrowing the broader analog/power portfolio for at least the next two quarters.

GFS gains a higher-quality, longer-duration utilization customer in mature-node power management, an area less exposed to leading-edge logic cyclicality. This is strategically constructive for Singapore fab loading and customer diversification, but MPWR alone is unlikely to move consolidated GFS earnings; the investable read-through is stronger for GFS’s ability to monetize geographically diversified, non-China manufacturing at attractive contract terms. Competitively, ON and Texas Instruments face limited direct displacement in high-performance AI voltage-regulation modules, while Vicor is the closer product-level risk if MPWR converts supply assurance into additional hyperscaler socket wins.

Immediate upside in MPWR is constrained by crowded bullish positioning and target-price momentum. Over the next 1–3 months, the catalyst is third-quarter enterprise-data growth and gross-margin progression despite supply-chain investment; over 6–18 months, qualification of the next-node platform and sustained AI power content per server determine whether the revenue ceiling is genuinely raised rather than merely de-risked. Falsification would be enterprise-data growth decelerating materially for two consecutive quarters, gross-margin guidance falling despite revenue growth, or evidence that hyperscaler AI capex shifts from training clusters toward lower-power inference deployments.

Key Decisions for Investors

• Maintain MPWR as a core AI-power long only on post-earnings consolidation rather than chase analyst-target momentum; add if the stock retraces 10–15% without a cut to enterprise-data guidance. Underwrite a 12-month outcome on sustained data-center growth and stable-to-up gross margin; exit/reduce on two quarters of enterprise-data deceleration or a meaningful margin-guide reset.
• Use a relative-value expression: long MPWR / short TXN in equal dollar risk over 3–6 months. MPWR has greater AI-server power-content sensitivity, while TXN remains more exposed to broad industrial and automotive normalization; cover if industrial demand inflects sharply upward or MPWR’s enterprise-data growth loses its premium.
• Accumulate GFS selectively as a secondary beneficiary of diversified mature-node demand, but treat the MPWR agreement as a confirmation signal rather than an earnings catalyst. Reassess after GFS reports Singapore utilization, wafer pricing, and long-term agreement economics; avoid sizing up solely on the customer announcement.
• Monitor Vicor (VICR) for negative read-through: if MPWR cites design wins in high-current AI power stages or accelerating enterprise-data mix, consider a tactical VICR short versus MPWR long. The trade is invalidated by VICR reporting renewed hyperscaler design-win momentum or a material margin recovery.