Citi Research has reduced its semiconductor conviction list to four companies it views as best positioned through mid-2026, citing two principal themes: continued AI infrastructure investment and analog businesses that can drive margin improvement from within. The research note places particular emphasis on how these dynamics create differentiated paths to outperformance in what Citi characterizes as an otherwise uneven industry environment.
Broadcom emerges as Citi’s highest-ranked semiconductor pick. The firm points to Broadcom’s exceptional earnings momentum, noting that C27 consensus EPS rose 25% after the company reported December-quarter results. Citi links that revision to accelerating demand for Broadcom’s custom AI chips, and expects that momentum to compound through 2027. With a $475 price target that implies roughly 40% upside from prevailing levels, Citi presents Broadcom as the group’s most conviction-worthy way to participate in AI infrastructure spending.
NVIDIA remains the archetypal AI exposure in Citi’s view. The research house assigns a $270 price target for NVIDIA, which implies 52% upside and represents the highest expected total return across Citi’s semiconductor coverage. Citi highlights the upcoming GTC conference as a short-term catalyst and identifies investor attention on inference workloads, the architectural transition from DRAM/HBM to SRAM, and the debate over co-packaged optics versus copper. Overall, Citi frames NVIDIA as a foundational holding for portfolios seeking meaningful AI exposure.
Texas Instruments is singled out within the analog cohort for its internal improvement path. Citi’s bull case for TXN is driven not by a recovery in cyclical demand but by the company’s ability to expand gross margins internally - a form of self-help that lends a defensive character relative to cyclically exposed semiconductor names. Citi also notes TXN’s nearly 3% dividend yield as an additional income buffer uncommon among semiconductor equities. The $235 price target corresponds to about 25% upside in Citi’s scenario.
Monolithic Power Systems completes the quartet based on a power-semiconductors product cycle Citi believes remains in early stages. As AI data centers raise power density requirements, the firm argues MPWR’s specialized power solutions become increasingly important to infrastructure builds. Citi assigns a $1,350 price target to Monolithic Power, implying near-term upside of about 33%, and describes it as the highest-conviction analog name for investors seeking AI buildout exposure without direct ownership of the large GPU providers.
Collectively, the four names reflect Citi’s dual emphasis: direct participation in AI infrastructure via companies with bespoke AI chip demand, and exposure to analog businesses that can improve profitability through internal measures as end-market demand fluctuates. Citi’s view positions Broadcom and NVIDIA as primary AI infrastructure plays, while Texas Instruments and Monolithic Power Systems provide a route to AI-related end markets through analog and power-semiconductor dynamics.
Disclosure: The article reflects Citi Research’s product-level assessments and price targets as presented in the research note. No additional facts or forecasts have been introduced beyond those included in Citi’s analysis.