Promising Semiconductor Companies Beyond Big Four | Generated by AI

Home 2026.06

Question: Besides Marvell, Broadcom, SK Hynix, and SanDisk, what other semiconductor companies look promising?

Answer:

Your four picks cluster around two themes — custom AI silicon/networking (Marvell, Broadcom) and memory/storage (SK Hynix, SanDisk). The most interesting names elsewhere in the stack fall into a few layers:

Foundry / manufacturing

TSMC (TSM) is the one almost every analyst list leads with. It controls about 72% of the chip foundry market per Counterpoint Research, making it indispensable for the AI infrastructure build-out, and posted a 40% year-over-year revenue increase in Q1, primarily AI-driven. Its strength in 3nm and 2nm makes it essential to the global supply chain, manufacturing for Nvidia, Apple, AMD, Qualcomm and emerging AI startups. The caveat: geopolitical risk from being Taiwan-based, plus the sector’s inherent cyclicality.

Compute (GPUs/accelerators)

Nvidia (NVDA) and AMD (AMD) remain the core compute plays. Banks highlight Nvidia, AMD, Broadcom, and Marvell among the key beneficiaries spanning AI compute and networking — so AMD and Nvidia are the natural complements to your Broadcom/Marvell exposure.

Memory (direct peer to SK Hynix)

Micron (MU) is the closest analog to SK Hynix. Its HBM3E and advanced memory products are seeing record demand from GPU and accelerator makers, with tight HBM supply boosting pricing power and deepening Nvidia partnerships.

Equipment and EDA (“picks and shovels”)

This is the layer people often miss. Applied Materials (AMAT), Lam Research (LRCX), Cadence (CDNS), and Synopsys (SNPS) are flagged as core equipment and EDA beneficiaries of AI data-center capex into 2026. ASML belongs here too as the EUV lithography monopoly — every leading-edge fab (TSMC, Samsung, Intel) depends on it. These names benefit regardless of which chip designer wins, which makes them structurally less risky than betting on a single architecture.

Adjacent infrastructure

Amphenol (APH) gets mentioned alongside Nvidia and Micron as positioned for the 2026 demand spike — connectors/interconnects for AI servers. Also worth knowing: global semiconductor sales are projected to surge 26.3% YoY to $975.4B in 2026, with AI accelerators, HBM, advanced logic, and supporting infrastructure as primary growth engines.

A few caveats from me: I’m not a financial advisor, and these are analyst views, not recommendations. Several of these names are already in overbought territory, making timing and risk management important, and many of these stocks have run hard over 12–18 months — a lot of AI optimism is priced in. If you want broad exposure without single-name risk, semiconductor ETFs like SOXX or SMH are commonly suggested as a base before layering individual positions.

Given your interest in GPU compute, the equipment/EDA layer (ASML, AMAT, LRCX, Cadence, Synopsys) is probably the most under-discussed and worth researching — it’s the part of the stack that doesn’t care whether NVIDIA, AMD, or custom ASICs win the inference war.

References:

Back Donate