1. Interconnect Technologies (Interconnect / Connectivity)
High-speed connectivity (optical + electrical) between GPUs in AI data centers.
• $ALAB Astera Labs: Leader in PCIe retimers and CXL, very strong in AI servers.
• $MRVL Marvell Technology: One of the strongest players in optical DSP and interconnects.
• $CRDO Credo Technology: Leader in the Active Electrical Cable (AEC) market.
• $AVGO Broadcom: Very dominant in high-speed networking chips and silicon photonics.
• $NVDA Nvidia: Integrated interconnect solutions with NVLink (mostly within its own ecosystem).
• $MTSI MACOM Technology Solutions: Specialist in RF, analog, and optical semiconductors. Strong in components used in AI data center switches and optical modules.
• $LITE Lumentum: Active in optical engines, lasers, and transceivers. Plays a role in the optical side of AI networking. • $COHR Coherent: Specializes in laser and photonic components. Participates in the AI optical interconnect supply chain.
• $AXTI AXT Inc: Supplier of Indium Phosphide (InP) substrates. Provides critical materials for high-speed optical modules.
Note: $ALAB and $MRVL are the most talked-about names in this field recently.
2. Advanced Packaging
A critical area in AI chips (especially HBM + logic integration) (2.5D/3D packaging).
• $TSM TSMC: The clear leader with CoWoS technology (most chips like Nvidia Blackwell, etc., are packaged here).
• $INTC Intel: A strong alternative with EMIB and Foveros technologies (don't listen to those who call it junk).
• $ASX ASE Technology: The world's largest OSAT company, with high advanced packaging capacity. • $AMKR Amkor Technology: A significant OSAT player, especially for US and global customers.
3. Memory (Especially HBM)
The most critical component for AI training and inference (High Bandwidth Memory).
• $SKHY SK Hynix: Clear leader in the HBM market (Nvidia's main supplier, with around 50-60% market share).
• Samsung Electronics: Has very large capacity, strong in HBM3E and HBM4.
• $MU Micron Technology: Rapidly ramping up in HBM3E and HBM4, the only major US-based player.
• $SNDK SanDisk: Currently the cleanest pure-play in the NAND side. Demand for high-speed enterprise SSDs in AI data centers is very strong.
• $SIMO Silicon Motion: A fabless company that designs NAND flash controllers. It doesn't produce its own NAND chips, but it makes the controller chips, considered the "brain" of SSDs (especially enterprise and AI-focused ones).
SK Hynix is clearly ahead in the HBM segment.
4. Semiconductor Manufacturing and Testing Equipment
An area expected to boom with factory investments and increased capacity in the photonics sector.
Manufacturing Equipment (Wafer Fab Equipment):
• $ASML: Holds a monopoly position in EUV and High-NA lithography (most critical equipment).
• $AMAT Applied Materials: Leader in deposition and materials engineering.
• $LRCX Lam Research: Very strong in etching and thin film deposition.
Test & Inspection:
• $AEHR Aehr Test Systems: Leader in burn-in and reliability testing systems. Critical in the testing process of advanced packaging and AI chips.
• $KLAC KLA Corporation: Leader in process control and metrology.
• $TER Teradyne: Stands out in testing equipment.
These equipment companies will receive significant support in 2026-2027 as the photonics sector also grows.
5. Energy (Data Center Power)
• $BE Bloom Energy: One of the fastest companies to install solid oxide fuel cells (can be commissioned within months). A leader in onsite power for data centers. • $VST Vistra: Merchant power generator, very powerful in AI data center load growth. Moreover, it is currently very inexpensive in basic terms. • $CEG Constellation Energy: America's largest nuclear power producer. A key player in data center contracts. • $GEV GE Vernova: Offers fast installation options in turbine engines. Also a key player in AI power infrastructure. • Siemens Energy: Similarly active in gas turbine and energy solutions.
