Semiconductors & Chip Manufacturing

AGI Sector Scan — Mapping every narrow sub-sector of semiconductors to identify US-listed companies positioned to capture outsized demand as recursive self-improvement drives exponential compute scaling.
Module 02 of N 15 Sub-Sectors 67 Companies US-Listed (incl. ADRs) May 2026

Executive Summary

Semiconductors sit at the absolute foundation of AGI. Every training run, every inference call, every recursive self-improvement loop ultimately bottlenecks on silicon. The AGI buildout is creating the largest sustained demand shock the semiconductor industry has ever seen — dwarfing the PC, smartphone, and cloud eras combined.

Not all sub-sectors benefit equally. GPU/accelerator designers and HBM memory makers face demand that is literally insatiable for the next 2-3 years. Equipment makers and foundries enjoy derivative demand but with longer lead times. Analog, power, and materials companies see real but more moderate tailwinds. This report maps each sub-sector's AGI exposure, supply dynamics, and key investable names.

15
Sub-Sectors
67
Companies
8
High Conviction
5
Medium Conviction
2
Low (Indirect)

Table of Contents

High   Medium   Low
1

GPU & AI Accelerators

High

How It Works

GPUs and custom AI accelerators (ASICs) perform the massively parallel matrix math that powers training and inference of large neural networks. Modern AI accelerators pack thousands of compute cores alongside high-bandwidth memory interfaces, optimized for the throughput-over-latency tradeoff that deep learning demands. NVIDIA dominates with its CUDA ecosystem, but Google, AMD, and startups compete with custom silicon.

Supply / Demand Dynamics

  • AI demand driver: This is THE bottleneck. Every dollar of AI capex starts with accelerator silicon. AGI-scale training clusters require 100,000+ GPUs. Recursive self-improvement multiplies demand non-linearly.
  • Supply constrained: Severely. TSMC CoWoS packaging and HBM supply are the binding constraints. Lead times remain 6-12 months. NVIDIA's Blackwell and next-gen Rubin architectures are sold out years ahead.

Key US-Listed Companies

NVIDIA NVDA AMD AMD Broadcom AVGO Marvell MRVL Intel INTC Alphabet (TPU) GOOGL Amazon (Trainium) AMZN
2

AI Inference & Edge Chips

High

How It Works

Inference chips are optimized for running trained models at low latency and low power, rather than training them. As AGI-level models get deployed to billions of users and autonomous agents, inference compute will dwarf training compute. Edge inference pushes this into devices — phones, cars, robots — using specialized NPUs (neural processing units) and low-power accelerators.

Supply / Demand Dynamics

  • AI demand driver: Extreme. Once AGI-class models exist, inference demand scales with every user, every agent, every API call. This is the recurring revenue layer of the AGI stack.
  • Supply constrained: Moderate. Less CoWoS-intensive than training GPUs, but still competes for leading-edge fab capacity. Edge chips use mature nodes (5nm-7nm) with better availability.

Key US-Listed Companies

NVIDIA NVDA Qualcomm QCOM AMD AMD Broadcom AVGO Lattice Semi LSCC
3

FPGAs & Programmable Logic

Medium

How It Works

FPGAs (Field-Programmable Gate Arrays) are reconfigurable chips that can be rewired after manufacturing. They sit between general-purpose CPUs and fixed-function ASICs: more flexible than ASICs but less power-efficient. In AI, they serve niche roles in inference acceleration, network preprocessing, and rapid prototyping of new architectures before committing to full custom silicon.

Supply / Demand Dynamics

  • AI demand driver: Moderate. FPGAs see AI tailwinds in data center SmartNIC offload and inference at the edge, but GPUs and custom ASICs dominate the core AI workload. They remain important for telecom and defense AI applications.
  • Supply constrained: Mild. Most FPGAs use mature process nodes. Supply normalized after the 2021-2022 chip shortage.

Key US-Listed Companies

AMD (Xilinx) AMD Lattice Semi LSCC Intel (Altera IPO) INTC Microchip Tech MCHP
4

Memory: HBM, DRAM & NAND

High

How It Works

AI accelerators need massive memory bandwidth to feed their compute cores. HBM (High Bandwidth Memory) stacks multiple DRAM dies vertically with through-silicon vias, delivering 5-10x the bandwidth of standard DRAM. Every NVIDIA H100/B200/Rubin GPU requires multiple HBM stacks. Standard DRAM handles system memory in servers, while NAND flash stores training datasets and model checkpoints.

Supply / Demand Dynamics

  • AI demand driver: Extreme for HBM, strong for server DRAM. HBM demand is growing 100%+ annually as every AI GPU ships with 4-8 HBM stacks. HBM now commands ~5x the ASP per bit vs. commodity DRAM.
  • Supply constrained: Severely for HBM. Only three companies globally can make it (SK Hynix, Samsung, Micron). HBM yield rates are lower than DRAM, and HBM3E/HBM4 require advanced packaging. HBM is sold out through 2027.

Key US-Listed Companies

Micron MU SK Hynix (ADR) HXSCL Samsung (OTC ADR) SSNLF Western Digital WDC Seagate STX
5

Foundries (Contract Chip Manufacturing)

High

How It Works

Foundries manufacture chips designed by others. Every fabless AI chip company (NVIDIA, AMD, Broadcom, Qualcomm) depends on foundries to turn their designs into physical silicon. Leading-edge AI chips require the most advanced process nodes (3nm, 2nm), which only TSMC and Samsung can currently produce at scale. Intel is attempting to re-enter as a foundry.

Supply / Demand Dynamics

  • AI demand driver: Enormous. AI chips consume a large and growing share of leading-edge wafer capacity. TSMC's AI-related revenue is growing 3x faster than its overall business.
  • Supply constrained: Yes. Building a new fab costs $20-40B and takes 3-4 years. Advanced node capacity is allocated years ahead. CoWoS advanced packaging is the tightest bottleneck within foundries.

Key US-Listed Companies

TSMC TSM Intel Foundry INTC GlobalFoundries GFS Samsung (OTC ADR) SSNLF UMC UMC
6

Semiconductor Equipment: Lithography

High

How It Works

Lithography tools project circuit patterns onto silicon wafers using light. EUV (extreme ultraviolet) lithography uses 13.5nm wavelength light to print features at 3nm and below. Each EUV machine costs $350M+ and is essentially the most complex machine humanity has ever built. Without lithography, no advanced chips exist.

Supply / Demand Dynamics

  • AI demand driver: Direct. Every new AI chip fab needs dozens of EUV tools. TSMC's Arizona, Japan, and new fabs all require ASML's latest High-NA EUV machines.
  • Supply constrained: Extremely. ASML has a literal monopoly on EUV. They produce ~50-60 EUV systems per year. Backlog stretches years. This is the single hardest bottleneck in the semiconductor supply chain to replicate.

Key US-Listed Companies

ASML ASML Lasertec (OTC) LSRCY Onto Innovation ONTO
7

Semiconductor Equipment: Deposition & Etch

High

How It Works

After lithography patterns a wafer, deposition tools add thin films of materials (metals, insulators, barriers) and etch tools selectively remove material to form 3D circuit structures. Advanced nodes require hundreds of deposition and etch steps per wafer. HBM manufacturing also requires specialized deposition for through-silicon vias and bonding layers.

Supply / Demand Dynamics

  • AI demand driver: Strong. More fabs = more equipment orders. Advanced packaging (CoWoS, HBM stacking) adds incremental deposition/etch steps beyond what traditional chip manufacturing requires.
  • Supply constrained: Moderate. These are oligopoly markets (Lam, Applied Materials, Tokyo Electron dominate), with backlogs elevated but not as extreme as ASML.

Key US-Listed Companies

Applied Materials AMAT Lam Research LRCX Tokyo Electron (ADR) TOELY Veeco Instruments VECO CVD Equipment CVV Intevac IVAC
8

Semiconductor Equipment: Inspection & Metrology

High

How It Works

Inspection and metrology tools find defects and measure critical dimensions on chips during manufacturing. At 3nm and below, a single particle or dimensional error can kill a chip. These tools use electron beams, optical imaging, and X-rays to verify that every layer of every chip meets spec. Yield depends entirely on catching defects early.

Supply / Demand Dynamics

  • AI demand driver: Strong. AI chips are large, complex, and expensive — yield is critical. A single NVIDIA Blackwell die costs thousands; undetected defects are hugely costly. Advanced packaging adds new inspection requirements.
  • Supply constrained: Moderate. KLA dominates with 50%+ market share. Backlog is healthy but these tools are less bottlenecked than litho.

Key US-Listed Companies

KLA Corp KLAC Onto Innovation ONTO Nova Ltd NVMI Camtek CAMT Bruker BRKR Nordson NDSN
9

Semiconductor Equipment: Test (ATE)

Medium

How It Works

Automatic Test Equipment (ATE) validates that finished chips work correctly before they ship. Testers apply electrical signals to chips and check responses against expected patterns. AI accelerators, with their thousands of I/O pins and high-speed interfaces, require sophisticated and expensive test solutions. Testing also ensures HBM stacks and advanced packages meet specs.

Supply / Demand Dynamics

  • AI demand driver: Moderate. More AI chips = more test time needed. But test is a smaller fraction of total chip cost than fabrication equipment, and test capacity scales more linearly.
  • Supply constrained: Mild. Two companies (Teradyne and Advantest) dominate. Capacity additions are more incremental than building new fabs.

Key US-Listed Companies

Teradyne TER Advantest (ADR) ATEYY Cohu COHU FormFactor FORM PDF Solutions PDFS
10

EDA & Design Software

High

How It Works

Electronic Design Automation (EDA) tools are the software used to design, simulate, and verify chips before they go to manufacturing. No chip — including every AI accelerator — can be built without EDA. These tools handle logic synthesis, place-and-route, timing analysis, power optimization, and physical verification. The EDA market is an effective duopoly (Synopsys + Cadence) with Siemens EDA a distant third.

Supply / Demand Dynamics

  • AI demand driver: Strong and compounding. More AI chip designs = more EDA licenses. Every hyperscaler designing custom AI silicon (Google, Amazon, Meta, Microsoft) is a new EDA customer. AI is also being used inside EDA tools themselves, creating a virtuous cycle.
  • Supply constrained: Not in the physical sense, but intellectual moats are enormous. It takes decades to build competitive EDA software. Switching costs are near-infinite. This is a toll-booth business on all semiconductor innovation.

Key US-Listed Companies

Synopsys SNPS Cadence Design CDNS Ansys ANSS PDF Solutions PDFS
11

Semiconductor Materials (Wafers, Photoresists, Gases)

Medium

How It Works

Semiconductor manufacturing consumes highly pure raw materials: silicon wafers (the substrate), photoresists (light-sensitive chemicals for lithography), specialty gases (for deposition and etching), CMP slurries (for polishing), and various wet chemicals. EUV lithography requires entirely new photoresist chemistries. Materials quality directly determines chip yield.

Supply / Demand Dynamics

  • AI demand driver: Moderate. More wafer starts = more materials consumed, but this is a derivative play. Materials companies have lower margins and less pricing power than equipment or design companies.
  • Supply constrained: Selectively. EUV photoresists and ultra-pure specialty gases have limited suppliers. Silicon wafer supply has been tight but is normalizing. Most materials companies are Japanese-dominated (JSR, Shin-Etsu, SUMCO), limiting US-listed options.

Key US-Listed Companies

Entegris ENTG DuPont (electronics) DD Air Products APD Linde LIN Cabot Microelectronics CMP Photronics PLAB Shin-Etsu (OTC ADR) SHECY
12

Analog & Mixed-Signal Semiconductors

Low

How It Works

Analog chips convert real-world signals (voltage, current, temperature, light) into digital data and vice versa. Mixed-signal chips combine analog and digital functions. They handle power management, signal conditioning, sensor interfaces, and data conversion. Every electronic system needs analog chips, but they are not on the critical path for AI compute itself.

Supply / Demand Dynamics

  • AI demand driver: Indirect. AI data centers need power management ICs, voltage regulators, and high-speed data converters. But analog chips are a small fraction of AI server BOM cost. The bigger growth driver for analog remains automotive, industrial, and IoT.
  • Supply constrained: No. Analog uses mature process nodes (28nm+) with ample capacity. The 2021-2023 analog shortage has fully corrected, and inventory levels have normalized.

Key US-Listed Companies

Texas Instruments TXN Analog Devices ADI Monolithic Power MPWR ON Semiconductor ON Skyworks SWKS MaxLinear MXL Renesas (ADR) RNECF Microchip Tech MCHP
13

Networking & Interconnect Chips

High

How It Works

AI training clusters require thousands of GPUs working in parallel, connected by ultra-fast networking. Networking chips handle switching, routing, and serialization/deserialization (SerDes) at 400G/800G speeds. InfiniBand and Ethernet-based fabrics move gradients between GPUs during distributed training. Optical transceivers convert electrical signals to light for data center interconnects.

Supply / Demand Dynamics

  • AI demand driver: Very strong. Networking scales super-linearly with GPU count — a 100K-GPU cluster needs a massive fabric. The shift to 800G and 1.6T Ethernet, plus InfiniBand dominance in training, drives huge switch/transceiver demand.
  • Supply constrained: Moderate. High-speed SerDes and optical components are tight. Custom AI networking ASICs compete for leading-edge fab capacity. NVIDIA (Spectrum-X, ConnectX) has significant share in AI networking.

Key US-Listed Companies

Broadcom AVGO Marvell MRVL NVIDIA (networking) NVDA Arista Networks ANET Cisco CSCO Coherent COHR Lumentum LITE II-VI / Coherent COHR Credo Technology CRDO Astera Labs ALAB Alphawave Semi AWSE
14

Advanced Packaging (CoWoS, Chiplets, 2.5D/3D)

Medium

How It Works

Advanced packaging integrates multiple chiplets, HBM stacks, and interposers into a single package using technologies like CoWoS (Chip-on-Wafer-on-Substrate), InFO, and hybrid bonding. This is how NVIDIA's Blackwell connects two GPU dies with HBM stacks. It extends Moore's Law by combining separately manufactured components rather than shrinking transistors further.

Supply / Demand Dynamics

  • AI demand driver: Critical. CoWoS is literally the tightest bottleneck for AI GPU production today. Every B200/B300 and competitor chip needs advanced packaging. Demand far exceeds supply.
  • Supply constrained: Extremely. TSMC's CoWoS capacity is the #1 supply constraint in the AI chip industry. They are tripling capacity but still cannot keep up. Most packaging companies are not US-listed, which limits investable options.

Key US-Listed Companies

TSMC (CoWoS) TSM Amkor Technology AMKR Kulicke & Soffa KLIC Onto Innovation ONTO Besi (ADR) BESI Camtek CAMT
15

Power Semiconductors (SiC & GaN)

Medium

How It Works

Power semiconductors manage electrical energy: converting, regulating, and distributing power. Silicon carbide (SiC) and gallium nitride (GaN) are wide-bandgap materials that handle higher voltages and temperatures more efficiently than traditional silicon. AI data centers consume enormous power (50-100MW per facility), and every rack needs efficient power delivery from grid to chip.

Supply / Demand Dynamics

  • AI demand driver: Moderate but growing. AI data center power density is exploding (30-120kW per rack). Efficient power conversion saves millions in electricity costs. GaN is gaining traction in server power supplies. SiC is more EV-focused but benefits from grid infrastructure buildout for data centers.
  • Supply constrained: SiC substrate supply is tight (dominated by Wolfspeed and Coherent). GaN capacity is scaling rapidly. Overall, the bottleneck here is less severe than in compute or memory.

Key US-Listed Companies

Wolfspeed WOLF ON Semiconductor ON Monolithic Power MPWR GaN Systems (acq. by Infineon) IFNNY Navitas Semi NVTS Vishay Intertechnology VSH Power Integrations POWI MACOM Technology MTSI

Methodology & Rating Key

Each sub-sector is rated on its direct exposure to AGI-driven demand over the next 2-3 years, considering: (1) how directly the sub-sector's products enable AI compute scaling, (2) supply constraint severity and pricing power, (3) concentration of demand among AI-specific customers, and (4) structural barriers to supply response.

HIGH Direct, large, and sustained demand from AGI buildout. Supply constrained or oligopoly structure gives pricing power. Revenue growth driven primarily by AI. MEDIUM Real AI tailwind but either indirect (derivative demand), smaller in magnitude, or diluted by larger non-AI end markets. Benefits are real but not transformative to revenue mix. LOW Tangential AI exposure. Products are consumed in AI systems but are a small fraction of BOM, face ample supply, and have limited pricing power tied to AI specifically.

Company lists aim for exhaustive US-listed coverage. Some companies appear in multiple sectors (e.g., NVDA in GPU Accelerators and Networking). Tickers are as of May 2026. OTC ADRs are included where they are the primary US trading vehicle for globally significant companies. This report is for research purposes and does not constitute investment advice.