AI Chip 13Fs: Nvidia, AMD, Broadcom, Marvell, Astera Inside

US-traded AI chip equities form the most dynamic semiconductor corner of institutional 13F positioning. Nvidia (NVDA), Advanced Micro Devices (AMD), Broadcom (AVGO, networking plus custom ASIC), Marvell Technology (MRVL, custom ASIC plus networking), Astera Labs (ALAB, AI connectivity), plus Credo Technology (CRDO, AI connectivity) anchor the cohort. Multi-year emerging GPU compute scaling, custom ASIC growth, plus emerging emerging hyperscaler capex dynamics drive distinctive institutional positioning. Reading AI chip 13F positioning requires understanding the GPU-vs-ASIC framework plus the multi-year demand dynamics.

The AI chip business model

AI chip companies operate four primary economic engines:

1. GPU compute scaling. Multi-year emerging GPU compute scaling drives Nvidia plus AMD operator economics. Multi-year emerging Nvidia Hopper H100 plus emerging emerging Blackwell B200/GB200 (2025 ramp) plus emerging emerging Rubin (2026 plan) drive multi-year emerging GPU compute generation cycles. Multi-year emerging Nvidia Data Center revenue exceeded $115B in FY2025 plus emerging emerging GPU plus emerging Networking plus emerging emerging Software (CUDA, NIM) plus emerging emerging emerging emerging emerging full-stack positioning drives multi-year emerging emerging operator scaling.
2. Custom ASIC growth. Multi-year emerging custom ASIC growth drives Broadcom plus Marvell operator economics. Multi-year emerging Broadcom Google TPU plus emerging emerging Meta MTIA plus emerging emerging TikTok ASIC plus emerging emerging Apple emerging emerging plus emerging emerging emerging emerging Marvell Amazon Trainium plus Microsoft Maia plus emerging emerging emerging emerging Cisco emerging drive multi-year emerging custom ASIC revenue. Multi-year emerging emerging Broadcom AI revenue ($12B+ FY2024) plus emerging emerging Marvell AI revenue ($1.5B+).
3. Hyperscaler capex dynamics. Multi-year emerging hyperscaler capex dynamics drives AI chip demand. Multi-year emerging Microsoft plus emerging emerging Google plus emerging emerging Amazon plus emerging emerging Meta combined capex reached $400B+ (2025 projected) vs $260B (2024). Multi-year emerging emerging Stargate (OpenAI plus Microsoft plus SoftBank plus emerging emerging Oracle) $500B over 4 years plus emerging emerging emerging emerging xAI Colossus expansion plus emerging emerging Anthropic Project Rainier plus emerging emerging emerging emerging Mistral plus emerging emerging emerging emerging Cohere drive multi-year emerging emerging AI infrastructure capex.
4. AI connectivity emerging. Multi-year emerging AI connectivity emerging drives Astera Labs plus emerging emerging Credo Technology operator economics. Multi-year emerging PCIe Gen 6 plus emerging emerging emerging emerging Compute Express Link (CXL) plus emerging emerging emerging Ultra Ethernet plus emerging emerging emerging InfiniBand plus emerging emerging emerging emerging optical interconnect plus emerging emerging emerging emerging emerging emerging linear pluggable optics (LPO) drive multi-year emerging AI connectivity TAM.

Major US AI chip names

Nvidia (NVDA)

Dominant AI GPU plus emerging emerging Hopper H100 plus emerging emerging Blackwell B200/GB200/GB300 plus emerging emerging Rubin (2026 plan) plus emerging emerging CUDA software plus emerging emerging Networking (Mellanox InfiniBand plus Spectrum-X Ethernet) plus emerging emerging Omniverse plus emerging emerging Drive plus emerging emerging Project Digits plus emerging emerging NIM microservices.

Advanced Micro Devices (AMD)

Diversified Data Center (MI300X plus MI325X plus MI350X plus MI400) plus emerging emerging EPYC server CPU plus emerging emerging Ryzen plus emerging emerging Radeon plus emerging emerging Xilinx FPGA plus emerging emerging Pensando DPU plus emerging emerging Silo AI emerging acquisition.

Broadcom (AVGO)

Diversified networking (Tomahawk, Jericho, Trident switches) plus emerging emerging custom ASIC (Google TPU, Meta MTIA, TikTok plus emerging emerging emerging Apple, plus emerging emerging emerging others) plus emerging emerging VMware (acquired November 2023) plus emerging emerging Symantec plus emerging emerging CA Technologies plus emerging emerging emerging Brocade. Multi-year emerging Hock Tan CEO leadership.

Marvell Technology (MRVL)

Diversified custom ASIC (Amazon Trainium plus Microsoft Maia plus emerging emerging emerging others) plus emerging emerging Ethernet plus emerging emerging optical PHY plus emerging emerging storage controllers plus emerging emerging Aquantia plus emerging emerging Inphi plus emerging emerging Innovium acquisitions. Multi-year emerging Matt Murphy CEO leadership.

Astera Labs (ALAB)

Pure-play AI connectivity plus emerging emerging Aries PCIe retimers plus emerging emerging Taurus Ethernet plus emerging emerging Leo CXL plus emerging emerging Scorpio fabric switches. Multi-year emerging Jitendra Mohan founder-CEO leadership.

Credo Technology (CRDO)

Diversified AI connectivity plus emerging emerging Active Electrical Cables (AECs) plus emerging emerging optical DSPs plus emerging emerging Linear Pluggable Optics (LPO) plus emerging emerging Microchip PCIe switches plus emerging emerging operational scaling. Multi-year emerging Bill Brennan founder-CEO leadership.

How institutional managers position around AI chips

Three patterns appear across smart-money 13Fs:

Pattern 1: GPU-leadership concentration

NVDA-concentrated growth manager positions reflect quality GPU compute leadership thesis.

Pattern 2: Custom-ASIC positioning

AVGO, MRVL-concentrated growth manager positions reflect custom ASIC growth thesis.

Pattern 3: AI-connectivity positioning

ALAB, CRDO-concentrated growth manager positions reflect AI connectivity scaling thesis.

How to read AI chip 13F positioning

Three rules apply:

Rule 1: Identify segment exposure

GPU vs custom ASIC vs connectivity have distinct dynamics.

Rule 2: Watch hyperscaler capex

Multi-year hyperscaler capex drives AI chip demand.

Rule 3: Cross-check generation cycle

Multi-year GPU plus ASIC generation drives operator revenue.

What AI chip positioning signals

1. GPU-leadership conviction. Concentrated NVDA positions signal GPU leadership thesis.
2. Custom-ASIC conviction. Concentrated AVGO, MRVL positions signal custom ASIC thesis.
3. AI-connectivity conviction. Concentrated ALAB, CRDO positions signal AI connectivity thesis.

For real-time tracking of AI chip 13F activity, see the institutional signals feed.