AI Chips

AI Semiconductor Segments

GPUs & Accelerators

Companies designing AI training and inference chips — GPUs, custom ASICs, and accelerator platforms powering data centers worldwide.

NVIDIA logo

NVIDIA

Dominates with H100/B200 GPUs — ~80% of the AI accelerator market.

AMD logo

AMD

Challenges NVIDIA with MI300X/MI350 Instinct accelerators for data centers.

Intel logo

Intel

Competes with Gaudi 3 accelerators and upcoming Falcon Shores.

Broadcom logo

Broadcom

Designs custom AI ASICs (Google TPU, Meta MTIA) for hyperscalers.

Qualcomm logo

Qualcomm

Leads edge AI inference with Snapdragon and Cloud AI 100.

Foundries

The fabrication giants manufacturing AI chips at cutting-edge process nodes — a critical bottleneck in the AI supply chain.

TSMC logo

TSMC

Manufactures ~90% of leading-edge AI chips at 3nm/2nm nodes.

Samsung Electronics logo

Samsung Electronics

Second-largest foundry with 3nm GAA technology and HBM production.

Intel logo

Intel

Building Intel Foundry Services for external customers at Intel 18A node.

Memory (HBM)

High Bandwidth Memory is the critical bottleneck for AI accelerators — these three companies control the entire HBM market.

SK Hynix logo

SK Hynix

Market leader in HBM — supplies majority of NVIDIA's HBM3E chips.

Samsung Electronics logo

Samsung Electronics

Second-largest HBM producer, ramping HBM3E capacity aggressively.

Micron Technology logo

Micron Technology

Only US memory maker — rapidly expanding HBM3E production.

Equipment

The picks-and-shovels of the AI chip boom — companies making the machines that make the chips.

ASML Holding logo

ASML Holding

Monopoly on EUV lithography — every advanced AI chip requires ASML machines.

Applied Materials logo

Applied Materials

Largest equipment maker — deposition and etch tools in every fab.

Lam Research logo

Lam Research

Dominates plasma etch — critical for advanced chip patterning.

AI Chip Supply Chain: Who Does What

Where each company plays across the AI chip stack — from design (NVIDIA, AMD, Apple), to foundry (TSMC, Samsung), to HBM memory (SK Hynix, Micron), to advanced packaging (TSMC CoWoS), to networking (Broadcom, Marvell). Hover any cell for details.
Chip Design
Fabrication
Memory / HBM
Adv. Packaging
Equipment
Networking
Chip Design
Fabrication
Memory / HBM
Adv. Packaging
Equipment
Networking
NVIDIA logoNVIDIA
AMD logoAMD
Broadcom logoBroadcom
Intel logoIntel
Qualcomm logoQualcomm
TSMC logoTSMC
Samsung logoSamsung
SK Hynix logoSK Hynix
Micron logoMicron
ASML logoASML
Applied Materials logoApplied Materials
Lam Research logoLam Research

Solid bars = active in this stage. Faded = notable relationship but not primary business. Hover for details.

AI Accelerator Market Share

Revenue share of the AI accelerator market (the chips that train and run AI models) in 2024. NVIDIA holds roughly 90%; AMD's MI300 is the only credible alternative; Google's TPU (Tensor Processing Unit) is internal-only and not counted in merchant share. Sources: Mercury Research, industry estimates.
NVIDIA88%
AMD8%
Intel1%
Others3%

AI accelerator market share by revenue, 2024. Sources: industry estimates, Mercury Research.

Data Center Revenue by Company

Quarterly revenue from each chipmaker's data center segment — the AI-driven business line. NVIDIA's data center revenue surpassed Intel's entire company revenue in 2024; AMD's data center segment includes both EPYC CPUs and MI300 GPUs. Sources: company earnings reports.

Quarterly data center segment revenue in $ billions. Sources: company earnings reports (NVIDIA fiscal calendar shifted to align with calendar quarters).

AI Revenue as % of Total Revenue

AI-related revenue as a share of each company's total revenue — a quick read on AI exposure. NVIDIA is approaching 90% AI-dependent; TSMC and SK Hynix are mid-range; broader semi companies like Texas Instruments are still mostly non-AI.

AI-related revenue as a percentage of total company revenue. Higher = more dependent on AI demand. Sources: company earnings reports, analyst estimates.

Semiconductor Capex: The Fab Investment Race

Annual capex (capital expenditure on new fabs and equipment) at TSMC, Samsung, and Intel. Each leading-edge fab costs $10–$30B; this chart shows the combined trillion-dollar bet that AI chip demand is durable.

Annual capital expenditure in $ billions. Samsung figures include total semiconductor division capex (memory + foundry). 2025E = estimates. Sources: company annual reports, earnings calls.

Ask Sterling

Register for a premium account to gain access to Sterling AI.

Get Started

Things you can ask Sterling:

Summarize Tesla's latest earnings reportWhy did NVIDIA's margins expand?Compare Apple vs Microsoft's cash flowWhat's driving the EV sector growth?
AI Chips - Reports | Sterling