trackaistocks.com
Get alerts
Home / Chips & Compute

AI chip stocks

Chips & Compute

GPUs, accelerators, CPUs, foundries

The processors that train and run AI models, and the factories that make them. Demand for AI compute starts here; everything else in the stack exists to feed, connect, power and house these chips.

CompanyTicker
NVIDIA
ChipsNetworking
Designs the GPUs and systems most AI models are trained and run on.
NVDA
Advanced Micro Devices
Chips
Instinct accelerators and EPYC server CPUs; the main GPU alternative to NVIDIA.
AMD
Broadcom
ChipsNetworking
Custom AI accelerators for hyperscalers and the Ethernet switch silicon that connects them.
AVGO
Taiwan Semiconductor (ADR)
Chips
Manufactures nearly all leading-edge AI chips and does the advanced packaging (CoWoS) they need.
TSM
Arm Holdings (ADR)
Chips
Licenses the CPU architecture used in many AI servers and every smartphone.
ARM
Marvell Technology
ChipsNetworking
Custom silicon for cloud customers and the optical/interconnect chips inside AI data centers.
MRVL
Intel
Chips
Xeon server CPUs, Gaudi accelerators, and a foundry business seeking AI customers.
INTC
Qualcomm
Chips
On-device AI in phones and PCs through Snapdragon processors.
QCOM
Microsoft
HyperscalersSoftwareChips
Primary: Hyperscalers & Cloud
Azure AI infrastructure, the OpenAI partnership, and Copilot across its software.
MSFT
Amazon
HyperscalersChips
Primary: Hyperscalers & Cloud
AWS rents AI compute and builds its own Trainium and Inferentia chips.
AMZN
Alphabet
HyperscalersChipsSoftware
Primary: Hyperscalers & Cloud
Gemini models, Google Cloud AI services, and in-house TPU accelerators.
GOOGL
Meta Platforms
HyperscalersChips
Primary: Hyperscalers & Cloud
Among the largest AI capital spenders; trains Llama models for its own products.
META
Super Micro Computer
Data centersChips
Primary: Data Centers
Builds GPU servers and racks, including liquid-cooled systems.
SMCI

Companies whose primary category is elsewhere but that also belong here are listed with their primary tag.

How chips companies earn from AI

This is where AI spending turns into revenue first. A hyperscaler or an AI lab decides how many accelerators to buy; NVIDIA and AMD design them, Broadcom and Marvell design custom versions for single customers, TSMC manufactures nearly all of them and does the packaging that joins each chip to its memory, and Arm, Intel and Qualcomm supply the CPUs that sit beside the accelerators or run smaller models on devices. Prices are high and volumes are set by how fast data centers can be built, so revenue in this layer moves in large steps rather than smooth curves.

The companies differ in how much of their business is AI. NVIDIA's data-center line is over 90% of revenue. Broadcom and AMD report AI or data-center revenue as a majority. TSMC reports the share of its revenue from high-performance computing. Intel and Qualcomm report nothing labelled AI at all, which is why their pages rate exposure as partial. Read the reported figure on each company page before assuming a chip company is an AI company.

Editorial, written from the reported figures on each company page. Company names link to those pages: NVDA, AMD, AVGO, TSM, ARM, MRVL, INTC, QCOM.

The numbers right now

Reported by the companies; each line links to the release or filing it comes from.

What to watch in chips

  • NVIDIA's Q3 FY2027 report against the $108.0 billion outlook, which assumes no China data-center compute revenue. source
  • Whether custom accelerators (Broadcom, Marvell) keep growing faster than merchant GPUs; the AI revenue guide at Broadcom is the cleanest quarterly read. source
  • TSMC's monthly revenue releases and any change to the US$60-64 billion capital budget, which set the ceiling on how many chips everyone else can ship. source

Quotes

Quotes by TradingView

Compare

Charts by TradingView

Latest 10-K, 10-Q and 8-K filings in Chips & Compute

Source: SEC EDGAR