Companies.

Anthropic

Anthropic – AI research: Claude reasoning models and enterprise alignment solutions.

OpenAI

OpenAI – Frontier models: ChatGPT consumer distribution and developer API platform.

xAI

xAI – Frontier models: Memphis Colossus computing cluster scale-out and Grok assistant.

Google

Google – Full-stack player: Gemini models, Google DeepMind research, and custom TPUs.

Meta

Meta – AI research: Llama open weights ecosystem and Meta Superintelligence Labs.

Amazon

Amazon – Cloud AI: AWS cloud hosting, custom Trainium and Inferentia silicon.

Apple

Apple – Consumer distribution: On-device private machine learning and Siri integration.

Microsoft, AI distribution: Azure cloud, the OpenAI partnership, and Copilot woven across Windows, Office, and GitHub.

Oracle, cloud plus SaaS: OCI cloud infrastructure scaling into AI compute, on top of a sticky database, ERP, and applications franchise.

Netflix

Netflix – Consumer personalization: Recommendation systems, localized translation pipelines, and content previsualization.

NVIDIA

NVIDIA – AI compute infrastructure: Custom graphics processing units, CUDA platform, and NVLink interconnects.

AMD, the second accelerator: EPYC server CPUs and the Instinct MI-series GPUs as the credible alternative to NVIDIA.

Cerebras

Cerebras – Alternative accelerators: Wafer-scale engine (WSE) AI compute systems and SRAM-based inference architectures.

SpaceX

SpaceX is becoming an AI-infrastructure company: reusable launch (Starship) drives $/kg-to-orbit toward the point where space-based compute pencils out, and the announced AI1 satellite is the first spacecraft designed as a data center.

CoreWeave, the GPU cloud: a specialized neocloud renting NVIDIA capacity (Hopper, Blackwell) to AI labs and enterprises.

Nebius, the European GPU cloud: a neocloud (spun out of the former Yandex assets) renting NVIDIA capacity to AI customers.

IREN, power-first compute: a data-center operator (rooted in bitcoin mining) converting owned, energized sites into AI and HPC GPU-cloud capacity.

Broadcom

Broadcom – Custom silicon & fabric: Network switching systems, optical transceivers, and hyperscaler accelerator co-design.

Arista, the AI back-end fabric: high-performance Ethernet switching that scales out GPU clusters.

Marvell, custom silicon and optics: hyperscaler custom-accelerator co-design, electro-optics, and data-center networking.

Astera Labs, connectivity silicon: PCIe and CXL retimers, smart cable modules, and fabric switches that move data inside and between AI servers.

Vertiv, data-center power and cooling: power distribution, busways, UPS, and the liquid cooling that high-density AI racks now require.

Eaton, electrical infrastructure: switchgear, busways, and power-management gear for data centers, plus electrification across industry, aerospace, and vehicles.

GE Vernova, electricity supply: gas turbines, grid equipment, and wind, the generation and transmission side of the AI power crunch.

Wolfspeed

Wolfspeed – Power semiconductors: Silicon carbide (SiC) transistors and solid-state grid conversion hardware.

Micron

Micron Technology – Memory manufacturing: High Bandwidth Memory (HBM) and enterprise DRAM supply.

SK Hynix

SK Hynix – Memory manufacturing: High Bandwidth Memory (HBM) silicon design and fabrication.

Seagate

Seagate Technology – Storage hardware: Hard disk drive (HDD) systems and HAMR technology for cold data scale.

Western Digital

Western Digital – Storage hardware: Enterprise hard disk drive (HDD) and NAND flash memory systems.

TSMC

TSMC – Semiconductor fabrication: Advanced logic wafer manufacturing and CoWoS packaging.

Samsung Electronics

Samsung Electronics – Semiconductors & memory: Dynamic random-access memory (DRAM), HBM supply, and logic foundry.

Tower Semiconductor

Tower Semiconductor – Specialty foundry: Silicon photonics manufacturing and transceiver chips.

ASML, the lithography chokepoint: the sole supplier of EUV scanners that every leading-edge chip depends on.

Amkor, outsourced packaging: an OSAT that assembles and tests chips, including the advanced packaging AI accelerators depend on.

Lumentum

Lumentum – Optical networking: Indium phosphide lasers and co-packaged optics (CPO) transceiver components.

Coherent

Coherent – Optical materials: Transceiver laser arrays and advanced optical networking materials.