AI companies

Research and product lab behind ChatGPT, GPT-5, DALL-E, Sora, Codex, and the OpenAI API. Founded by Sam Altman, Greg Brockman, Ilya Sutskever, and others; deeply capitalised by Microsoft. Sets the pace on frontier LLMs for most of the industry.

Alphabet's AI portfolio spans DeepMind (Gemini, Gemma, Veo, Imagen), Google Cloud (Vertex AI, NotebookLM), and Search/Workspace integrations. One of the two or three most-resourced AI programs in the world; publishes broadly and open-sources its Gemma model line.

Largest OpenAI shareholder and the distribution channel for GPT-5 into enterprise via Copilot and Azure OpenAI. Also ships its own Phi small-model family, VS Code with GitHub Copilot, and the Bing/Copilot search stack.

Runs the largest open-weight LLM program in the world through the Llama series (3.1, 4). Also ships Movie Gen, Segment Anything, Emu, and consumer-facing Meta AI assistants across Instagram/WhatsApp/Facebook. Yann LeCun leads long-horizon research.

Enterprise AI reach through AWS Bedrock (multi-vendor model hosting) and the Nova family (Nova Pro, Nova Lite). Backer of Anthropic. Historically a compute + hosting play more than a frontier-model program.

Ships the Qwen open-weight family — Qwen 2.5 72B, Qwen 2.5 Coder 32B — under a permissive licence. Strong bilingual EN/ZH and best-in-class small-model code performance.

Safety-forward AI lab founded by former OpenAI researchers Dario and Daniela Amodei. Ships the Claude family of models (Opus, Sonnet, Haiku) and the Claude Code developer tool. Constitutional-AI training and interpretability research are its signature technical bets.

Maker of Cursor, the AI-first fork of VS Code. Model-agnostic (routes across Claude, GPT-5, Gemini). Currently one of the most-used developer AI tools.

Consumer design incumbent. Magic Studio bundles image/text/video AI features across the templates workflow. Enormous distribution moat via existing user base.

Originators of Stable Diffusion. Post-restructuring focus on image + audio open weights (SD3, Stable Audio). Community fine-tunes remain a major driver of image-gen.

Ships Doubao (consumer chat) and Seedream (image gen) built on the same foundation-model stack that powers TikTok/Douyin AI features. Dominant in mainland Chinese consumer AI.

Voice-synthesis and voice-agent platform. Ships text-to-speech, voice cloning, dubbing, and Conversational AI. Widely adopted for content creation and voice-first apps.

Code-hosting incumbent (Microsoft-owned). Ships Copilot (Chat + IDE + PR reviews + Workspace) and the broader Copilot ecosystem. Deepest developer distribution in the industry.

Text-to-video pioneer. Ships Gen-3 Alpha, Gen-4, and the Aleph editing product. Major creative-industry adoption.

Kai-Fu Lee's frontier lab. Ships the Yi family (open-weight Yi-1.5 6B/9B/34B, closed Yi Lightning). Bilingual EN/ZH; Yi Lightning briefly topped LMSys Chatbot Arena.

Creative-tools incumbent. Ships Firefly (image/video gen), Generative Fill in Photoshop, and generative video in Premiere. Commercial-safe training data is a differentiator.

On-device generative AI baked into iOS/macOS via Apple Intelligence. Uses a hybrid of small on-device models and cloud-fallback (partnering with OpenAI). Focus is privacy-first rather than frontier benchmarks.

Ships the Ernie family and Wenxin chat. Historically the leading Chinese AI player before DeepSeek/Doubao caught up. Deep integration with Baidu Search grounding.

Enterprise-focused frontier lab. Command R+ is the flagship, purpose-built for RAG and tool-use. Strong compliance posture (SOC 2, HIPAA) is its enterprise wedge.

Data lakehouse vendor. Ships Mosaic AI (fine-tuning + serving + evals), Vector Search, and the DBRX open-weight model. Enterprise integration is the wedge.

Chinese frontier lab that broke the training-cost assumption with DeepSeek V3 and the open reasoning model R1. Publishes weights, papers, and API access at prices 50-100× cheaper than Western peers.

Collaborative design tool. Ships Figma AI (design generation, design-to-code) and the Weave beta for creative workflows. Design-team standard.

The GitHub of open models. Hub hosts hundreds of thousands of model weights + datasets; also ships Inference Endpoints, AutoTrain, Spaces (hosted demos), and the transformers/diffusers libraries.

Watson-era enterprise AI reborn as watsonx — a hosted, governance-heavy platform for regulated industries. Ships Granite open-weight models alongside a multi-vendor model catalog.

Text-to-image product with an aesthetic house-style. Historically Discord-first; now a full web app. Bootstrapped and highly profitable.

European (French) frontier lab founded by former Meta and Google DeepMind researchers. Ships Mistral Large, Mixtral (MoE), and Mistral Nemo. La Plateforme API offers EU-data-residency options attractive to GDPR-heavy buyers.

Chinese frontier lab. Ships the Kimi K2 1T-parameter open-weight MoE model. Long-context strengths.

Open-source workflow automation. Fair-code licence + self-hosted-by-default. Native AI/agent nodes make it one of the leading LangChain-adjacent orchestration tools.

Docs/wiki incumbent. Ships Notion AI (writing assist, Q&A over workspace) and Notion Calendar. Distribution via existing team-workspace users.

AI search product. Uses a routing layer over multiple frontier LLMs plus proprietary web-index and citation logic. Consumer + enterprise search wedges.

Cloud IDE + hosting. Ships Replit Agent (prompt-to-app), Replit Ghostwriter, and the broader Bounties + Deployments platform. Consumer/prosumer developer wedge.

Data-cloud vendor with Cortex — SQL-native LLM and vector operations run directly against warehouse data. Also ships the open Arctic MoE model.

Text-to-music generation. Consumer creative product with a strong prosumer/hobbyist community. Rapid model iteration cadence.

Text-to-music generation. Direct competitor to Suno; different audio-quality trade-offs and licensing posture.

Elon Musk's AI company. Ships the Grok family with tight X (Twitter) integration for real-time information. Positions itself as less-restricted / more-direct-tone alternative to the frontier incumbents.

Spin-out from Tsinghua University. Ships the GLM family (GLM-4, GLM-4-9B open weights) and ChatGLM consumer chat. Strong Chinese quality with vision variants.

Israeli frontier lab. Ships the Jamba hybrid Mamba-Transformer models with dramatically faster long-context throughput. Enterprise-focused positioning.

Jira/Confluence maker. Ships Atlassian Intelligence and Rovo AI agents across the team-collaboration stack.

SMB CRM/marketing incumbent. Breeze AI ships across the platform. Strong SMB distribution.

Visual workflow automation (formerly Integromat). AI actions across a 1500+ app catalog. Popular with non-developers.

GPU monopoly-adjacent hardware vendor. Also ships open-weight models (Nemotron), inference frameworks (TensorRT-LLM, Triton), and NIM microservices. Every AI company is downstream of Nvidia in some way.

CRM incumbent. Ships Agentforce, Einstein, and CRM-embedded AI. Enterprise distribution + org-graph as the moat.

Frontend hosting + Next.js maker. Ships v0 (design-to-code), the AI SDK, and Vercel AI Gateway. Developer distribution via Next.js.

Workflow-automation incumbent. Now a first-class AI-workflow platform (Zapier Agents, AI actions). 7000+ integration catalog is the moat.