The AI Stack
Sign in

OpenAI

AI research organization developing large language models like GPT series, ChatGPT, DALL-E, and Sora.

Updated May 2026

Overview

Website
openai.com
Founded
2015
Headquarters
San Francisco, CA
Ownership
Private
Segment
Frontier Foundation Model Labs

Product overview

OpenAI builds frontier AI models including the GPT family of large language models, multimodal tools like DALL-E for image generation, Whisper for speech recognition, Sora for text-to-video, and consumer products like ChatGPT. These are used by individual consumers, developers via API, and enterprises for tasks like content generation, coding assistance, customer support, and research. It stands out through its hybrid nonprofit/for-profit structure, massive scale backed by Microsoft, and focus on deploying safe AGI while catalyzing generative AI mainstream adoption via ChatGPT.

Revenue model

Freemium model with free tier; consumer subscriptions (Plus $20-23/mo, Pro $200-229/mo, Go $8/mo); team/business $25-30/user/mo; custom enterprise licensing; API usage-based pricing per token (e.g. GPT-4o $2.50/$10 per 1M input/output tokens). 2025 ARR ~$12-13B from subscriptions (majority), API, and enterprise.

Moat

OpenAI's primary competitive moat is its large user base of 8-900 million users, which provides distribution gravity for developers and enterprise adoption; however, this advantage is fragile due to shallow engagement, lack of daily habit formation, and absence of network effects or switching costs. The company is attempting to build durable moats through vertical data partnerships and value-based enterprise pricing models, but currently lacks proprietary technology, a clear technological lead, or a self-reinforcing platform ecosystem.

Active layers

Stack lineage

What powers OpenAI

L5Orchestration
  • Temporal

    OpenAI uses Temporal as a reliable foundation layer underneath long-running, mission-critical applications.

L3Data
  • Pinecone

    Listed as a Pinecone customer building AI-powered applications.

L1Silicon
  • Broadcom

    OpenAI deploys multiplanar network designs using NVIDIA Spectrum-X Ethernet in conjunction with MRC, leveraging the NVIDIA Spectrum-X Multiplane capability for hardware-accelerated load balancing.

  • NVIDIA

    We are finalizing an investment and partnership agreement with OpenAI. There is no assurance that we will enter into an investment and partnership agreement with OpenAI or that a transaction will be completed.

What OpenAI powers

L4Models
  • Together AI

    Together AI hosts OpenAI's models — e.g. openai/gpt-oss-120b appears in their public model catalog.

  • OpenRouter

    OpenRouter hosts OpenAI's models — e.g. openai/gpt-chat-latest appears in their public model catalog.