OpenAI owns ChatGPT, with Microsoft holding a minority stake valued at $13 billion as of 2023
OpenAI, founded in 2015, developed and owns ChatGPT outright. The organization operates as a capped-profit entity after transitioning from its original non-profit structure, with Microsoft as its largest investor through a $13 billion commitment announced in January 2023. This partnership gives Microsoft exclusive rights to integrate ChatGPT technology into its products, but OpenAI retains full ownership and operational control of the model.
How OpenAI was structured from the start
OpenAI launched with $1 billion in pledged commitments from its founding group, which included Elon Musk, Sam Altman, and other entrepreneurs focused on AI safety. The founders explicitly chose a non-profit structure to prioritize responsible AI development over profit maximization. This decision reflected their stated concern that advanced AI capabilities could concentrate in too few corporate hands, creating systemic risks.
The non-profit approach shaped OpenAI’s early work in concrete ways. Between 2015 and 2018, the organization published most of its research openly, allowing the global academic community to cite, critique, and build upon their findings. This transparency accelerated progress across the entire AI field. Papers on transformer architectures and language modeling from OpenAI researchers became foundational texts for subsequent development at competing labs.
However, the non-profit model created a practical problem: training advanced AI systems requires tens of millions in annual computing costs. By 2019, OpenAI transitioned to a “capped-profit” structure, where investors can earn returns up to a defined cap, with any excess flowing back to the non-profit parent. This hybrid model allowed the organization to raise capital while maintaining its governance structure and mission alignment.
The technical foundation of ChatGPT
ChatGPT is a large language model built on transformer neural network architecture, specifically a variant called GPT-3.5 and later GPT-4. Transformers process text by weighing the importance of different words simultaneously rather than sequentially, which allows them to capture long-range relationships between concepts in a text passage.
OpenAI trained ChatGPT on approximately 300 billion text tokens drawn from books, websites, academic papers, and other public internet sources. The training used unsupervised learning, meaning the model wasn’t given explicit right answers but instead learned to predict the next word in a sequence billions of times. After this initial training phase, OpenAI used reinforcement learning from human feedback (RLHF), where contractors rated different model outputs and the system was fine-tuned to prefer highly-ranked responses.
This training approach produces responses that sound natural because the model has internalized statistical patterns about how humans write and communicate. When you ask ChatGPT a question, it’s not retrieving stored answers but calculating the most probable next sequence of words based on patterns learned during training. The model can apply these patterns to novel topics it hasn’t explicitly seen, which explains its ability to write code, analyze documents, or generate creative content on demand.
Why the Microsoft partnership matters for ownership structure
Microsoft’s $13 billion investment is the largest capital commitment to any AI company to date, but it doesn’t grant Microsoft ownership of OpenAI. Instead, Microsoft receives a revenue-sharing agreement: when customers use ChatGPT through Microsoft’s platforms—particularly Azure cloud services and Microsoft 365 applications—Microsoft keeps a percentage of the subscription revenue.
The partnership also gives Microsoft exclusive rights to OpenAI’s technology for certain applications. Microsoft integrated GPT-4 into Copilot, its AI assistant embedded across Windows, Office, and Bing. However, OpenAI can still license ChatGPT to other companies and operates its own consumer-facing product at openai.com, where users pay for premium access.
This arrangement preserves OpenAI’s independence while giving Microsoft commercial leverage. For B2B professionals, this matters because it means ChatGPT development decisions flow through OpenAI’s product roadmap, not Microsoft’s priorities. When OpenAI releases new model versions, improvements to context windows, or API price changes, these decisions originate from OpenAI’s leadership.
Practical implications for business users
Understanding ownership clarifies which company is responsible for ChatGPT’s capabilities, limitations, and future direction. OpenAI controls when new features roll out, how the model is trained, what data it learns from, and what safeguards are implemented. If you’re evaluating ChatGPT for business use—whether for customer support automation, document analysis, or code generation—you’re evaluating a product developed and maintained by OpenAI, with Microsoft as a financial partner rather than owner.
If you’re an enterprise customer considering ChatGPT integration through Microsoft Azure, your contract runs with Microsoft, but the underlying model and its updates come from OpenAI. This distinction affects support channels, pricing negotiations, and what roadmap requests you should direct where.
Start by identifying one specific workflow in your organization where ChatGPT or similar models could reduce manual work. Document the current time investment and error rate, then run a pilot using either OpenAI’s API directly or through Microsoft’s Azure OpenAI service. Track whether the model’s actual output quality matches your use case, since ChatGPT performs well on some tasks (summarization, first-draft writing) and poorly on others (complex calculations, factual recall about events after its training data cutoff).
If you have insights about AI implementation in your industry or strategies for evaluating AI tools in practice, LinkedIn Daily is actively seeking contributor submissions. Visit the LinkedIn Daily write-for-us page to pitch a story about how your organization is using AI in production environments.
