How OpenAI Makes Money

From research lab to commercial engine

Sam Altman, Elon Musk, Greg Brockman, Ilya Sutskever and several other technologists founded OpenAI in December 2015 with a pledge of roughly $1 billion in funding commitments. The founders set out to research artificial general intelligence in a way that would broadly benefit humanity rather than concentrate power in a single corporation. Musk left the board in 2018 over strategic disagreements, and the organization increasingly needed capital far beyond what donations could supply. 1

That capital gap forced a structural change. In 2019, OpenAI created a "capped-profit" subsidiary, OpenAI LP, that could raise investment while still reporting to the original non-profit board. Microsoft became the anchor investor that year with a $1 billion commitment, later expanding its stake to roughly $13 billion across subsequent funding rounds. The arrangement gave OpenAI the computing budget needed to train ever-larger models while keeping a nominal mission-driven governance layer in place.

OpenAI's mission statement remains "to ensure that artificial general intelligence benefits all of humanity."

The release of ChatGPT in November 2022 changed the company's trajectory entirely. What began as a research demo attracted a million users within five days and turned OpenAI into a household name. By 2025, ChatGPT had grown past 900 million weekly active users, and OpenAI's annualized revenue reportedly reached roughly $20 billion by mid-2026. 2

The Microsoft relationship

Microsoft's investment bought more than equity exposure. OpenAI's models run on Microsoft's Azure cloud infrastructure, and Microsoft has embedded OpenAI's technology into Copilot across its productivity suite, Bing search and GitHub. The companies restructured their commercial agreement in 2026, adjusting revenue-sharing terms as OpenAI worked toward converting its capped-profit entity into a full public benefit corporation. 3

That dependency cuts both ways. Microsoft gains a stake in what could become one of the most valuable technology companies ever built, while OpenAI gains reliable access to the specialized chips and data centers that large language model training requires. Neither party has suggested the arrangement is close to unwinding, even as OpenAI diversifies its compute suppliers to include Oracle and other cloud providers.

Betting big on compute and a possible IPO

Training and running frontier AI models costs enormous sums, and OpenAI's 2025 financials illustrate the gap: roughly $13 billion in revenue against an estimated $34 billion in total costs, most of it computing infrastructure. To close that gap, OpenAI raised a $122 billion funding round that valued the company at $852 billion. 4

In June 2026, OpenAI filed a confidential registration statement with the U.S. Securities and Exchange Commission, a step widely read as preparation for a public offering that could value the company as high as $1 trillion. Sam Altman has framed the spending as necessary to stay ahead of rivals racing toward more capable models, though critics note the company has never posted an annual profit. The pending listing would mark an unusual moment for an organization that began life as a non-profit committed to open research.

Widening the product lineup

OpenAI's early releases, GPT-3, DALL-E and Codex, established the company as a research-driven lab publishing frequent papers and open tools. Codex eventually evolved into GitHub Copilot's underlying engine through the Microsoft partnership, showing how research output could be repackaged into commercial products. The Dactyl robotic hand project, meanwhile, demonstrated OpenAI's reinforcement-learning work outside of language models, even though robotics never became a primary revenue source.

Today's product lineup is broader: the GPT-4 and GPT-5 model families, DALL-E for image generation, Sora for video generation, and enterprise tools sold through the API and ChatGPT Enterprise tier. OpenAI also operates a nonprofit-funded safety research arm that studies alignment and misuse risks separately from its commercial teams. The company continues to publish some research openly, though it has scaled back full model-weight releases compared with its earliest years, citing safety concerns.

Competing in a crowded field

Google DeepMind, backed by Alphabet's balance sheet and years of reinforcement-learning research, remains OpenAI's most resourced rival. Anthropic, founded by former OpenAI researchers including Dario and Daniela Amodei, has built a comparable enterprise business around its Claude model family and safety-focused positioning. Meta AI takes a different approach, releasing openly licensed Llama models that undercut the commercial pricing both OpenAI and Anthropic depend on. 5

Governance and safety scrutiny

OpenAI's non-profit board briefly ousted Sam Altman as chief executive in November 2023, only to reinstate him days later after employees threatened to resign en masse and Microsoft signaled support for his return. The episode exposed how uneasy the balance remains between OpenAI's mission-driven governance and its commercial ambitions. Regulators in the European Union, the United States and elsewhere have since pressed the company on data use, copyright and model safety testing, adding compliance costs to an already expensive business. 6

Key Partners

Microsoft anchors OpenAI's partner network as both its largest investor and primary cloud infrastructure provider through Azure. Chip suppliers including Nvidia and, more recently, custom silicon partnerships supply the processing power that model training requires. Universities and research institutions collaborate on safety and alignment research, while enterprise partners across finance, software and retail integrate OpenAI's API into their own products.

Key Activities

OpenAI's core activity is training and refining large language models, which requires continuous research into model architecture, data curation and reinforcement learning from human feedback. The company also builds and maintains consumer-facing products, chiefly ChatGPT, and developer-facing infrastructure like the API and fine-tuning tools. Safety research runs alongside product development, with dedicated teams studying misuse risks and alignment before major model releases.

Key Resources

OpenAI's most valuable resource is its trained model weights, the product of years of research and enormous computing spend that competitors cannot easily replicate. Access to specialized chips and data center capacity, largely through the Microsoft Azure partnership, ranks close behind. The company's research talent, drawn from top academic and industry labs, and its accumulated training data pipelines round out its key resources.

Value Propositions

For developers and enterprises, OpenAI offers API access to state-of-the-art language and image models without the cost of building comparable systems in-house. Consumers get a general-purpose assistant in ChatGPT that handles writing, coding, research and everyday questions through a simple chat interface. Businesses using ChatGPT Enterprise gain data privacy guarantees and administrative controls suited to corporate deployment. Researchers benefit from published papers and, in some cases, open tools that extend OpenAI's work. Governments and NGOs gain access to AI capabilities they could not otherwise afford to develop independently.

Customer Relationships

OpenAI maintains self-service relationships with millions of ChatGPT subscribers who sign up online and manage billing without human contact. Developers access support through documentation, community forums and a tiered support system tied to API usage volume. Enterprise and government customers work with dedicated account teams that negotiate custom contracts and data-handling terms. The company also engages the public through its blog and safety publications to shape perceptions of responsible AI development.

Channels

The ChatGPT website and mobile apps serve as OpenAI's primary consumer channel, reaching hundreds of millions of users directly. The API serves as the main developer and enterprise channel, letting third parties embed OpenAI's models into their own software. Microsoft's product suite, including Copilot and Bing, acts as an indirect distribution channel that reaches users who never visit OpenAI's own properties. The company's blog, research publications and conference appearances support awareness and recruiting.

Customer Segments

Individual consumers who pay for ChatGPT Plus or use the free tier make up OpenAI's largest user base by volume. Software developers and enterprises building applications on the API represent the segment generating the most predictable recurring revenue. Large corporations purchasing ChatGPT Enterprise or custom API agreements contribute high-value contracts. Governments, NGOs and research institutions form a smaller but strategically important segment tied to OpenAI's stated mission.

Cost Structure

Computing infrastructure, including chip procurement and data center capacity, represents OpenAI's largest and fastest-growing cost, reportedly reaching tens of billions of dollars annually. Salaries for research and engineering talent make up the second-largest expense category, reflecting intense competition for AI specialists. Legal, safety compliance and administrative costs round out the structure as the company navigates regulatory scrutiny in multiple jurisdictions.

Revenue Streams

ChatGPT subscriptions across the Plus, Team, Pro and Enterprise tiers generate recurring consumer and business revenue. API usage fees, charged per token processed, form the backbone of OpenAI's developer and enterprise business. Licensing and commercial partnerships, including the Microsoft agreement, contribute additional revenue tied to how OpenAI's technology gets embedded into third-party products.

• 1OpenAI's founding history and early donors
• 2ChatGPT's rapid early adoption
• 3Details of the Microsoft-OpenAI partnership restructuring
• 4OpenAI's funding round and valuation figures
• 5Overview of Anthropic's founding and business
• 6Account of Sam Altman's brief ouster and reinstatement