What is OpenAI?

By Jaeden Schafer · 11 min read

What is OpenAI?

OpenAI is the most influential AI research laboratory in the world. It created ChatGPT (the fastest-growing consumer app ever), GPT-4 (arguably the most capable language model available), DALL-E (revolutionary image generation), and Whisper (state-of-the-art speech recognition). Its decisions shape how the entire AI industry evolves.

I’ve integrated OpenAI’s API into AI Box because their models deliver the best results for most use cases. But OpenAI itself is fascinating—it’s a company founded on non-profit ideals that’s become a capitalist powerhouse, with internal drama that reads like a tech thriller. Understanding OpenAI matters if you’re building on AI, investing in AI, or just curious about how the future gets built.

At its core, OpenAI develops large language models and AI tools, distributing them through ChatGPT (consumer), the API (developers), and enterprise channels. The company employs some of the world’s top AI researchers and has fundamentally reshaped what’s possible with AI in less than a decade.

Company History & Evolution

2015 – Idealistic Founding: OpenAI was founded in December 2015 by Sam Altman, Elon Musk, Dario Amodei, and others as a non-profit research lab focused on ensuring AI development benefited humanity. The mission statement emphasized “safe AI”—ensuring advanced AI systems would remain beneficial as they became more powerful. Very idealistic, somewhat naive in hindsight.

2016-2018 – Early Research: OpenAI published groundbreaking research on reinforcement learning and developed GPT (the original Generative Pre-trained Transformer). The company remained small, focused on research rather than commercialization. Elon Musk was on the board and contributed financially.

2018 – Elon’s Exit: Elon Musk left the board in February 2018, citing conflicts with Tesla and SpaceX work. He remained publicly supportive but no longer involved in governance. This was significant—it signaled that OpenAI was becoming more autonomous, and perhaps that Elon’s vision and the company’s direction were diverging. In hindsight, Elon leaving was a watershed moment for OpenAI’s commercial trajectory.

2019 – The Pivot to Commercialization: OpenAI created a “capped-profit” structure—technically a non-profit subsidiary but with a for-profit arm that could return investor capital up to a specific multiple (initially 100x). This was OpenAI’s first real pivot toward commercialization while claiming moral high ground. It was clever structuring: raise capital like a startup, claim non-profit values, but let investors earn substantial returns.

2020-2022 – API & Scaling: OpenAI released the API, allowing developers worldwide to build with GPT-3. This was transformational. I’ve built multiple products on this API. Companies like Stripe, Zapier, and hundreds of others integrated OpenAI models. Revenue scaled rapidly. The API became a reliable cash machine.

November 30, 2022 – ChatGPT Launch: Everything changed. ChatGPT became public and instantly changed the world. Reached 100 million monthly active users in two months—faster than any consumer app in history. Everyone suddenly understood what AI could do. OpenAI went from respected research lab to household name overnight. Sam Altman became a celebrity. The board lost practical control of the narrative.

November 2023 – The Coup (That Wasn’t): This is the dramatic part that shaped everything that followed. The OpenAI board (led by non-profit directors including chairman Greg Brockman) abruptly fired Sam Altman as CEO, citing “not being candid in his communications.” No warning. A single blog post. The specific reasons remain murky—some speculate AI safety disagreements, others think it was commercial vs. research tensions.

The next 72 hours were chaos:

• OpenAI’s 700+ employees (including many top researchers) signed a letter saying they’d quit if Altman wasn’t rehired
• Satya Nadella (Microsoft CEO) instantly offered Altman a position to lead a new team at Microsoft
• Altman started accepting the Microsoft offer
• The board realized they had zero leverage
• Negotiations occurred over three days
• Altman was rehired as CEO with a completely restructured board
• The non-profit’s governance control dissipated entirely

This was essentially a hostile takeover by a company against its own non-profit structure. It revealed that organizational governance is fragile when talent concentration is this high. A few dozen researchers proved more valuable than a board of directors. It also revealed that Sam Altman is exceptionally good at political maneuvering and negotiation.

2024 – Moving to For-Profit: OpenAI is transitioning entirely to for-profit status, abandoning even the pretense of non-profit control. The capped-profit structure is becoming a traditional venture-backed business. Valuation reached $157 billion in early 2024 funding rounds.

Complete Product Lineup

GPT-4o (Omni): The current flagship. Multimodal (text, images, audio). Better reasoning than GPT-4 Turbo, faster inference, cheaper pricing. This powers ChatGPT’s best performance and is the model most developers should use for new applications.

GPT-4 Turbo: Previous flagship with 128K context window. Still excellent for specialized tasks. Available via API and ChatGPT Plus, though slowly being superseded by GPT-4o.

GPT-3.5 Turbo: Older but capable. Fast and cheap. Powers the free ChatGPT tier. Useful for simple tasks or cost-sensitive applications where speed matters more than perfect reasoning.

GPT-4 Vision: Specialized version of GPT-4 for image understanding. Being integrated into GPT-4 Turbo and replaced by the multimodal capabilities of GPT-4o.

DALL-E 3: Image generation model significantly superior to DALL-E 2. Follows detailed prompts accurately. I use this at AI Box for generating UI mockups and marketing visuals. Output quality is genuinely impressive, though it sometimes struggles with hands, text, and complex scenes.

Whisper: Speech-to-text model trained on multilingual data. Incredibly accurate even with heavy accents, background noise, and multiple speakers in a single audio. We integrated this into AI Box for voice note transcription. The API is straightforward and reliable.

Sora: Video generation model. Still in limited beta as of early 2025. Generates short videos from text descriptions. Honestly remarkable—it understands physics, continuity, and motion in ways that are slightly terrifying. Not broadly available yet, but will be industry-shifting when released.

Embeddings Models: Convert text into vector representations for semantic search, clustering, and recommendation systems. Used behind the scenes in many applications.

API Pricing Deep Dive

OpenAI uses token-based pricing. One token roughly equals 4 characters of English text or 0.75 words.

GPT-4o Pricing:
– Input: $0.15 per 1 million tokens ($0.00015 per 1,000)
– Output: $0.60 per 1 million tokens ($0.00060 per 1,000)
– Roughly 50% the cost of GPT-4 Turbo with superior performance

GPT-4 Turbo Pricing:
– Input: $0.01 per 1K tokens
– Output: $0.03 per 1K tokens
– Higher cost but still reasonable for serious applications

GPT-3.5 Turbo Pricing:
– Input: $0.50 per 1 million tokens
– Output: $1.50 per 1 million tokens
– So cheap that cost is rarely a constraint

DALL-E 3 Pricing:
– Standard images: $0.04-$0.12 per image depending on size and quality
– High-definition images: Higher cost
– Can generate multiple images in one call to reduce cost per image

Whisper Pricing: $0.02 per minute of audio

Volume Discounts: Major API users negotiate custom pricing. Microsoft gets preferential rates as a major investor. Most companies don’t reach this volume, but it exists as an option.

My advice: Start with GPT-3.5 Turbo or GPT-4o based on requirements. For prototypes, cost isn’t typically the limiting factor—quality and speed are. Once at scale, you can optimize pricing. Always track API costs carefully; they can surprise you at scale.

The Sam Altman Drama (2023-24)

The November 2023 firing and rehiring deserves deep analysis because it shaped everything about OpenAI’s future and revealed truths about how power actually works in tech.

The Setup: Sam Altman had become the public face of OpenAI. He was on interview shows, meeting with politicians, pitching investors. He raised OpenAI’s profile and its valuation. But this visibility also created friction with the research-focused board and non-profit structure. OpenAI’s original mission emphasized safety and long-term research over commercial growth.

The Firing: On November 17, 2023, the board fired Altman without warning, announcing it via blog post. The public reason: “not candid in his communications with the board.” This is vague enough to hide real issues. Possible factors: AI safety disagreements, frustration with commercial direction, concern about his public prominence, or internal politics. The board also briefly attempted to rehire chief scientist Dario Amodei as CEO, further suggesting this was about vision differences.

The Disaster: Within 24 hours, the situation spiraled. Altman was meeting with Microsoft leadership. His team began looking at external opportunities. Employees were threatening to leave. The board had underestimated how concentrated value was in Altman and his core team. They essentially held all the cards.

The Resolution: Altman negotiated his rehiring with complete board restructuring. A new board was assembled with commercial-focused directors rather than research-focused ones. The non-profit structure remained but lost practical control. This was a complete victory for Altman and commercial interests.

The Lesson: This incident revealed several truths: (1) Organizational governance structures are fragile when talent is this concentrated, (2) In AI, people matter more than institutions, (3) Money and power always win over stated values, (4) Board of directors have less actual power than we think if key talent is willing to walk.

Business Model & Valuation

Revenue Streams:

ChatGPT Subscriptions: $20/month for Plus, $30/user/month for Team, $200/month for Pro (new tier). Millions of paying subscribers generate substantial recurring revenue. Estimated to be the fastest-growing subscription service ever. At $20/month × 10+ million users, that’s $240+ million annualized from subscriptions alone.

API Usage: Developers and companies pay per token consumed. This scales infinitely. Major companies (Microsoft, Stripe, etc.) spend millions monthly. This is likely OpenAI’s largest revenue stream now and the most scalable one.

Enterprise Licenses: Custom deployments, priority access, SLAs, and custom fine-tuning for large organizations. Higher margins but smaller volume than consumer or API channels.

Valuation Reality Check: At $157 billion valuation (as of 2024 funding rounds), OpenAI’s valuation is… ambitious. OpenAI is profitable, but probably not at a scale that justifies this valuation under traditional metrics like revenue multiples or earnings multiples. It’s priced on belief in future dominance and ability to maintain a moat against competitors.

Whether that’s realistic is the key question. If OpenAI maintains AI dominance for 5-10 years, the valuation is cheap. If competitors catch up (Claude, Gemini, open-source models), the valuation is expensive. The outcome depends largely on execution and competitive dynamics.

Strategic Partnerships

Microsoft – The Most Important Partnership: Microsoft invested $10+ billion, gaining exclusive cloud access to OpenAI’s models and partnership for integrating into Office 365, Windows, and other products. This is mutually beneficial: OpenAI gets capital and massive distribution; Microsoft gets access to best-in-class AI. It’s also somewhat symbiotic—Microsoft can’t leave OpenAI easily, and OpenAI depends on Microsoft infrastructure. Neither company will abandon this lightly.

Apple Integration: Starting in 2025, Apple devices will integrate ChatGPT. This is huge for distribution—billions of iPhone, iPad, and Mac users gain access to ChatGPT. Apple gets best-in-class AI; OpenAI gets global distribution. This is the kind of integration that makes competitors nervous.

Enterprise Partnerships: Companies across finance, healthcare, consulting, and technology are building proprietary solutions on OpenAI’s API. These partnerships validate the technology and generate revenue. They also lock customers into OpenAI’s ecosystem through switching costs and integration depth.

Academic & Research Partnerships: OpenAI funds research at universities and collaborates with academic labs. This both builds credibility and creates a talent pipeline.

Market Position & Competitors

OpenAI’s dominance is real but eroding. Competitors are advancing rapidly:

Anthropic (Claude): Excellent reasoning, better honesty about limitations, longer context windows. Claude is competitive on most benchmarks and has passionate advocates. If Anthropic stays focused and funded, they’re the primary threat.

Google (Gemini): Deep pockets, tons of engineers, massive data. Gemini is competitive and improving. Google’s distribution advantage is huge—Gmail, Android, Search. If Google integrates Gemini properly, this could become the primary AI for billions of people.

Open Source (Llama 3.1, others): Meta’s Llama 3.1 is genuinely capable. Open-source models are advancing fast and have advantages: can run locally, can fine-tune, can customize. Disadvantage: require technical skill and compute resources.

OpenAI’s current moat is: (1) best models, (2) best product execution, (3) strategic partnerships, (4) first-mover advantage. These are real but not permanent. The industry will eventually commoditize. Question is whether OpenAI can maintain premium pricing and market share as competitors catch up.

Risks & Considerations

Dependency Risk: If you build your product entirely on OpenAI’s API, you’re dependent on their pricing, availability, and feature roadmap. Wise to have fallback options or competitive analysis in your roadmap.

Regulatory Risk: AI regulation is coming. OpenAI is a likely regulatory target given its prominence. Compliance costs could increase.

Safety & Alignment Concerns: OpenAI’s alignment and safety work is less transparent than some would prefer. Real questions remain about how to ensure advanced AI systems remain beneficial.

Market Concentration: OpenAI has enormous influence over AI development direction. This concentration of power raises questions about appropriate distribution of AI benefits.