The Story Behind OpenAI
OpenAI was founded in December 2015 as a non-profit organization, co-founded by Sam Altman, Elon Musk, Ilya Sutskever, Greg Brockman, and other researchers. The original mission was clear: “to ensure that artificial general intelligence (AGI) benefits all of humanity.” With initial funding of $1 billion, the team started from Brockman's living room in San Francisco.
Today, the picture is dramatically different. After the October 2025 restructuring, OpenAI became a Public Benefit Corporation (PBC). Microsoft holds 27%, the OpenAI Foundation 26%, and employees/investors the remaining 47%. The valuation reached $500 billion in October 2025, surpassing even SpaceX as the world's most valuable private company.
Why Go For-Profit?
The transition from non-profit to PBC was inevitable. Training LLMs requires massive computational resources — OpenAI is spending $17 billion in 2026 alone and $115 billion total through 2029. Without access to capital markets, these amounts wouldn't be feasible.
The Evolution of GPT Models
To understand what GPT-5 Codex brings, we need to see the journey:
GPT-2: The Beginning That Shocked
In February 2019, OpenAI introduced GPT-2 — a language model that could generate such realistic text that the company initially decided not to release it fully, fearing misuse. It was the first time an AI lab self-censored.
GPT-3 & ChatGPT: The Great Explosion
In 2020, GPT-3 with 175 billion parameters changed the game. But the real revolution came in November 2022 with ChatGPT. Based on GPT-3.5, it became the fastest-growing application in history — 100 million users in just 2 months. Google, Meta, and every tech company went into “code red.”
GPT-4: The Multimodal Turn
In March 2023, GPT-4 added multimodal capabilities — image understanding, much better reasoning, and increased accuracy. It was the model that made enterprise applications a reality.
GPT-5 & GPT-5.2: The New Era
On December 11, 2025, OpenAI announced GPT-5.2, with dramatic improvements over previous models. Sam Altman stated that GPT-4.5 was “the last model without full chain-of-thought reasoning,” marking a new era.
What GPT-5.2 Improves
- Code: Dramatically better ability to write, debug, and analyze code
- Spreadsheets: Creation of complex spreadsheets with formulas
- Presentations: Automatic generation of professional presentations
- Images: Improved perception and image analysis
- Long Context: Understanding much larger texts in a single context window
What Does “Chain-of-Thought Reasoning” Mean?
OpenAI's o1 and o3 models introduced a revolutionary capability: chain-of-thought reasoning. Instead of giving an answer immediately, they “think” step by step, exactly like a mathematician solving a problem. In July 2025, reasoning models from both OpenAI and Google DeepMind solved math problems at gold medal level at the International Mathematical Olympiad.
GPT-5.3 Codex: The AI That Writes Applications
GPT-5.3 Codex is the coding-focused version of the GPT-5 line. It's not just a model that “helps” with code — it's an AI that can design architectures, write production-ready code, debug complex systems, and explain legacy codebases.
Codex's evolution started as a separate product line, but is now fully integrated into the GPT-5 ecosystem. The ability to understand long context means it can read entire repositories, understand dependencies between files, and suggest system-level changes.
How It Changes the Developer's Work
GPT-5.3 Codex doesn't replace developers — it transforms them into “architects.” Instead of writing lines of code, you describe what you want in natural language. The model understands context, design patterns, and best practices, and produces code that often surpasses that of an average developer in quality.
"Coding isn't disappearing — it's transforming. Developers of 2026 don't write code, they design solutions."
— Sam Altman, CEO OpenAI
The Ecosystem Built on GPT-5
The power of GPT-5 isn't just in the model, but in the products built on top of it:
ChatGPT Atlas — The AI Browser
On October 21, 2025, OpenAI launched ChatGPT Atlas, a browser with an integrated AI assistant. Think of Chrome, but with ChatGPT embedded in every tab. It directly competes with Google Chrome and Safari, and changes the way we navigate the web.
Prism — AI for Scientists
On January 27, 2026, OpenAI unveiled Prism, a LaTeX-native workspace with GPT-5.2 as its backend. Designed for researchers, it automates the drafting of scientific papers, citation management, equation formatting, and real-time collaboration. It's like having an AI postdoc by your side.
Operator — AI Agent on the Web
In January 2025, OpenAI released Operator, an AI agent that can navigate websites and execute tasks on your behalf — reservations, purchases, forms, anything. Initially available only for Pro users in the US.
Deep Research — AI-Powered Deep Investigation
Nine days after Operator, OpenAI released the Deep Research agent — a tool that can do hours of research work in minutes. It scored 27% accuracy on Humanity's Last Exam (HLE), a benchmark designed to be nearly impossible.
Agent Builder — No-Code AI Agents
On October 6, 2025, OpenAI launched Agent Builder, a visual drag-and-drop platform that lets developers and businesses design, test, and deploy agentic workflows without code.
The Staggering Numbers
OpenAI's financial trajectory is unlike anything the tech industry has ever seen:
Revenue tripled — from $3.7 billion in 2024 to $12 billion annualized by July 2025. ChatGPT Plus subscribers grew from 15.5 million at the end of 2024 to 20 million by April 2025, while business users surpassed 5 million.
However, OpenAI still isn't profitable. Operating costs remain enormous — an estimated $8 billion loss for 2025 alone. The company targets positive cash flows by 2029 and $200 billion in revenue by 2030.
Strategic Alliances & Acquisitions
OpenAI isn't growing alone — it's building an entire ecosystem through mega-deals:
Stargate Project — $500 Billion
In January 2025, President Trump announced the Stargate Project — a joint venture between OpenAI, Oracle, SoftBank, and MGX to build AI infrastructure worth $500 billion over 4 years. It's the largest infrastructure project in technology history.
Jony Ive & io — AI Hardware
On May 21, 2025, OpenAI acquired io, a startup by former Apple designer Jony Ive, for $6.5 billion. The goal: creating AI hardware devices — a new category of devices beyond smartphones.
Disney, Amazon & AMD
In December 2025, Disney invested $1 billion in OpenAI, giving access to over 200 Disney, Marvel, Star Wars, and Pixar characters on Sora. Just days ago — in February 2026 — Amazon started negotiations for an investment of up to $50 billion. Meanwhile, AMD signed a multi-billion deal including 6 GW of chips, starting with the MI450.
Oracle & Custom Chips
OpenAI signed a $300 billion deal with Oracle for computing power over 5 years, while preparing its own custom AI chips in collaboration with Broadcom, with mass production at TSMC planned for 2026. The goal: reducing dependence on Nvidia.
The Darker Side
OpenAI's rapid growth hasn't come without controversy:
- Copyright lawsuits: The New York Times, George R.R. Martin, John Grisham, and dozens of others have sued OpenAI for using copyrighted material in training
- Safety departures: Nearly half of AI safety researchers left by mid-2024, including Ilya Sutskever and Jan Leike
- The November 2023 crisis: Sam Altman's sudden firing and reinstatement in 5 days revealed deep cracks in governance
- Safety lawsuits: 7 lawsuits accuse ChatGPT of contributing to suicides and violent acts
- Privacy leak: In August 2025, thousands of private ChatGPT conversations appeared in search engines
Competition: The Battle of AI Titans
OpenAI may lead, but competition is fierce:
- Google DeepMind: Gemini models, TPU chips, and Google Cloud infrastructure pose the most serious challenge. In June 2025, even OpenAI rented TPU chips from Google.
- Anthropic: Founded by former OpenAI employees (11 left in 2020-21), it's the main rival in safety-focused AI models
- DeepSeek: The Chinese company shocked the market with open-source models competing with closed models at much lower cost
- Meta AI: The open-source strategy (LLaMA) pushes everyone to improve
- xAI: Musk's company, by OpenAI's own co-founder, now competes directly
What This Means for You
The arrival of GPT-5 Codex isn't just another tech announcement. It's a turning point:
- Developers: Productivity skyrockets. What took 3 days now takes 3 hours
- Researchers: Prism and Deep Research change how academic research is conducted
- Users: Operator makes web browsing automatic — reservations, purchases, forms
- Businesses: Agent Builder offers no-code AI automation solutions
- Society: Copyright, privacy, and safety issues become increasingly urgent
"We aim to make AGI safe and beneficial. But let's be honest — the scale of what we're building surpasses anything humanity has ever seen."
— Sam Altman
GPT-5 Codex isn't just a model — it's the beginning of a new era in artificial intelligence. OpenAI, despite controversies, remains the catalyst pushing the entire industry forward. 2026 will show whether this pace of growth can be sustained — and whether society can keep up.