OpenAI

The company that put AI in everyone's hands

OpenAI is the company that turned a decade of quiet research on artificial general intelligence into a product hundreds of millions of people touch every week. ChatGPT, launched at the end of 2022, became the fastest-growing consumer application in history and is now the front door through which most of the world first experienced generative AI. By early 2026 it had crossed 900 million weekly active users and roughly a billion monthly users, a scale of adoption no software category had reached this quickly.

What makes OpenAI more than a single hit product is the stack beneath it. The company designs its own frontier models, runs the largest consumer AI service, operates a fast-growing developer and enterprise platform, and is now building both the data centers and the custom silicon to train the next generation. That vertical ambition, from research to chips to a device in your pocket, is the thesis: OpenAI is betting that whoever controls the best models and the compute to run them will define the next computing platform.

Source

From nonprofit charter to public benefit corporation

OpenAI was founded in December 2015 as a nonprofit research lab, with Sam Altman, Elon Musk, Greg Brockman, Ilya Sutskever and others pledging $1 billion and a mission to ensure AGI benefits all of humanity. As the cost of training models exploded, that structure buckled. In 2019 OpenAI created a capped-profit subsidiary and took a $1 billion investment from Microsoft, moving its workloads onto Azure and beginning one of the deepest partnerships in technology.

The tension between mission and money nearly broke the company in November 2023, when the board abruptly fired Altman only to reinstate him five days later under overwhelming employee and investor pressure. The episode reshaped OpenAI's governance and set the stage for a formal restructuring, completed in October 2025, that reorganized the for-profit as OpenAI Group PBC, a Delaware public benefit corporation controlled by the nonprofit OpenAI Foundation.

Under the recapitalization the Foundation retains special voting rights and appoints the entire board, while Microsoft holds roughly 27% of the for-profit, a stake valued in the hundreds of billions. It is an unusual arrangement for a company of this size, and one that investors will keep scrutinizing as OpenAI moves toward the public markets.

Source

The model engine and the agentic turn

OpenAI's product cadence is relentless. GPT-4 in 2023 set a new bar for reasoning; GPT-4o and the o1 reasoning family in 2024 introduced models that think step by step before answering; GPT-5 arrived in 2025. In 2026 the pace accelerated further: GPT-5.5, released in April, became the generally available flagship and the default ChatGPT model, built for agentic coding and computer use. The next-generation GPT-5.6 family, split into Sol for the most demanding domains, Terra as a balanced mid-tier, and Luna for speed and low cost, moved from a government-gated preview to full public availability on July 9, 2026.

The strategic shift underneath the version numbers is the move from chat to agents. Products like Codex for software engineering and the newly launched ChatGPT Work, which can gather context across a user's connected apps and work autonomously for hours to return finished spreadsheets, slides and documents, mark OpenAI's push to have AI complete multi-step tasks rather than just answer questions. This is the terrain where OpenAI competes most directly with Google's Gemini and Anthropic's Claude, and the model cadence is as much a response to that rivalry as it is a research roadmap.

Source

Two businesses under one roof

OpenAI runs what are effectively two companies. The first is the largest consumer AI product in the world: ChatGPT, with tiered plans from Free through Plus at $20 a month to Pro at $200, and roughly 50 million paying subscribers. The second is a business-and-developer platform serving more than a million organizations and around 4 million developers through per-token API access and per-seat enterprise plans. The enterprise segment now accounts for more than 40% of revenue and is the faster-growing half.

The moat is a combination of distribution, brand and model quality that few rivals can match at once. ChatGPT is close to a generic term for AI assistants, and that consumer mindshare feeds the enterprise funnel. The relationship with Microsoft, once exclusive, loosened in April 2026 when the companies ended exclusivity and capped revenue sharing, freeing OpenAI to run multi-cloud and pursue its own infrastructure. The open question is durability: switching costs for a chatbot are low, and competitors are shipping capable models at aggressive prices, so OpenAI has to keep winning on quality and product depth rather than lock-in.

Source

The compute gambit

OpenAI's defining wager is that frontier intelligence is bounded by compute, and it is spending on a scale to match. In January 2025 it announced Stargate, a joint venture with SoftBank, Oracle and MGX to invest up to $500 billion over four years in US AI data-center infrastructure, an umbrella now targeting roughly 10 gigawatts of capacity. That plan has been filling in with concrete sites: an expanded Oracle partnership worth more than $300 billion for about 4.5 GW, five additional US locations, and a custom AI accelerator program with Broadcom, fabricated by TSMC, rolling out from the second half of 2026.

In July 2026 OpenAI took the buildout a step further by announcing Project Camellia, its first wholly self-designed-and-built data-center campus, in Effingham County near Savannah, Georgia. The company plans an initial investment of roughly $20 billion rising past $30 billion at a full 3.2 GW scale, with power contracted through Georgia Power and phased delivery from 2028 to 2032. Owning the campus outright, rather than leasing capacity, signals a company trying to control its most important input rather than rent it, and it concentrates enormous execution and financing risk on OpenAI's own balance sheet.

Source

Record financing and the road to an IPO

The revenue curve is steep. OpenAI crossed a roughly $25 billion annualized run-rate in early 2026, generating on the order of $2 billion a month, up from about $13 billion recognized across 2025. Yet the company remains deeply unprofitable, because the compute needed to serve hundreds of millions of users and train ever-larger models consumes cash faster than subscriptions and API calls bring it in. Reconciling that growth with a credible path to profit is the central financial question hanging over the business.

To fund the ambition, OpenAI closed the largest private fundraise in history in March 2026: $122 billion at an $852 billion post-money valuation, with Amazon committing up to $50 billion, and Nvidia and SoftBank each investing $30 billion. It followed that in June 2026 with a confidential draft registration for an initial public offering that could value the company near or above $1 trillion, potentially the largest technology listing ever. In July 2026 OpenAI added Nubank founder David Velez and BNY chief executive Robin Vince to its boards, a move widely read as governance preparation for public-market scrutiny.

Source

What to watch next

Several threads will decide how the next year unfolds. On models, watch whether OpenAI can sustain its cadence toward an eventual GPT-6 while Gemini and Claude push hard on reasoning and agents. On the business, the key metric is whether enterprise and agentic products reach revenue parity with the consumer side, which OpenAI has said it expects by the end of 2026. On hardware, the first Jony Ive-designed io device, targeted for a second-half-2026 reveal, would mark OpenAI's entry into consumer devices, though shipping timelines have looked slippery.

The larger risks are execution and litigation. The Stargate and Project Camellia buildouts depend on partners, power and construction all arriving on schedule, and any IPO will force the compute burn into public view. In July 2026 Apple sued OpenAI and its io hardware unit over alleged trade-secret theft tied to unreleased Apple hardware, a reminder that OpenAI's expansion into devices and its aggressive talent recruitment carry legal exposure. For a company this consequential, the interesting questions are no longer whether the technology works, but whether the economics, the governance and the infrastructure can scale with it.

Source