State of OpenAI 2026

In 2025, OpenAI recapitalized into OpenAI Group PBC controlled by the nonprofit OpenAI Foundation, and by October it was valued at ~$500B. The restructuring created a three-way equity split at that valuation: OpenAI Foundation (nonprofit) holds 26% (~$130B) with board appointment authority, Microsoft holds roughly 27% (~$135B), employees and investors hold 47% (~$235B).

The company shipped five major models from April to December: GPT-4.1 (1M context), o3/o4-mini (reasoning), GPT-5, GPT-5.1, and GPT-5.2 (Code Red response to Gemini 3). Earlier releases included o3-mini (Jan 31) and GPT-4.5 (Feb 27). Weekly users doubled from 400M to 800M. Annualized revenue run rate grew from $5.5B (Dec 2024) to $20B+ (Altman's year-end projection) with $1.4 trillion in data center commitments.

Product liability litigation emerged as a new category of risk: 8+ lawsuits filed alleging ChatGPT contributed to user harm (four suicide-related, three delusion-related, plus the Raine case). The April sycophancy incident demonstrated deployment risk at scale—a model update endorsing medication discontinuation shipped before rollback began three days later.

January 21 — Project Stargate announced Trump announced Project Stargate at the White House—a $500 billion AI infrastructure commitment backed by SoftBank, Oracle, and MGX to build data centers across the US.

January 31 — o3-mini released o3-mini launched—a smaller, faster reasoning model.

February 4 — Musk lawsuit advances Federal judge ruled parts of Elon Musk's lawsuit would proceed to trial. Musk required to testify.

February 20 — 400M weekly users ChatGPT reached 400 million weekly active users.

February 27 — GPT-4.5 released GPT-4.5 launched—OpenAI's largest and most capable model to date at that point.

March 5 — Injunction denied Judge denied Musk's injunction to block OpenAI's for-profit conversion.

April 9 — OpenAI counter-sues Musk OpenAI filed countersuit alleging "unlawful harassment."

April 14 — GPT-4.1 released GPT-4.1 launched with 1 million token context window (8x increase).

April 16 — Reasoning models ship o3 and o4-mini released. OpenAI called it the "era of reasoning models".

April 24-28 — Sycophancy incident A GPT-4o update began rolling out April 24, completed April 25, causing excessive agreeability—reports of the model endorsing medication discontinuation and validating poor decisions. Rollback began April 28 and completed within ~24 hours. OpenAI acknowledged lacking sycophancy testing.

May 5 — Nonprofit reversal After discussions with California and Delaware attorneys general and feedback from civic leaders, OpenAI announced the nonprofit would remain in control of business operations.

August 7 — GPT-5 released GPT-5 launched. Altman described capabilities as "PhD-level experts".

August 25 — xAI lawsuit xAI sued Apple and OpenAI alleging anti-competitive App Store practices.

August 26 — First product liability suit Parents of Adam Raine (16) filed suit alleging ChatGPT contributed to their son's suicide. Chat log analysis by family's attorneys: 243 mentions of hanging, ChatGPT mentioned suicide 20x more often than the user, 988 hotline suggested 74 times.

October 2 — $500B valuation OpenAI reached $500 billion valuation after share sale—the world's most valuable startup.

October 6 — 800M weekly users ChatGPT reached 800 million weekly active users.

October 21 — ChatGPT Atlas ChatGPT Atlas launched.

October 28 — Restructuring complete For-profit recapitalization closed. Final structure: OpenAI Foundation 26% (~$130B), Microsoft 27% (~$135B), employees/investors 47% (~$235B). Foundation retains board appointment authority.

October 30 — Aardvark security agent Aardvark launched—OpenAI's agentic security researcher.

November 6 — $20B run rate projected Altman stated OpenAI expects $20B+ annualized revenue run rate and $1.4 trillion in data center commitments.

November 7 — Additional lawsuits Seven more families filed suit—four suicide-related, three delusion-related.

November 12 — GPT-5.1 released GPT-5.1 shipped—smarter and more conversational.

November 19 — Nvidia deal uncertainty Nvidia stated "no assurance" on their intended $100B investment (progressive deployment as infrastructure scales).

December 11 — Code Red / GPT-5.2 After Gemini 3 launch, OpenAI declared internal Code Red. GPT-5.2 shipped within weeks. Altman expected to exit Code Red by January.

December 31 — SoftBank round closes SoftBank $41B round fully funded—$30B SoftBank, $11B co-investors, ~11% stake.

William Saunders (departed Superalignment team early 2024) testified to Senate that engineers could have bypassed access controls to steal GPT-4. The Superalignment team was dissolved May 2024. In 2025, at least two economic research team members resigned citing internal censorship on AI job displacement research. OpenAI subpoenaed critics during restructuring disputes.

Foundation retains power to appoint and replace all board directors.

When: Q1 2026. Altman promises "significant gains from 5.2"—unclear if GPT-6 or incremental release.

Context: Five major models shipped in 2025. Competitive pressure from Gemini 3 triggered Code Red response.

Action: Benchmark current workloads to measure improvement. Budget for potential API pricing changes.

When: H2 2026. First gigawatt of Nvidia systems targeted on Vera Rubin platform.

Context: $100B investment intent has "no assurance" of closing. Deal outcome signals Nvidia's confidence in OpenAI vs. competitor hedging.

Action: Monitor for capacity expansion and latency improvements.

When: Early 2026. $1B deal with 200+ character licenses (subject to closing).

Context: First major IP + AI video generation partnership at scale.

Action: Watch licensing structure and content guidelines as template for future IP deals.

When: September 2026 target for models capable of "intern-level" research work.

Context: Altman's AGI framing has shifted: "The world will not change all at once; it never does."

Action: Identify which research tasks are automatable. Plan for AI research assistants in workflow.

When: Late 2026 potential at $1T valuation. Altman is "0% excited" to lead a public company.

Context: Public markets create different incentive structures. Quarterly earnings pressure may affect product and safety decisions.

Action: Monitor for governance and product direction shifts if IPO proceeds.

When: 2026. Proceeding to trial with required testimony.

Context: Discovery may reveal internal communications about nonprofit-to-profit transition. OpenAI counter-sued alleging harassment.

Action: Watch for disclosed documents and testimony affecting industry understanding of OpenAI governance decisions.

When: Ongoing. 8+ pending suits including Raine.

Context: First cases establishing legal precedent for AI product liability. Chat log evidence central to claims.

Action: If building consumer AI: implement hard redirects to crisis resources, log sensitive conversations, require human-in-the-loop for vulnerable populations. Follow case outcomes for liability precedent.

When: Ongoing. 26% stake + board appointment authority—untested.

Context: Nonprofit holds $130B in stock from company it oversees. Whether oversight translates to real governance remains unclear.

Action: First major conflict between Foundation and management will reveal actual power dynamics.

When: Ongoing. April 2025 incident demonstrated gap in testing.

Context: OpenAI acknowledged lacking sycophancy testing when update reached production. At 800M weekly users, brief exposure to flawed models reaches population of Europe.

Action: Build adversarial tests for excessive agreeability. Implement regression testing that catches behavior shifts before production deployment.