This page records what named people have said, on the record, about when AGI or superintelligence will arrive. It has 20 entries as of November 13, 2026. The years named run from 2026 to 2047, and several speakers decline to name a year at all. The entries are not comparable one to one, because each speaker is predicting a different thing: "powerful AI", "AGI", "a system that learns like a human" and "superintelligence" have different definitions, set out in AGI vs ASI vs superintelligence. This page reports the statements. It does not rate them.
Lab Leaders and Lab Staff
| Who | What they said | Year or range named | Date and venue |
|---|---|---|---|
| Sam Altman, OpenAI | Superintelligence is possible "in a few thousand days", and it may take longer | No year. A few thousand days from late 2024 | September 23, 2024, essay "The Intelligence Age" |
| Sam Altman, OpenAI | OpenAI is confident it knows how to build AGI as traditionally understood and is turning its aim to superintelligence | No year | January 2025, blog post "Reflections" |
| Sam Altman, OpenAI | Systems that can work out novel insights are likely in 2026, and robots that do real-world tasks may arrive in 2027 | 2026, 2027 | June 10, 2025, essay "The Gentle Singularity" |
| Sam Altman, OpenAI | Internal goals of an automated AI research intern by September 2026 and a true automated AI researcher by March 2028 | 2026, 2028 | Late October 2025, company livestream, as reported. Sources give October 28 to 30 |
| Dario Amodei, Anthropic | "Powerful AI" could come as early as 2026, and could take much longer | 2026 at the earliest | October 2024, essay "Machines of Loving Grace" |
| Anthropic (company) | Expects powerful AI systems to emerge in late 2026 or early 2027 | 2026 to 2027 | March 6, 2025, submission to the White House science office |
| Dario Amodei, Anthropic | Powerful AI could be as little as 1 to 2 years away, and could be considerably further out | 2027 to 2028 at the earliest | January 2026, essay "The Adolescence of Technology" |
| Jack Clark, Anthropic | Now thinks it more likely that full recursive self-improvement becomes possible in 2027. He previously expected 2028 | 2027 | October 12, 2026, newsletter Import AI 476 |
| Demis Hassabis, Google DeepMind | A 50% chance of AGI within the decade, and a timeline of five to 10 years, with one or two more breakthroughs possibly needed | 2031 to 2036 for the five to 10 year range | January 2026, World Economic Forum, Davos, as reported by Fortune |
| Demis Hassabis, Google DeepMind | AGI is "probably only a few short years away" | No year | July 14, 2026, essay for the DeepMind Institute |
| Shane Legg, Google DeepMind | A 50% chance of AGI by 2028 | 2028 | October 26, 2023, Dwarkesh Patel podcast |
| Mark Zuckerberg, Meta | Developing superintelligence "is now in sight", and the rest of this decade is likely the decisive period | No year. By 2030 implied as the decisive period | July 30, 2025, letter "Personal Superintelligence" |
| Elon Musk | AI smarter than any individual human by the end of 2026, and smarter than all of humanity combined within about five years | 2026, and by about 2031 | January 22, 2026, World Economic Forum, Davos, as reported |
| Mustafa Suleyman, Microsoft AI | Medical superintelligence will arrive "in the next few years" | No year | November 6, 2025, essay "Towards Humanist Superintelligence" |
| Ilya Sutskever, Safe Superintelligence | "5 to 20" years for a system that learns as well as a human and then becomes superhuman | 2030 to 2045 | November 25, 2025, Dwarkesh Patel podcast |
Researchers, Forecasters and Surveys
| Who | What they said | Year or range named | Date and venue |
|---|---|---|---|
| Geoffrey Hinton | Superintelligent AI could arrive in 4 to 19 years, with a good chance it is here in 10 years or less | 2029 to 2044 | April 26, 2025, CBS interview, as reported by one outlet |
| Yann LeCun | Human-level intelligence will not be reached by training large language models or by training on text alone | No year | January 2026, AI House, Davos, as reported by Fortune |
| Andrej Karpathy | AGI is about a decade away | About 2035 | October 17, 2025, Dwarkesh Patel podcast |
| Daniel Kokotajlo and co-authors | The "AI 2027" scenario places artificial superintelligence in December 2027. The authors later said 2027 was their most likely single year and their medians were later, with one author at 2030 | 2027 in the scenario, later medians | April 3, 2025, ai-2027.com, with notes added through December 2025 |
| 2,778 AI researchers (survey) | Aggregate forecast of a 10% chance by 2027 and a 50% chance by 2047 of machines outperforming humans at every task | 2047 for the 50% point | Survey paper first posted January 5, 2024 |
How to Read the Table
- The years in the third column are ours where the speaker gave a span. "5 to 20 years" said in November 2025 is shown as 2030 to 2045. The speaker's own words are in the second column.
- The targets differ. Amodei's "powerful AI" is defined by a list of capabilities and he avoids the term AGI. Sutskever's answer is about a system that learns like a person. The survey asks about machines that outperform humans at every task. Only some entries are about superintelligence in the sense of far exceeding human ability.
- Statements age. Altman's 2026 item and OpenAI's September 2026 research-intern goal now refer to the present. In September 2026 OpenAI said it had met the intern goal and was making strong progress toward March 2028, according to a court filing that quotes the company's post. That is the company's own assessment.
- The same person moves. Clark moved from 2028 to 2027. The AI 2027 authors moved later. Hassabis gave five to 10 years in January 2026 and "a few short years" in July 2026.
Where the Statements Matter
Several of these statements are quoted in a complaint filed on November 12, 2026 that asks a California court to stop five companies from building more capable systems. See the lawsuit to halt frontier AI development. The organisations whose stated goal is superintelligence are listed in the superintelligence lab tracker, and Sutskever's company is covered in SSI funding and valuation.
Methodology
Each entry is a dated, attributable statement. Where the speaker published it, we read the original: Altman's The Intelligence Age, Reflections and The Gentle Singularity. Amodei's Machines of Loving Grace and The Adolescence of Technology. Anthropic's March 2025 submission. Import AI 476. Hassabis's July 2026 essay. Zuckerberg's letter. Suleyman's essay. The Dwarkesh Patel interviews with Legg, Sutskever and Karpathy. AI 2027. And the survey paper. Four entries rest on press reports of spoken remarks: the Davos statements by Hassabis and LeCun (Fortune), Musk at Davos (Euronews), Altman's livestream goals (TechRadar, with the months as quoted from OpenAI's September 2026 post in the complaint cited below) and Hinton's CBS interview (PPC Land). The dates of two Altman posts are not shown on the pages and are taken from citations in the November 12, 2026 complaint. Read on November 13, 2026.