AI Thinkers·CEO, OpenAI
Sam Altman
We are on an unbroken exponential curve toward superintelligence, and the transition can be managed without catastrophe if the benefits are broadly distributed. AI is not a product category but a new form of infrastructure: intelligence itself becoming a utility, as cheap and ubiquitous as electricity. The goal is a compression of centuries of scientific and economic progress into a single generation. His policy orientation holds that getting the distribution of this abundance right — not slowing the technology — is the defining governance challenge.
Key positions
Superintelligence is a matter of years, not decades.
"It is possible that we will have superintelligence in a few thousand days; it may take longer, but I'm confident we'll get there." In "The Gentle Singularity" (Jun 2025) he argued "we are past the event horizon; the takeoff has started" while insisting the experience will "feel impressive but manageable." His framing is deliberately reassuring: the singularity is real and near, but arrives as a smooth curve.
"The Intelligence Age," ia.samaltman.com, September 2024; "The Gentle Singularity," blog.samaltman.com, June 2025
Intelligence will become too cheap to meter — a utility with super-exponential economic value.
"Towards intelligence too cheap to meter" (July 2024). "Three Observations" (Feb 2025) formalised the economics: AI costs fall ~10x every 12 months — token cost from GPT-4 to GPT-4o dropped ~150x in 18 months — while the socioeconomic value of linearly increasing intelligence is "super-exponential."
Sam Altman on X (@sama), July 2024; "Three Observations," blog.samaltman.com, February 2025
Capital and labour will be dangerously disrupted — and this requires deliberate policy intervention.
"Moore's Law for Everything" (2021) warned: "The balance of power between capital and labor could easily get messed up, and this may require early intervention." His proposed solution at the time — an American Equity Fund — positioned him as one of the few major tech leaders to acknowledge distributive risk in concrete policy terms. His April 2026 "New Deal" proposal repeated this at a larger scale.
"Moore's Law for Everything," moores.samaltman.com, March 2021; OpenAI "Industrial Policy for the Intelligence Age," April 2026
Compute and energy infrastructure — not algorithms — are the current binding constraint.
AI models will continue to improve on scaling laws, so the bottleneck is energy production. Altman pushed the Stargate project (17 gigawatts of AI-dedicated power), backed nuclear fusion startup Helion Energy ($375M), and supports fission company Oklo. He has said AI will require "an energy breakthrough, including fusion."
Popular Science, February 2024; DataCenter Dynamics, 2024; Fortune, September 2025
OpenAI's closed model approach was a historical mistake.
Following DeepSeek's January 2025 open-weight release: "I personally think we have been on the wrong side of history here and need to figure out a different open source strategy." His hedge — "not everyone at OpenAI shares this view, and it's also not our current highest priority" — reflects genuine internal tension.
TechCrunch, January 2025; VentureBeat, January 2025
Safety at OpenAI is about civilisational distribution — not primarily technical alignment.
His "vibes don't match a lot of the traditional AI-safety stuff." Safety framing is macro-societal: the danger is concentration of power — by authoritarian states or by a single company including OpenAI. However, OpenAI's superalignment team received only 1–2% of compute and was dissolved in May 2024, raising questions about how seriously technical safety was resourced.
New Yorker investigation, 2024; Sam Altman, "Reflections," blog.samaltman.com, January 2025
In their own words
“It is possible that we will have superintelligence in a few thousand days; it may take longer, but I'm confident we'll get there.”
“Anyone in 2035 should be able to marshal the intellectual capacity equivalent to everyone in 2025.”
“Once you see AGI you can't unsee it. It has a real 'ring of power' dynamic to it... The only solution I can come up with is to orient towards sharing the technology with people broadly, and for no one to have the ring.”
“We are now confident we know how to build AGI as we have traditionally understood it.”
Tensions with other thinkers
Where this view genuinely conflicts with others in the field.
Amodei: safety requires hard technical infrastructure and capability gates. Altman: safety is primarily a distribution and governance problem.
Altman has pivoted toward open ("wrong side of history"); Amodei maintains closed models are necessary for safety.
LeCun argues scaling transformers will not reach AGI; Altman treats continued scaling as the near-certain path to superintelligence.
Karpathy says 10+ years; Altman says "a few thousand days" (roughly 5–8 years).