Digital Economy

What is the 'gentle singularity' of Sam Altman the CEO of OpenAi?

ChatGpt's dad outlines a scenario where super-intelligent AI will transform the global economy by 2030, making artificial intelligence 'too cheap to measure'

13 June 2025

SAM ALTMAN, CHIEF EXECUTIVE OFFICER, OPEN AI

3' min read

Sam Altman does not mince words: 'We are beyond the event horizon, the take-off has begun'. In a long post on his personal blog, the OpenAI CEO paints a future where artificial super-intelligence will not only be a reality, but will become so cheap as to be comparable to the cost of electricity. A vision that, if realised, would completely reshape the global economic balance with 'an enormous acceleration' of scientific progress and increased productivity. A line of reasoning that opens with the flywheel effect: 'From now on, the tools we have already developed will help us find further scientific insights and create better artificial intelligence systems.

The mechanism Altman describes is that of a self-propulsive flywheel: the tools already built help find new scientific discoveries and create better AI systems. We are not yet at full recursive improvement, but 'an early version' of this process. But we are not far, according to Altman, from 'robots that can build other robots (and, in a sense, data centres that can build other data centres)'.

Scientists already report that they are two to three times more productive than in the pre-AI period. 'If we can do the research work of a decade in a year, or in a month, then the pace of progress will obviously be very different,' Altman points out.

How much does ChatGPT consume

In an optimistic prediction, Altman foresees an era in which Ai will be much cheaper than today, where an average ChatGPT query consumes 0.34 watt-hours of energy - as much as an electric oven in just over a second - and uses about 0.000085 gallons of water, equal to a fifteenth of a teaspoon. Figures that seem derisory today, but take on different proportions when multiplied by hundreds of millions of daily users. "The latest version of ChatGPT is already more powerful than any human being who has ever lived," Altman observes. And the numbers prove him right: hundreds of millions of people rely on these systems daily for increasingly complex and strategic tasks. Altman points out that AI will become as cheap as electricity, but a large network is needed - not just data centres, but also abundant renewable energy.

The acceleration timeline

According to the OpenAI CEO's predictions, 2025 has already seen the arrival of agents capable of performing real cognitive work - 'writing computer code will never be the same again'. 2026 should bring systems capable of entirely new insights, while 2027 could see robots operating in the physical world. But it is in the 2030s that Altman predicts the real breakthrough: 'Intelligence and energy - ideas and the ability to realise them - will become extremely abundant'. Two factors that have historically been major constraints on human progress.

Economic Challenges and Opportunities

The picture is not without its critical aspects. Altman acknowledges that 'whole classes of jobs will disappear', but argues that the world will become so rich so fast that we can 'seriously consider new political ideas that we could not afford before'. The historical analogy is that of the Industrial Revolution: a subsistence farmer of a thousand years ago would look at many of our current jobs as 'fake jobs', thinking that we are just playing for entertainment. Altman hopes that a thousand years from now our descendants will see their jobs as 'very fake', but incredibly important and satisfying for those who do them.

The road to distributed superintelligence

The strategy proposed by Altman to manage this transition has two fundamental pillars. The first is to solve the problem of alignment: to ensure that AI systems learn and act towards what we 'collectively really want in the long run', i.e., that AI acts for us, not against us (Altman cites social media feeds as an example of misaligned AI - incredibly effective at keeping us glued to the screen, but exploiting brain mechanisms that go against our long-term preferences). And then to make superintelligence cheap, widely available and not too concentrated in individual people, companies or countries. 'Giving users a lot of freedom, within broad boundaries for society to decide, seems very important,' Altman concludes.

'We are building a brain for the world,' Altman declares referring to the entire AI industry. A highly customised and easy-to-use brain for everyone, where the limit will be good ideas rather than technical capabilities. The final promise is bold: 'Intelligence that is too cheap to measure is well within our reach. And if current predictions seemed crazy in 2020, Altman reminds us that even describing in 2020 where we would be today would sound "probably crazier than our current predictions of 2030". The gamble is clear: to scale 'smoothly, exponentially, and without a hitch through superintelligence'. A gentle singularity that, if realised, would forever transform our relationship with work, creativity and progress itself.