One question shapes the Davos Super Intelligence Summit 2027. What each of its four panels asks, and the public data that explains why.
The summit's FAQ puts its purpose in one sentence: "How do we turn increasingly powerful intelligence into shared prosperity, while preserving human agency and earning society's trust?"
Shared prosperity is the aim. Human agency and public trust are conditions on how it is reached: people keep real control over decisions, and the public comes to trust the systems involved. A system can raise output and still fail either condition, and the question is phrased so that the aim cannot be separated from them.
The same ideas run through the OECD AI Principles, adopted in 2019, updated in May 2024 and now backed by 47 adherents. Principle 1.1 lists among its aims "reducing economic, social, gender and other inequalities". Principle 1.2 says "AI actors should implement mechanisms and safeguards, such as capacity for human agency and oversight". The principles leave open how a company, a fund or a ministry turns this into decisions about budgets, products and people. The summit's four panels take that step for agents, robots, infrastructure and social impact in turn.
Public opinion surveys give a rough measure of both conditions. In surveys fielded in 2024 and published by the Pew Research Center in April 2025, 73% of AI experts in the United States expected AI to have a positive impact on how people do their jobs over the next 20 years. Among US adults the figure was 23%. The two groups did agree on one thing. More than half of adults (55%) and of experts (57%) "say they want more control over how it is used in their lives". That is the human agency condition, in the public's own words.
The summit describes AI as having moved from a software conversation to a board-level infrastructure decision. Stanford's AI Index 2026 supplies the numbers. Organisational adoption of AI reached 88%. Generative AI reached 53% population adoption within three years, at a pace that varies by country and tracks GDP per capita. Documented AI incidents rose to 362, from 233 the year before.
For boards, these numbers move AI from the innovation budget into risk and operations. The incident count rose by more than half in a year, and the gap in adoption between countries matters to anyone operating across markets. Stanford HAI states the stake directly: AI's benefits "won't be evenly distributed unless we guide its development thoughtfully".
On the choice of place, the FAQ says Davos "is the rare setting where capital, policy, enterprise demand, scientific credibility and public narrative are concentrated in one week".
The first panel deals with systems that act. An agent is software that plans and carries out a task on its own, such as filing a claim, changing a booking or running code. The Journal's piece on AI agents vs agentic AI covers the terms in more detail.
In the words of the AI Index, "AI agents made a leap from 12% to ~66% task success on OSWorld, which tests agents on real computer tasks across operating systems, though they still fail roughly 1 in 3 attempts on structured benchmarks." A system that succeeds two times in three is useful. It also needs someone who can check it, stop it and answer for it.
In this panel, human agency comes down to two practical points: who approves what an agent does, and who is accountable when it gets something wrong. The longer horizon in the panel's title, artificial general intelligence, asks the same thing at a larger scale.
The second panel covers AI in machines that move and handle things. The International Federation of Robotics counted a record 5 million industrial robots in operation worldwide in 2025. The Journal's guide to physical AI sets out the full figures and what they leave out.
A machine's mistake on a factory floor, in a warehouse or in a hospital has a physical cost, to goods or to people. That changes what buyers need to know: how the system was tested, who is liable, and how the people working beside it are trained. Trust here is built from test results and safety records.
The third panel covers what AI runs on. The summit lists four subjects under it: AI chips and compute, data centres, energy and scaling, and sovereign AI infrastructure.
The AI Index shows how concentrated the first two are. It counts 5,427 data centres in the United States, more than ten times any other country, and finds that a single company, TSMC, fabricates almost every leading AI chip. Then there is power. Data centres used about 485 TWh in 2025, roughly 1.5% of global electricity consumption, according to the International Energy Agency, which projects about 950 TWh in 2030. The Journal's piece on AI and climate change covers what that demand means for emissions.
The AI Index also records a policy response, with national AI strategies expanding and state-backed investment in AI supercomputing rising, "a sign of growing ambitions for domestic control over AI ecosystems", while model production stays concentrated in the US and China. Access to compute and power affects who can share in the gains, and the OECD's commentary on Principle 1.1 names that risk: "This stewardship role should strive to address concerns about inequality and the risk that disparities in technology access increase existing divides within and between developed and developing countries."
The fourth panel covers social impact. On work, the International Labour Organization finds that one in four workers worldwide is in an occupation with some exposure to generative AI, and it sees the transformation of jobs, more than their full automation, as the most likely impact.
The public is less sure. Pew's 2026 global survey reports:
"In 34 of 37 countries Pew Research Center surveyed earlier this year, people tend to believe AI will lead to fewer jobs rather than more jobs."
On inequality, Pew found no country in which more than one adult in four expects AI to narrow the gap between rich and poor. On both jobs and inequality, these surveys record more worry than hope.
Before Davos, an executive could put one question from each panel to their own organisation:
A logistics group and a bank will answer these differently, and so will companies under different regulators. That spread of answers is the reason to compare them with people from other sectors.
The Davos Super Intelligence Summit 2027 is an invitation-only summit held on Tuesday 19 January 2027, from 16:00 to 20:00 CET, at the Grandhotel Belvedere, Promenade 89, 7270 Davos, and organised by Davos Premier. Directions, access and invitation requests are covered in the Journal's guide to finding the summit.
The future of AGI and agentic AI; physical AI and robotics; AI infrastructure, which covers AI chips and compute, data centres, energy and scaling, and sovereign AI infrastructure; and social impact through innovation. Each takes up one part of the summit's question, as the sections above explain.
Decisions about AI now involve capital, regulation, enterprise buyers, science and public opinion at the same time. The summit FAQ describes Davos as the rare setting where all of these are concentrated in one week. With organisational adoption of AI at 88% according to the AI Index, few boards can leave the subject off their agenda.
The name uses the word, and the summit's question speaks of increasingly powerful intelligence. The programme approaches it through four practical panels, from agents and robots to compute and social impact. For what superintelligence means and how it differs from AGI, see AGI vs ASI.
The programme splits the summit question across four conversations on the evening of 19 January. The topics listed under each one are in the programme.