Abundance, and what it would take

Before you read. "Abundance" is the most hopeful idea in the AI debate: that AI, together with cheap clean energy, could make many things that are scarce today (expert advice, good teaching, medical breakthroughs, reliable power) available to almost everyone. This article takes that hope seriously. It also treats it as a set of conditions, not a promise. Nothing here will happen by itself. For each benefit we set out what would need to go right, and what you could watch to see whether it is.

What "abundance" means here

In everyday speech, abundance simply means plenty. In the AI debate it has a more specific meaning: the idea that the cost of intelligence, and of the energy that powers it, could fall so far that things now limited by scarce expertise or scarce power become cheap and widely shared.

Two traditions use the word, and it helps to keep them apart:

• Technology-led abundance. The entrepreneur Peter Diamandis and the writer Steven Kotler argued in Abundance: The Future Is Better Than You Think (Free Press, February 2012) that fast-improving technologies could meet basic needs for far more people (search-verified). Today's AI leaders often speak in similar terms (see below).
• Building-led abundance. In Abundance (Avid Reader Press, March 2025), the journalists Ezra Klein and Derek Thompson argue that plenty depends on actually building housing, infrastructure and clean energy, and that rules and institutions often block this (search-verified). Their book is mainly about the United States, not AI, but it makes a point that runs through this article: abundance is a construction project, not a forecast.

Who makes the case, in their own words

The hopeful case is not a fringe view. It is made by people who lead some of the world's most advanced AI laboratories, and by at least one Nobel laureate. They also, notably, attach conditions to it.

Demis Hassabis, co-founder and chief executive of Google DeepMind, shared the 2024 Nobel Prize in Chemistry for AI-based protein structure prediction (Nobel Prize press release, 9 Oct 2024). Interviewed by 60 Minutes in April 2025, he described AI leading to what he calls "radical abundance", which the programme summarised as "the elimination of scarcity". On disease he said: "one day maybe we can cure all disease with the help of AI", adding that this was "within reach. Maybe within the next decade or so, I don't see why not." In the same interview he named two worries: "bad actors" repurposing AI "for harmful ends", and keeping control of increasingly autonomous systems (CBS News transcript, first broadcast 20 Apr 2025). → Tracker: CR-K24, CR-K25

Dario Amodei, chief executive of Anthropic, argued in "Machines of Loving Grace" (October 2024) that powerful AI could compress "the next 50-100 years of biological progress" into 5–10 years. He is careful to say that AI "isn't magic fairy dust" and that real-world constraints (experiments, regulation, human institutions) will slow things down (darioamodei.com). → CR-T02

Sam Altman, chief executive of OpenAI, wrote in "Three Observations" (February 2025) that "the cost to use a given level of AI falls about 10x every 12 months, and lower prices lead to much more use", and that "the price of many goods will eventually fall dramatically." He also wrote, just as plainly, that "increasing equality does not seem technologically determined and getting this right may require new ideas" (blog.samaltman.com). In "The Gentle Singularity" (June 2025) he predicted that in the 2030s "intelligence and energy… are going to become wildly abundant" (blog.samaltman.com). → CR-K26, CR-K05

Marc Andreessen's Techno-Optimist Manifesto (October 2023) describes the goal as making intelligence and energy extremely cheap, and argues that slowing AI has human costs (a16z). See The accelerationists. → CR-K02

A fair summary: the leading proponents believe abundance is possible, and most of them say in the same breath that it depends on safety, on sharing the gains, and on the physical world keeping up.

The evidence so far: what has actually got cheaper

The abundance case is not only a vision. Several of its building blocks have measurable track records. Here is what the best available data show, as of 28 September 2026.

What this shows, and what it doesn't. Costs of AI, solar and batteries really have fallen very steeply, and early studies of AI at work and in classrooms show real gains, especially for less experienced people. None of this yet shows abundance in the full sense. The education and work studies are single settings; the health evidence is thin; and falling prices for AI have been accompanied by much greater use.

The honest counterweights

A fair account of the abundance case must include the serious arguments against it.

• The economy may change more slowly than the technology. The MIT economist Daron Acemoglu estimated that AI's effect on total factor productivity (a broad measure of how efficiently an economy turns inputs into output) would be "nontrivial but modest": "no more than a 0.66% increase" over ten years, and possibly under 0.53% (NBER Working Paper 32487, 2024; published in Economic Policy, 2025). At the other end, Goldman Sachs economists projected in March 2023 that generative AI could raise global GDP by about 7% over ten years (search-verified). Both are forecasts. The gap between them is itself the point: nobody yet knows. → CR-K27, CR-K28
• Cheaper often means more, not less. As the price of AI falls, use rises. The Stanford AI Index 2026 puts AI data-centre power capacity at 29.6 GW in 2025, "about what it takes to power the entire state of New York at peak demand" (AI Index 2026). The International Energy Agency projects data-centre electricity use to more than double to about 945 TWh by 2030 (IEA, Energy and AI). → CR-N01
• Capability is uneven. The International AI Safety Report 2026 calls today's systems "jagged": excellent at some hard tasks, weak at some simple ones (executive summary, 3 Feb 2026). Robots, for example, succeed in only 12% of real household tasks such as folding clothes or washing dishes (AI Index 2026).
• Plenty doesn't share itself. The IMF has warned that AI could widen inequality without deliberate policy (IMF, 14 Jan 2024). Acemoglu's paper predicts AI will widen the gap between capital and labour income. Microsoft's own data show adoption of 28.8% in the Global North against 16.2% in the Global South (Microsoft, Sep 2026).
• Physical limits are real. Energy, chips, water, land, and the time it takes to run real-world experiments and clinical trials all cap how fast "intelligence" can turn into things people can use. Amodei's own essay makes this point.

Each hoped-for benefit, and what it would take

For each benefit below: what proponents hope for, what has happened so far, the preconditions that would have to be met, and indicators you could watch. Where a claim can be checked over time, it links to our public tracker.

1. Expert help for everyone

The hope. Good advice (on health, law, money, paperwork, a business plan) stops being a privilege. Anyone with a phone can get patient, accurate help in their own language.

So far. Use is spreading faster than any previous technology, and prices for a fixed level of capability have fallen by orders of magnitude. Arabic-first models exist in the UAE, Saudi Arabia and Qatar (Jais, ALLaM, Fanar). But the International AI Safety Report 2026 notes that systems can still make things up and give wrong answers with confidence, and that people can over-trust them (executive summary).

2. Faster science and medicine

The hope. AI shortens the path from question to cure: Amodei's "50–100 years" of biology in 5–10, Hassabis's "end of disease" within reach.

So far. AlphaFold is a genuine, Nobel-recognised breakthrough used worldwide. The AI Index 2026 reports AI moving "toward actual discovery", with AI-related publications in the natural, physical and life sciences up 26–28% year on year, and the first end-to-end AI weather-forecasting pipeline (AI Index 2026). Whether any medicine discovered mainly by AI has yet completed full regulatory approval is something we have not yet verified; it is on our list to check, and it is one of the clearest signs to watch.

3. A patient tutor for every learner

The hope. Every child and adult has a tutor that adapts to them, at almost no cost.

So far. The Nigeria trial is an encouraging result, but it is one six-week study in one place, and teachers guided the sessions, so the effect of the AI tutor cannot be separated from the effect of the whole programme. The UAE teaches AI as a subject from kindergarten to Grade 12, with about a quarter of the curriculum on ethics (Gulf News, 23 Aug 2025). Stanford reports that only half of US middle and high schools have AI policies and just 6% of teachers say those policies are clear (AI Index 2026).

4. Abundant clean energy

The hope. Cheap solar, storage and nuclear power, planned and run with the help of AI, make electricity plentiful and clean, including for AI itself.

So far. Solar and battery costs have fallen dramatically. The IEA judges that "concerns that AI could accelerate climate change appear overstated, as do expectations that AI alone will address the issue", and estimates that AI-based fault detection can cut outage durations by 30–50% (IEA). In Abu Dhabi, Masdar and EWEC reached financial close in July 2026 on a US$6.1bn project combining 5.2 GW of solar with 19 GWh of battery storage, designed to deliver 1 GW of power around the clock, and expected to be operational in 2027 (Masdar, 13 Jul 2026; Masdar describes it as a world first). → CR-K29

5. Higher productivity, broadly shared

The hope. People produce more in less time, wages rise, and new kinds of work appear.

So far. The gains in individual studies are real and often largest for less experienced workers. At the level of whole economies, the International AI Safety Report 2026 finds "no effect on overall employment" yet, but "some signs of declining demand for early-career workers in some AI-exposed occupations". Stanford researchers using payroll data report that employment of 22–25-year-olds in highly AI-exposed occupations is about 19% below where it would be had it kept pace with less-exposed peers (June 2026), while stressing that these are "descriptive patterns, not causal estimates" (Stanford Digital Economy Lab, Aug 2026). → CR-K30

6. Better, faster public services

The hope. Government that is quicker, fairer and available around the clock, in the citizen's language.

So far. Gulf governments have set some of the world's most ambitious public targets: the UAE aims to run 50% of federal government sectors, services and operations on agentic AI within two years of April 2026, with every federal employee trained and a stated principle that "people come first" (Dubai Media Office, 23 Apr 2026). Abu Dhabi aims to be "the world's first fully AI-native government across all digital services by 2027" (its own description) (DGE). → CR-G08, CR-G02

Where the Gulf fits: building for the age of superintelligence

Nobody can say whether or when AI far more capable than today's will arrive (see The singularity, explained). What the Gulf is doing is building the conditions under which abundance could land here if it does: energy, computing power, skills and public services.

• Energy. Barakah's four nuclear units generate, by the operator's account, up to a quarter of the UAE's electricity (ENEC, 5 Sep 2024). The round-the-clock solar-and-storage project above is expected in 2027.
• Computing. Large AI campuses are under construction, but under real stress. The 2026 regional war brought strikes on AWS data centres in the UAE and Bahrain, and Reuters reported a redesign of the 5 GW UAE–US AI Campus into dispersed, hardened sites (CNBC, 15 Sep 2026; Reuters, 11 Sep 2026). Stargate UAE's builder said in August 2026 that it was still on track to hand over capacity to customers in 2026, with two 100 MW buildings enclosed (Khaleej Times, 13 Aug 2026). As of 28 September 2026 we have found no official announcement that the first 200 MW is in service. → CR-C01, CR-C02
• Skills. AI is taught from kindergarten in the UAE (Gulf News, 23 Aug 2025); Saudi Arabia has introduced an AI curriculum for more than six million public-school students (Saudi Gazette, 24 Aug 2025); Microsoft has pledged to help train one million people in the UAE by the end of 2027. → CR-E04
• Dependence. Advanced chips arrive under US export licences with security conditions (The National, 20 Nov 2025). Abundance built on licensed hardware is real, but it is not the same as self-sufficiency.

Building for abundance is not the same as achieving it. The tracker exists so that readers can see the difference.

What we don't know yet

• Whether AI costs will keep falling at recent rates, or level off.
• Whether productivity gains in individual studies add up to faster growth for whole economies.
• Whether the benefits will reach lower-income people and countries, or concentrate.
• Whether energy, water and chip supply can keep pace with demand, especially in a region under geopolitical pressure.
• Whether any of the most ambitious forecasts (the end of disease within a decade, a century of biology in ten years) will prove right. These are sincere expert views, not measurements.

What this means for you

• You can be hopeful without being naive. Some of the building blocks of abundance (cheaper AI, cheaper solar, cheaper batteries) are already measurable facts.
• Ask "what would have to happen?" When you hear a big promise, look for the preconditions. Our tables are one way to do that.
• Benefits are already within reach in small ways. A free or low-cost AI tool can help you draft a letter, understand a document, or practise a language today. Check important answers, especially on health, law and money.
• Abundance is something societies build. Whether its gains are shared is decided by choices about training, access, energy and rules, and people everywhere can have a say in those choices.

Related: The accelerationists · Risks and the conditions they need · Will AI take my job? · Where are we now? (tracker)