The Abundance That AI May Promise Is Not Free

Opinion by: Merav Ozair, PhD, blockchain and AI senior advisor.

Elon Musk and Peter Diamandis support the idea that “everything will be free.” They purport to believe that AI abundance will end poverty and provide a universal high income.

Others in the mega tech ecosystem mention the coming abundance. Demis Hassabis, for example, says AI could spark a “renaissance” of “radical abundance.”

Politicians at the World Economic Forum 2026 in Davos liked Musk’s vision. They were thrilled that their economic problems would soon be “set free.” This story is quite appealing. Who doesn’t like to get things for free? 

What does it truly mean? Would all economic activities have no cost? Would all corporations become altruistic and seek no profit?

Let us unpack the narrative.

The cost of production can be cheap, but never zero

Let’s put things in perspective. In the age of AI abundance, products and services will not arrive out of “thin air.” They would still need labor, materials, energy and infrastructure.

The advances in AI and other emerging technologies may lead to very cheap energy and highly automated production. This evolution will result in the marginal cost of most digital and even physical goods approaching zero. 

This is due to three main factors. First is the automation of labor, where machines and AI handle almost all production, logistics and many services. The second is advanced manufacturing and AI distribution, like 3D printing, robotics and AI logistics systems that drastically reduce waste and inventory, making “enough for everyone” technically feasible. Lastly, abundant energy — fusion or ultra‑cheap solar makes energy so affordable that it stops being the bottleneck. 

Because energy underlies everything physical, all other costs fall. 

Plans are already in place. Elon Musk is now prioritizing lunar manufacturing and AI, with a goal of over 1,000 gigawatts of solar power. Using solar energy instead of nuclear power will reduce energy cost to almost zero. The catch: the initial cost to establish the infrastructure on the moon is very high, and it would need to overcome major challenges.

Related: Energym AI dystopia goes viral as crypto projects tout user-owned AI agents

Under those conditions, it is plausible that education resources become somewhat free to the user because they are AI‑generated and infinitely replicable once the system is built. A large fraction of healthcare becomes extremely cheap, once the appropriate AI and robot infrastructure exists. 

At the level of physics and engineering, if the real bottlenecks — energy and automation — are abundant, costs collapse, but they do not completely disappear.

Infrastructure is the missing layer that no one talks about

Robotics and energy need to run at scale and speed to create an “abundance” of everything for everyone. For this, it needs infrastructure.

Automation and robotics run on what Jensen Haung calls “AI factories.” This is AI infrastructure, representing a shift towards treating AI development as an industrial process, enabling organizations to continuously train and refine AI models for better safety and efficiency.

They are specialized, high-performance computing data centers designed to “manufacture” intelligence by converting raw data into trained AI models and tokens, rather than simply storing data. Using advanced GPUs and massive interconnected infrastructure, they are the engines of AI applications such as autonomous vehicles, robotics and generative AI.

AI factories are expensive. They need a lot of money to build and run. Companies that have already set up the infrastructure will keep growing and improving. For example, Nvidia is five times more profitable than IBM was in the 1980s, with only a tenth of the staff. Productivity and profits will increase, because AI greatly boosts efficiency. Investments will go to those who own AI models, platforms and especially the infrastructure.

This will lead to the biggest concentration of wealth in history.

Major players include tech giants like Nvidia, AWS and SpaceX. They will continue to dominate the market, making it tough for newcomers to compete.

Governments are also involved. China is using its huge solar energy capacity to boost the energy-heavy AI boom. This creates a unique “AI and energy” ecosystem. Here, artificial intelligence optimises renewable energy generation, while solar power supports data centres. China is seen as a leader in renewable energy use.

Cheap energy is not cheap 

Energy is the fuel that runs AI factories, which are the engine of all robotics, automation and AI applications that will generate abundance. Energy fuels the infrastructure, and infrastructure runs the AI applications. Therefore, energy is the real bottleneck. Without cheap energy, this “free” theory fails.

Currently, electricity is the primary form of energy used to run the infrastructure. China is aggressively integrating renewable energy into its infrastructure and other regions are expanding renewable-powered energy into data centers as well. Electricity generation and grid capacity for AI-scale infrastructure is very costly and not scalable. To reach abundance at scale, energy must be very cheap and scalable.

What are the options?

Fission energy is a type of or nuclear energy. It is fully mature, providing stable power, but produces radioactive waste. It carries the risk of nuclear proliferation, and safety concerns regarding meltdowns. It is cheaper than current fossil-based electricity sources but still has a tangible cost, and, like the other electricity sources it is limited, and not scalable.

Fusion energy involves merging light atoms to create energy, mimicking the sun, while traditional nuclear energy splits heavy atoms. Fusion offers nearly limitless, cleaner energy without long-lived high-level waste. 

Fusion is inherently safer with no risk of a runaway chain reaction. 

The caveat, however, is that fission is what’s currently being used. Creating nuclear fusion for energy is extraordinarily expensive and requires upfront investments of hundreds of billions of dollars, and it is still experimental and likely decades away from large-scale commercial use.

Unlike nuclear fission, nuclear fusion is scalable. It is cheap but not does not cost zero. Someone has to pay the upfront costs to build the infrastructure, to create it and then maintain it. 

Elon Musk is going to the moon

Lunar solar power provides ample energy without atmospheric issues. Yet, it has high costs for launching, building and maintaining in a vacuum. Musk’s plan is to move all production, including the AI factory, to the moon.

The moon has low gravity and plenty of resources, making it the cheapest place for AI infrastructure.

Robots will terraform and build infrastructure. Humans will come to oversee and expand, while AI data centres will fuel the space economy.

With Starlink, SpaceX, Optimus robots and xAI, Musk is in a strong position to make this happen.

However, machines for making advanced AI chips need to reach the moon. These bus-sized machines require very precise conditions.

The solution is a new method called Atomically Precise Manufacturing (APM). This builds atom by atom and aligns with Musk’s “first principle” thinking.

If successful, this could unlock unlimited solar energy and raw materials from the moon and asteroids. There would be no thermal limits or atmospheric interference.

This could lead to boundless AI at a low cost. Experts say that if lunar fabrication works, it could create a trillion-dollar, or even hundreds of trillions, opportunity.

Who will benefit most from this hundred-trillion-dollar chance? Will it be shared fairly?

The soft prison of “free”

When you have centralized infrastructures and systems, whoever owns the infrastructure sets the terms of engagement. Strongly centralized systems can provide extensive “free” services, but in exchange, they often demand high control over speech, movement, data and economic choices. Non‑authoritarian welfare states may trade some individual autonomy for security and guaranteed services. Many “free” digital services today are funded by surveillance, profiling and behavioral manipulation — your data and attention are the real price. 

In a world of AI abundance, the infrastructure may be government owned. It may be owned by corporations. It could be owned through a public-private partnership. Either way, the infrastructure is centralized and the centralized power will dictate the distribution terms — how AI abundance is distributed, who gets what, under what conditions. If they wish to, they can abruptly “shut the valve” and nothing is distributed either to an individual or a group. Your dependency on their services becomes a “soft prison” stripped of your autonomy and self-sovereignty.

It might be a hundred-trillion-dollar opportunity, but the owner of the centralized infrastructure will get the lion’s share and will dictate what will trickle down to the masses.

They say if something is “free”, you are the product. This remains true in a world of sheer abundance. In that world, the product is your self-sovereignty.

Opinion by: Merav Ozair, PhD, blockchain and AI senior advisor.