Tech leaders often describe a future in which AI makes goods and services essentially free — a world of abundance that ends scarcity and lifts incomes. It’s an alluring vision, but ‘free’ rarely means zero cost. We should separate falling marginal prices from the full set of economic, technical and political realities behind production and distribution.
Lower marginal costs, not zero costs
Advances in AI, automation, robotics and advanced manufacturing can drive marginal costs for many digital and physical products toward near zero. Software, content generation and highly automated production can reduce the labor and waste that used to drive per-unit prices. But that does not erase all costs. Materials, energy, skilled labor for maintenance and design, shipping, and the capital to build and run systems remain real expenses.
The unanswered layer is infrastructure
Delivering abundance at scale requires powerful, specialized infrastructure: data centers, training clusters, robotic factories, precision manufacturing equipment and resilient networks. These are costly to build, power and maintain. Jensen Huang’s image of ‘AI factories’ is useful: raw data and compute are converted into deployed models and robotics that actually produce value. Whoever builds and controls these factories gains outsized advantages — productivity, market share and profit — and wealth becomes concentrated in the hands of infrastructure owners.
Energy is the limiting factor
Energy is the fuel of this economy. Current grids and generation capacity are expensive and finite compared with the appetite of large-scale AI and automation. Mature options like fission have trade-offs (waste, safety, political issues); fusion remains experimental and distant. Massive solar or other renewables could help, but deployment and storage at planetary scales require enormous capital. Energy will be cheap only in relation to who pays for the capital and ongoing upkeep — it won’t be literally free.
Off‑Earth production is possible but costly
Ambitious concepts like lunar-based manufacturing or solar farms in space could unlock huge resource and energy advantages. Low lunar gravity and local resources might let robots build infrastructure that is infeasible on Earth. But the technical challenges and upfront costs to establish and operate such off-world systems are staggering. If they succeed, the value will accrue primarily to those who own the infrastructure and the supply chains that support it.
The hidden cost of ‘free’ is control
Centralized infrastructure implies centralized control. Whether run by corporations, states, or hybrid partnerships, owners define access, pricing, rules and distribution. ‘Free’ consumer services already exact payment in attention and data; in a world dominated by a few infrastructure owners, the price could be deeper: surveillance, behavioral steering, limits on speech, or even the ability to shut off essential services. Dependence on a single provider creates a soft prison where guaranteed access comes at the cost of autonomy and self-sovereignty.
What matters next
Abundance in production does not automatically deliver equitable access. The major economic question is not only whether goods can be produced cheaply but who owns the means of production and how benefits are allocated. If infrastructure and energy remain concentrated, the large gains from AI and automation will likely accrue to the owners, not to everyone.
Policymakers, technologists and civil society should focus on governance, public investment, open platforms and ownership models that prevent monopoly control of critical infrastructure. Otherwise, the promise that ‘everything is free’ may simply be a reframing: when something seems free, you — or your autonomy — may be the product.
Opinion by Merav Ozair, PhD, blockchain and AI senior advisor.