Blogs / The Great Value Inversion
The Great Value Inversion

Design · 5 Min Read

The Great Value Inversion

As Intelligence Becomes Cheaper to Produce, Who Will Control What Remains Scarce?

More access, whose authority?

Between November 2022 and October 2024, the price of querying an AI model matching GPT-3.5’s performance on the Massive Multitask Language Understanding benchmark fell from $20 to $0.07 per million tokens. Stanford’s AI Index recorded a reduction of more than 280 times. This was not a measurement of human intelligence becoming cheaper. It showed how rapidly access to a specified level of model performance became less expensive. [web:50]

For people producing words, code, interfaces, or analysis, the change raises an uncomfortable question. If some outputs associated with expertise become easier to reproduce, where might economic advantage accumulate?

Cheaper generation does not settle what happens to expertise or wages. In Denmark, Anders Humlum and Emilie Vestergaard found rapid AI adoption and changes in work, but no statistically significant average effects on earnings or recorded hours across 11 exposed occupations, using records through December 2024. Their March 2026 working paper does not promise permanent protection from displacement. It shows why changing tasks and changing compensation must not be treated as the same event. [web:83]

A different tension concerns the systems producing these capabilities. Wider access and unequal authority could develop together. More people might use sophisticated tools without gaining influence over the infrastructure, finance, technology, or commercial arrangements supporting them. The benefits of access would be real. So would the limits of participation.

I approach this question from work across digital experiences and mineral systems. The interface and its material conditions are not separate worlds. I think of their possible reorganization as the Great Value Inversion: not a demonstrated economic law, but an inquiry into what happens when increasingly reproducible outputs depend on systems that remain difficult to expand or influence.

For mineral-producing regions, the consequential question is whether becoming more important to technological production also creates greater authority over its terms.

Every algorithm lives in hardware

Every algorithm lives in hardware.

For an algorithm to do anything in the world, physical systems must execute it. An apparently immediate response rests on machines, electricity, networks, cooling equipment, buildings, land, and labor.

Kate Crawford’s Atlas of AI helped establish this material and political perspective, connecting AI with mineral extraction, human work, data acquisition, and concentrated power. AI’s physical foundations are not a new discovery. What matters here is how dependence on them may shape economic relationships. [web:30]

The International Energy Agency’s 2026 assessment estimates that data centers consumed 485 terawatt-hours of electricity in 2025, with its central projection reaching approximately 950 terawatt-hours in 2030. These figures cover data centers generally, not AI alone. The latter figure is a conditional projection, not an observed outcome. [web:131]

The systems supporting this expansion cannot all develop at software speed. Grid connections, advanced chip manufacturing, memory, equipment, and capital can constrain it. Different components require different materials, many also used in transportation, electrification, and other industries. AI is one source of demand within larger systems, not an explanation for every mineral-market change. [web:88][web:131]

My mineral-trajectories perspective follows these relationships across extraction, processing, manufacturing, use, maintenance, and disposal. An interface-level account can explain what a product enables while saying little about whose work, infrastructure, or environment makes it possible. I do not think we can understand technological development fully if we stop where the user’s experience begins.

This does not mean every application translates into additional mining. Demand passes through existing capacity, inventories, contracts, investment decisions, and competing uses. Those relationships determine how a digital choice becomes a material consequence.

They also complicate the assumption that essential materials must become increasingly expensive.

Strategic does not necessarily mean expensive

The IEA’s Global Critical Minerals Outlook 2025 reported that lithium demand grew by nearly 30% in 2024. Yet lithium prices had fallen by more than 80% since 2023. Graphite, cobalt, and nickel prices also declined by 10–20% in 2024. Expanding supply pressured prices despite demand growth associated largely with broader energy applications. [web:44]

Strategic importance does not guarantee sustained price appreciation.

A material can be essential but well supplied relative to current demand. It can be inexpensive while processing capacity remains concentrated. A disruption can be consequential even when the affected input normally accounts for a small cost.

Nor does geological availability establish usable supply. A deposit must be developed, financed, processed, transported, and connected to buyers. The constraint may sit in an industrial process, an electricity connection, or a commercial arrangement rather than underground.

The IEA reported that the average share of the top three refining countries for key energy minerals increased from around 82% in 2020 to 86% in 2024. Supply expanded while the industrial capacity transforming materials became more concentrated. [web:44]

Scarcity is not a substance leaving human minds and settling in the earth. Different constraints become visible as production changes. The important distinction is between contributing an essential input and controlling a relationship through which value can be captured.

That distinction changes the development question. Rather than asking only whether mineral prices will rise, we need to ask what producing regions can negotiate, build, and govern.

Being necessary is not the same as being powerful

Possessing minerals is not the same as extracting them. Extraction does not confer processing capability. Processing does not necessarily provide technology ownership, market access, or control over contracts. Profitable production does not establish that public institutions or nearby communities receive an adequate share of its benefits.

Zambia makes the issue concrete. Copper supports electrical infrastructure, but its relevance extends far beyond AI. UNCTAD’s February 2026 assessment describes a major copper-producing economy where commodity dependence has delivered limited job creation and continuing exposure to volatility and external shocks. [web:42]

The assessment identifies potential opportunities in copper articles, machinery, industrial chemicals, fabricated metals, and related services. These are possibilities for diversification, not completed achievements. Finance, infrastructure, technology, skills, and market access affect whether firms can pursue them. [web:42]

This is a different proposition from waiting for copper prices to rise. It concerns converting mineral-related activity into durable productive capabilities. Supplying industrial services or equipment may build knowledge and commercial relationships without requiring an immediate leap from a mine to a semiconductor factory.

It does not follow that every mineral should be processed locally. Commercial viability, electricity requirements, environmental consequences, and public financial exposure still matter. A project can promise value addition while transferring risks that its prospective benefits do not justify.

Higher export revenue can support public investment. It does not by itself reveal the durability or distribution of gains. A country can strengthen its negotiating position while communities near extraction sites remain unable to influence land use, protections, compensation, or revenue decisions.

National authority and community authority are related, but they are not interchangeable.

The provocative possibility is that a region could become more necessary to technological production without gaining proportionate influence over its terms. Avoiding that outcome requires more than recognition of a resource’s importance. It concerns the capabilities, contracts, and institutions connecting it to what follows.

What are we actually designing?

This brings the investigation back to technological development.

A feature is not only an interface decision. Depending on its scale and implementation, it can become a recurring demand on computing resources, a procurement requirement, or part of the justification for new infrastructure. Its material implications depend on how those relationships operate, not simply on the feature’s existence.

The significance is not that every designer should become a commodity analyst. It is that the boundaries used to evaluate a product can hide some of what the product helps organize.

Choices about capabilities, frequency of use, hardware maintenance, repair, and useful lifespan can participate in shaping demand. Their effects require evidence; identifying a possible influence is not the same as measuring it.

For me, the useful position lies between claiming that designers control these systems and treating material consequences as someone else’s concern. I want to understand where a design decision enters a larger arrangement, what it helps make routine, and which consequences remain outside the authority of the people making it.

Designers do not independently control ownership, financing, or industrial policy. Their influence operates through organizations and institutions. But those limits do not make the relationships irrelevant to design. They help identify where a product decision ends and a question of procurement, investment, or public authority begins.

The Great Value Inversion, then, is not a prediction that minerals defeat algorithms. It asks whether the conditions of producing increasingly accessible cognitive capabilities will alter who can shape technological development.

There is no necessary outcome. Greater access could coexist with concentrated infrastructure power. Some producing regions could negotiate stronger capabilities and public benefits; others could contribute more while remaining dependent on external finance, processing, and buyers. These possibilities can overlap, including within the same country.

A meaningful change would concern not only the volume of material supplied or the revenue earned, but the ability to influence the relationships producing those outcomes. That includes whether affected communities participate in consequential decisions rather than merely absorb their effects.

If the material foundations of intelligence become more important, will the people and places sustaining them gain only a larger role in producing technological value, or greater authority to decide who benefits from it?.

Selected sources