Science does not understand biology

There is a category error running through the centre of the AI revolution. It is not a technical error or a financial error. It is a philosophical one, and it is the most expensive philosophical error in four hundred years of human civilisation.

It forces us to confront a fundamental question: can you build an exponential digital future on top of a physical supply chain bound by biological and geological time? What happens when the timeline of a spreadsheet collides with the timeline of the earth's crust? We are moving toward a wall at extraordinary speed, with our eyes fixed firmly on the dashboard.

The substrate tells the truth

When we examine the capital expenditures of the leading cloud and infrastructure pioneers, we are not looking at isolated corporate strategies; we are looking at the vanguard of a systemic economic inversion. A single data centre built in Chicago required 2,177 tonnes of copper. That is 27 tonnes per megawatt of installed capacity, and that facility was built in 2009, long before AI workloads existed at scale. The racks running large language models today draw between 20 and 60 kilowatts each, compared to the 4 to 6 kilowatts of a legacy data centre. The copper requirement per site is not growing incrementally; it is compounding.

By 2035, the cumulative copper locked into data centres alone could surpass 4.3 million tonnes. Experts project copper demand from AI-powered facilities will average approximately 400,000 tonnes annually over the next decade, peaking at 572,000 tonnes in 2028.

Global mined copper supply reached 22 million tonnes in 2024. Supply is projected to peak in the late 2020s at just over 24 million tonnes, before falling below 19 million tonnes by 2035 as ore grades decline and mines close.

The numbers do not reconcile. They belong to different centuries. An AI data centre can be built and operational in roughly 18 to 23 months. A new copper mine takes an average of 17.9 years to move from discovery to production. For the most recent cohort of mines still in development, that figure has surged to 28 years. One project, discovered in 1990, is not expected to begin operations until 2030 or later. Another, if it meets its timeline, will have taken 37 years from discovery to first production.

If an AI data centre requires less than two years to build, but the copper mine required to feed it takes nearly three decades to permit and extract, where does the missing physical matter come from? Can capital compress geological time? Can a venture raise accelerate the rate at which rock yields to a drill? If the spreadsheet and the substrate are speaking completely different languages, which one do we expect to break first? This is not a logistics problem. It is a physics problem. You cannot permit a mine faster than the rock allows you to reach it. You cannot accelerate the biological and hydrological systems that make extraction viable in the first place. Industry outlooks already warn of a 25 to 30 percent copper shortfall by 2035.

The greatest admission

The leadership of OpenAI has publicly stated that the primary bottleneck for AI is no longer chips or talent, but infrastructure constraints: raw energy and compute capacity. The proposed solution is to use AI itself to solve the problem.

Read that again. The machine that requires the infrastructure will design the infrastructure the machine requires. This is not a strategy. It is a proof of the very ignorance it is meant to resolve. When the most powerful voices in artificial intelligence identify the problem as physical and then propose a digital solution, the category error is complete.

This operational blindness extends across the entire infrastructure landscape. Billions of dollars are being committed monthly by major tech ecosystems to secure computing capacity running to mid-2029. But when an enterprise capitalises billions for localised digital infrastructure, where do the millions of gallons of daily evaporative cooling water actually come from? If that water is drawn from a hyper-local aquifer, what happens to the root-zone moisture of the surrounding topsoil? If an organisation balances its ledger by funding a watershed restoration project three hundred miles away next season, does that water retroactively flow back under the local community's feet today? Are we measuring ecological continuity, or are we simply trading a local physical crisis for a macro corporate asset? The biological systems sustaining the regions in which they are built do not appear in the investment memorandum. They appear, eventually, in the drought reports. By then, the capital has already moved.

The inversion

On 12 June 2026, SpaceX completed the largest IPO in history at a valuation approaching $1.8 trillion.

Pollinators, according to scientists, contribute between $235 and $577 billion in annual global food production. Trees contribute approximately $1.3 trillion annually to the global economy through timber, carbon sequestration, agricultural yield support, and urban energy and healthcare cost reduction.

The organisms and systems that make agriculture, water, breathable air, and therefore human existence possible are not on any exchange. They carry no ticker symbol. They appear, when they appear at all, as a page at the back of a sustainability report.

The most valuable company in the world listed at one third of the annual economic contribution of every bee, bird, bat, and butterfly on the planet, offering a product whose physical existence depends entirely on systems it does not price, does not measure, and does not protect.

This is not a metaphor for misallocation. This is misallocation.

The world we inhabit is not merely unbalanced. It is inverted. We have placed at its apex the thing with the least reach and the shortest duration. We have placed at its base the thing without which nothing above it can exist.

Scarcity has arrived not because nature ran out. Nature did not run out, we simply stopped accounting for it. We built the most sophisticated measurement systems in history and pointed them at the things that matter least.

The recovery path

There is one, but it is not the one currently being discussed.

The recovery path is not more compute, a bigger grid, a new rare earth mine in a jurisdiction where the permits move faster. It is not a deal structure that pushes the hydrological risk to the next quarter's report. The recovery path runs through the thing that has been treated as a cost centre, a compliance obligation, a box to tick before the real work begins.

What has been called sustainability is not the edge of the business. It is the foundation of all business. What has been called natural capital is not a nice-to-have in a shareholder letter. It is the only capital that was ever real.

The timber, the water, the soil, the pollinator, the mycorrhizal network beneath the forest floor: these are not resources waiting to be extracted. They are the substrate on which every other form of value is built, and they operate on timelines that no algorithm can compress and no capital raise can accelerate.

Biological limits determine cycles. Those cycles are non-negotiable. They were non-negotiable in the fourteenth century and they are non-negotiable now. The only variable is how long the lag runs between the biological signal and the financial recognition.

Those who do not understand substrate will forfeit their spreadsheet.

The shift

We were told the shift would produce a thousand billionaires. Perhaps it will. But the shift that matters is quieter and more structural, and it is already under way.

There is a form of intelligence that predates artificial intelligence by four billion years. It does not run on copper or silicon, it does not require a data centre or a permit. It has solved, across geological time, every problem that AI is now being asked to solve: energy efficiency, resource allocation, waste elimination, systemic resilience, adaptive response to environmental change.

It is called Natural Intelligence, the intelligence of living systems. The only substrate on which the future can be built.

The greatest industry of the next century is indeed AI, Adaptation Intelligence.

How poetically and precisely ironic that it is the machine built to replace nature that will force us, finally, to remember it.

illuminem Voices is a democratic space presenting the opinions of leading Sustainability Thought Leaders, their views do not necessarily represent those of illuminem.

The world needs sustainability knowledge. At illuminem, no interest group or shareholder can influence our work. Thank you for supporting our mission to make high-quality and independent sustainability information free for all. Every contribution helps. Thank you for donating today.