Economic power is returning to the physical realm

FOR more than a decade, the assertion by venture capitalist Marc Andreessen that “software is eating the world” captured the logic of economic power. Digital platforms reshaped industries, software firms eclipsed manufacturers, and competitive advantage appeared increasingly detached from physical constraints.

That world view influenced corporate strategy and public policy alike. It is now breaking down. Software remains essential, but it is no longer scarce.

Artificial intelligence (AI) has turned software into an abundant input. Code can be generated, tested and optimised at next to no marginal cost. In many applications AI does not merely improve software; it replaces it. As software becomes commoditised, it loses its power to differentiate.

What is becoming scarce again is hardware – broadly defined to include infrastructure, energy systems, materials, manufacturing capacity, transport and supply chains.

It follows that hardware is becoming a more important component of the global economy, since economic power accrues not to what is most elegant, but to what is most constrained. The current volatility in oil prices and shipping costs is a stark reminder of the powerful impact physical constraints – in this case, the Strait of Hormuz – have on the global economy.

Software scales almost without friction. Hardware does not. Every additional data centre, factory, transmission line or robot requires land, energy, skilled labour, permitting and time. Marginal costs are unavoidable. Lead times are long.

These constraints are once again binding across the global economy. Energy makes the shift visible. Electrification, AI and cloud computing are driving a surge in electricity demand.

The International Energy Agency (IEA) estimates that electricity consumption from data centres could more than double by 2030. At the same time, global power grids are ageing and under-built. According to the IEA, annual investment in electricity networks must nearly double this decade to meet climate targets.

Such expansion requires a lot of materials. S&P Global forecasts that demand for copper could rise by 50 per cent by 2040, driven largely by electrification and grid expansion. Transmission alone will require tens of millions of kilometres of new cabling worldwide. Software offers no substitute.

Climate change reinforces the trend. Much of the world’s infrastructure was built for more stable weather. Power networks, transport corridors and water systems must now be hardened against extreme conditions. These investments are slow, capital-intensive and politically contentious.

Defence adds urgency. After decades of under-investment, Western countries are scrambling to rebuild military capacity. Recent conflicts have revealed an array of problems with “industrial throughput”: the ability to produce hardware, from missiles to spare parts, quickly and efficiently.

Rearmament is partly a software problem, but it depends crucially on factories, machine tools, chemicals, skilled workers and reliable supply chains – many of which have been allowed to atrophy.

This has geopolitical consequences. Industrial capacity confers strategic autonomy. China has understood this for years. It combines scale with deep process know-how, dominating large parts of battery manufacturing, solar equipment and critical-materials processing. In rare earths used in magnets, it accounts for roughly 60 per cent of global mining output and about 91 per cent of processing. Similar concentrations exist across other industrial inputs.

Other economies retain important strengths. Taiwan, Japan and South Korea dominate critical segments of semiconductor manufacturing. Germany and Switzerland remain leaders in advanced machinery and industrial processes. These advantages rest not on software alone, but on decades of accumulated engineering expertise.

America is attempting reindustrialisation through subsidies and industrial policy. It may yet succeed. But manufacturing capacity is cumulative. It depends on tacit knowledge, skilled labour and dense supplier ecosystems. Software can be reshored in months. Factories cannot.

The idea of a weightless digital economy has always been misleading. AI runs on data centres that consume vast quantities of electricity and water. Transport remains stubbornly physical, reliant on ships, ports, railways and lorries. Even technologies often presented as dematerialising, such as 3D printing, increase the importance of advanced materials, feedstocks and energy rather than freeing processes from them.

Mining, long neglected by technology investors, is becoming strategically central again. Yet, extraction is only part of the story. Processing is the true choke point – and the most geographically concentrated part of the supply chain. Control over processing increasingly determines control over downstream manufacturing.

AI and quantum technologies will speed up innovation in materials science, producing new alloys, compounds and manufacturing techniques. Paradoxically, this will increase, not reduce, demand for physical inputs. As technology reshapes materials themselves, predicting future demand becomes harder, raising the premium on resilience and industrial depth.

Software still matters. But as it becomes commoditised, economic power is flowing back to the physical layer of the economy. Control over energy, materials, infrastructure, robotics and industrial capacity will define competitiveness in the coming decade.

This shift has implications for investors, companies and governments. It favours long-term capital over rapid scalability, engineering depth over digital polish, and resilience over optimisation. It also helps explain why manufacturing and resource security are regaining strategic importance after decades of neglect.

The next era of global competition will be decided not by who writes the best code, but by who can build, power and supply the physical systems on which the digital world ultimately depends.

©2026 The Economist Newspaper Limited. All rights reserved

The writer is a tech investor and strategic adviser to Divergent Technologies. He was chairman of the supervisory board of Deutsche Bank from 2012 to 2022.