For the past two years, the AI story has been told almost entirely through upside. Faster productivity. Smarter software. A new growth cycle. But buried inside that optimism is a number that deserves far more scrutiny: $4 trillion.

That is how much analysts now estimate could be spent on AI data-centre infrastructure by the end of this decade. Not on apps or models, but on the hard, physical backbone of AI: chips, memory, power, land, cooling, and construction.

The question markets are only beginning to ask is simple: is the global economy ready to absorb that level of demand without consequences?

A build-out without precedent

AI has quietly turned Big Tech into one of the most capital-intensive industries in the world.

Hyperscalers are racing to secure capacity before rivals do. Data centres are being planned and built at a pace normally associated with wartime mobilisation or national infrastructure drives. Unlike past tech waves, this one cannot scale purely through code. It scales through factories, grids, and supply chains.

That matters, because those systems have limits.

The early warning signal: memory chips

One of the clearest stress points is already visible in memory markets.

DRAM inventories across major suppliers have been falling sharply since late 2024. Data centres consume enormous amounts of high-performance memory, and demand is rising faster than new capacity can come online. This is not a demand blip. It is structural.

As inventories shrink, prices rise. Those higher costs then flow straight into AI infrastructure budgets, and eventually into margins, pricing decisions, and investment returns.

This is how inflation begins in capital-heavy booms: quietly, upstream, long before it shows up in consumer data.

Bottlenecks create inflation, not innovation

AI spending is hitting multiple choke points at once:

• Chips: advanced logic and memory are both supply-constrained
• Power: electricity demand from data centres is surging faster than grid upgrades
• Construction: specialised facilities require scarce materials and skilled labour
• Financing: higher interest rates raise the cost of funding multi-year projects.

Each bottleneck reinforces the others. When supply cannot expand quickly, prices do the adjusting.

This is not demand-pull inflation driven by consumers. It is cost-push inflation, driven by corporate investment at scale. Historically, that type of inflation is harder for central banks to contain without slowing growth.

When returns fall, narratives change

The AI trade assumes that today’s spending leads smoothly to tomorrow’s profits. But that equation breaks if costs rise faster than revenues.

A former senior executive at Meta Platforms recently warned that cost blowouts could lower investor returns and, in turn, reduce the flow of capital into AI. That is a crucial point.

Capital is patient only when returns justify it. If margins compress and project paybacks extend, the tone shifts. Spending becomes more selective. Expansion plans get delayed. Enthusiasm cools.

This is how themes mature. Not through collapse, but through arithmetic.

Why markets may be underpricing the risk

So far, markets have treated AI capex as a pure growth signal. More spending equals more opportunity. But they have been slower to price the macro side effects.

• Rising input costs keep inflation higher for longer
• Persistent inflation limits central banks’ ability to cut rates
• Higher rates compress valuations, especially for long-duration tech assets

In other words, AI does not need to fail to hurt markets. It simply needs to succeed at a scale that strains supply.

The strategic rethink ahead

None of this argues that AI is a mirage. It argues that the easy phase of the AI trade may be ending.

As costs rise, investors will start asking harder questions:

• Which companies can pass costs on?
• Who controls their own supply chains?
• Who earns returns on AI, and who just funds it?

The $4 trillion question is not whether AI will transform the economy. It is whether the economy can absorb that transformation without reigniting inflation and forcing a reset in policy and valuations.

For now, markets are still celebrating the promise. But the pressure is building in the plumbing underneath.

That is usually where the story turns.