Are We Building the Next Bubble in Data Infrastructure?

The AI revolution is reshaping global infrastructure. Data centers, transmission lines, and gigawatts of new capacity are being built at unprecedented speed. Capital expenditure (capex) on AI and digital infrastructure has surged to levels that not only transform the industry — but also prop up entire economies.

But here's the concern: are we entering a bubble phase, where construction and capex run far ahead of sustainable revenue?

AI Capex and GDP: The Deutsche Bank Warning

According to a recent Deutsche Bank analysis (summarized by Fortune, Sept 2025, via Rohan Paul), the U.S. economy is being kept out of recession largely by AI CapEx. Strip out tech-related spending, and real GDP growth in 2024–2025 hovers close to 0%.

This means much of today's economic momentum comes not from AI productivity or software value creation, but from building data centers and power infrastructure.

The problem? For this to keep boosting GDP, the sector would need to accelerate quarter after quarter — a parabolic path Deutsche Bank sees as "unlikely and unsustainable."

The $800 Billion Gap

As a former Bain & Company consultant, I enjoy reading Bain's papers. Bain & Company's 2025 Technology Report ("How Can We Meet AI's Insatiable Demand for Compute Power?") estimates that by 2030, the industry will need $2TN in annual revenue to support sustainable ratios of revenue to CapEx. Yet even factoring in efficiency gains, there's an $800 BN shortfall between expected revenues and infrastructure requirements.

Put differently: the world is building data infrastructure faster than it is generating the cashflows to pay for it.

Tech Stocks and Market Dependence

As Fortune highlighted, half of the S&P 500's gains this year have been driven by a handful of AI and tech stocks. This level of concentration means any slowdown in capex or AI adoption could ripple through financial markets.

The parallel with past bubbles is clear: when growth is driven by investment and speculation rather than productivity, the correction can be sharp.

A Softer Landing?

Goldman Sachs projects that AI productivity gains could lift GDP growth by 0.4% per year in the near term, and around 1.5% in total as adoption scales.

That's meaningful — but still far below the exponential expectations that current capital expenditure seems to be pricing in.

Lessons for Europe and Spain

For Europe, and Spain in particular, the risk is twofold:

• Imported Bubble Dynamics: With billions in foreign direct investment flowing into Spanish data centers, the economy benefits in the short run. But if global AI capex proves unsustainable, projects could stall, leaving stranded assets.

• Strategic Positioning: To avoid being left as mere hosts of unfinished infrastructure, Spain and the EU must focus on tying investment to real local value: R&D, intellectual property, and resilient energy systems.

Conclusion: Bubble or Industrial Revolution?

In the long run, I believe AI will be a transformative force — reshaping how we work, live, and build industries, just as the smartphone era gave rise to platforms like Uber and Airbnb that redefined mobility and hospitality.

But in the medium term, the data center sector faces a real risk: capacity may be built faster than AI adoption expands across enterprises. If demand lags investment, we could see overcapacity, asset price deflation, bankruptcies, and consolidation.

The key question is whether revenues will scale to match the billions being poured into infrastructure. Next week, we'll explore the other side of the debate: the rise of multi-billion inference contracts that are beginning to anchor more predictable revenue streams for the industry.