Global Investment Strategy | Global Financial Services

For the past 18 months, capital markets have rewarded companies that leaned hardest into artificial intelligence. Nvidia’s meteoric rise, Microsoft’s OpenAI partnership, and Alphabet’s pivot to Gemini have all been celebrated as smart, aggressive moves in the AI arms race. But recent tremors in the bond market suggest a new phase is beginning, one where investors are asking: how much AI investment is too much?

And more importantly: who’s footing the bill?

From optimism to overstretch

The excitement around AI has been almost existential for markets. Tech titans are pouring hundreds of billions into data centres, chips, infrastructure, and power, anything to secure their place in the new digital economy. But in the past few weeks, debt investors have started blinking.

A basket of bonds issued by the hyperscalers Alphabet, Meta, Microsoft, and Oracle has sold off, with spreads widening to their highest levels since April. The shift reflects growing unease not about the future of AI, but about the sheer cost of getting there.

As JPMorgan recently noted, building the AI infrastructure of tomorrow will cost more than $5 trillion. That’s a number so vast it will require not just public capital markets but also private credit and, potentially, government backing. This isn’t just an R&D budget line anymore. It’s one of the most capital-intensive projects in the history of technology.

Leverage: A signal or a warning?

Ironically, many of the same firms driving this buildout - Meta, Google, and Microsoft - have enormous cash reserves and generate impressive operating cash flow. In fact, JPMorgan estimates the group will collectively produce over $725 billion in operating cash flow by 2026, and already holds more than $350 billion in liquid assets.

So why raise more debt?

That’s the question now bothering fixed-income markets. In recent weeks, tech giants have raised tens of billions through bond issuance: Alphabet with $25bn, Oracle with $18bn, and Meta with a staggering $57bn across private and public channels.

For some investors, this is an efficient capital deployment strategy. Why use cash when borrowing costs remain manageable? For others, it's a warning sign: that even the best-capitalised companies are turning to leverage to sustain a pace of AI investment that may soon outstrip demand or profitability.

Oracle: A case study in stretching

No one embodies this tension more clearly than Oracle. The company has accumulated around $96 billion in long-term debt, primarily to support cloud leasing deals tied to OpenAI and other generative AI projects. In theory, these should generate $300 billion in revenue over the next five years. But that projection hinges on a narrow client base and massive computing requirements.

In recent weeks, Oracle bonds have fallen nearly 5%, even as broader tech credit has remained relatively stable. Credit rating agencies like Moody’s have flagged the risk of overdependence on a handful of AI firms. The market is waking up to the fact that AI demand, while transformative, may not be immune to overcapacity, pricing pressure, or energy constraints.

When markets start pricing AI risk

What we’re seeing now may not be panic, but it is discipline. The fact that bond spreads are widening, not tightening, indicates that investors are pricing risk appropriately. That’s healthy.

But it’s also the first real signal that markets are becoming more selective about AI. Up until now, every dollar spent on machine learning, cloud computing, or data infrastructure has been treated as forward-looking genius. Today, capital allocators are demanding more proof.

Can these data centres generate sustainable returns? How will they manage soaring power consumption? Will competition compress margins faster than expected?

The road ahead: More questions than answers

There’s no doubt AI will reshape the global economy, but there’s growing doubt about how fast, and at what cost.

For investors, that means paying closer attention to capital intensity, cash flow coverage, and leverage metrics across the Big Tech landscape. AI may be the future, but debt is real, and repayment doesn’t wait for the next breakthrough.

Markets wanted bold AI investment. Now they want accountability. As public and private capital lines blur, the next test won’t be about vision; it will be about discipline.

The AI revolution may be here, but the bill is coming due.