Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
The writer is deputy chief investment officer at Richard Bernstein Advisors
Meta’s 40-year bond sale last year was widely seen as a vote of confidence in Big Tech and artificial intelligence given the strong demand for the debt. Investors lined up to lend for four decades to a company considered among the world’s most financially secure.
But what looks like confidence may instead be complacency. Corporate bond investors are quietly absorbing enormous speculative risk and that is unlikely to bode well for equity holders.
For years, tech’s appeal stemmed from asset-light, cash-rich business models. Pristine balance sheets and minimal debt weren’t just attractive; they were defining features. Expansion through cash flow helped fuel their rise and underpinned investor confidence. But in the scramble for AI capacity, these companies have morphed into capital-intensive utilities with huge infrastructure commitments, both on and off balance sheets.
Meta, Microsoft, Amazon and others are spending tens of billions of dollars to build or lease hyperscale data centres that may define the next phase of computing. In fact, it most likely will. But this exponential, debt-funded capex also has the hallmarks of a late-cycle surge in spending that can trigger multiple re-ratings and credit stress.
History rarely rewards lenders who finance capital-intensive growth booms at their peak. In the late 1990s, telecom companies borrowed heavily to lay fibre-optic cables, confident that data demand would ensure adequate returns. Although the infrastructure transformed the economy, it generated little return on investment for years.
In the mid-2000s, wildcatters levered up to chase $100 oil. The shale revolution reshaped US energy production, yet the companies behind the build-out still endured long periods of weak returns and balance-sheet pressure. Both episodes ended the same way: overcapacity, writedowns and years of debt overhang, despite laying the groundwork for genuine economic change.
Today’s confidence in long-dated Big Tech debt rests on two assumptions. First, that AI will deliver productivity and revenue gains sufficient to justify the scale of investment. And second, that the infrastructure being built will remain relevant for decades. Neither is guaranteed. Forty years ago, rooms were filled with mainframes whose combined power now fits in an iPhone. Who is to say tomorrow’s chips won’t become exponentially more efficient, requiring far less space and cooling, or that computation won’t migrate back to local hardware?
Are we really at the peak of technological innovation? The technology bull would insist we aren’t. A true AI believer might argue the opposite: that accelerating innovation renders today’s infrastructure obsolete well before the debt backing it matures. If so, today’s massive server farms could become empty warehouses, reminiscent of the abandoned paper and textile mills dotting the US East Coast. The difference is that these facilities are being financed with some of the longest-dated debt the corporate market has ever seen.
A more cautious scenario is that AI adoption underwhelms. What many overlook is that abundant liquidity over the past decade has created surplus capital and non-productive investment across assets, including cryptocurrencies, private equity, meme stocks and even investment-grade credit. What if AI optimism is just another expression of liquidity-driven euphoria? And what happens if interest rates rise, liquidity tightens or the AI promise disappoints?
Credit investors should take note: they do not participate in the upside if ambitious AI investments succeed, yet they bear the downside if expectations fall short. Locking in low spreads for 30 or 40 years in sectors undergoing rapid technological change is a big risk.
Large technology companies remain among the strongest borrowers in global markets. But bond market mistakes rarely come from lending to obvious junk; they come from lending too confidently to companies seen as infallible. And the key danger isn’t default, but how risk is repriced as credit excess is unwound.
When the history of this period is written, the AI-era debt boom may look more like previous cycles of over-optimism and heavy capital spending. Years of highly accommodative policy have conditioned investors to expect favourable outcomes.
A generation has never experienced a true boom-bust cycle and has been rewarded repeatedly for buying every dip. But some of the highest returns in investing ultimately come from areas where capital is scarce, not where it is abundant. And today, capital flowing into Big Tech is anything but scarce.
