HSBC's Elhedery: AI Is Transformational, Data Center Deals Are Not All Equal

The AI infrastructure boom has produced a familiar pattern in capital markets: a technology narrative so compelling that underwriting standards risk becoming an afterthought. Georges Elhedery, CEO of HSBC, is pushing back on that tendency, even as he embraces the technology's upside.

Speaking to CNBC's Dan Murphy on Tuesday, Elhedery called AI "a major enabler to re-engineer and reimagine all our businesses, all our activities." He was unambiguous about the productivity case. "It's going to put a lot of power back in the hands of our colleagues because of the level of productivity it gives them, so it's going to be transformational, and we're certainly very much on a journey to adopt it," he said.

That is the easy part. Financing the physical buildout of AI, the server farms, power infrastructure, and leasing arrangements that make the technology usable at scale, is where Elhedery wants investors to slow down.

"One has to be selective," he said. "You have to look at the quality of the tenancy agreements. You have to look at the pricing of these contracts. Not all of them are of that level of virtuosity that you can support strongly. Some of them may be more challenging."

That is a notable statement from the chief executive of a bank with close to $3 trillion in assets. S&P Global ranked HSBC as the seventh-largest bank in the world and the largest in Europe last year. The London-headquartered lender is not a passive observer of this trend either. It operates data centers directly and finances others, putting it on both sides of the AI infrastructure trade. In its 2025 annual report, HSBC said it aims to become "the leading bank for fast-growing transition ecosystems," explicitly naming data centers and AI as part of its target customer base.

Why It Matters

The scale of capital flowing into this sector is difficult to overstate. McKinsey estimates global data center spending could hit $7 trillion by 2030. Big Tech firms are increasingly turning to private credit and debt markets to fund construction, rather than relying solely on balance sheet cash, a shift that has started to reshape fixed income markets more broadly.

JPMorgan Chase projects $4.1 trillion in AI-related debt issuance by the end of the decade. That is not a marginal add-on to corporate bond supply. It is a structural shift in how capital markets allocate credit, and it is happening at a moment when government borrowing costs are already elevated.

Mohamed El-Erian, chief economic adviser at Allianz and former CEO of PIMCO, flagged the resulting tension at the GZERO Summit in London earlier this month. Hyperscalers raised roughly $200 billion in bond financing last year. This year, that figure could approach a trillion dollars, according to El-Erian. "They are yield insensitive. They are spread insensitive because they are betting on something that is truly transformational," he said.

Governments, meanwhile, are not stepping back. They are borrowing heavily and refinancing existing debt at higher rates simultaneously, El-Erian noted. Two large, rate-insensitive borrowers competing for the same pool of capital is a dynamic that tends to push yields higher across the board. Bond yields across major economies have already hit multi-decade highs in 2026, driven by concerns over government deficits, sticky inflation, and more hawkish central bank postures.

The risk here is not that AI fails to deliver value. It is that capital gets misallocated along the way, financing data center capacity tied to weak tenants or poorly structured contracts, while the broader cost of credit rises for everyone else competing for the same financing.

Key Risks

Not every data center lease is created equal, and that is precisely Elhedery's point. A facility backed by a long-term contract with a creditworthy hyperscaler carries a fundamentally different risk profile than one dependent on a smaller, less established tenant or a short-duration agreement subject to repricing. Lenders and bondholders who fail to differentiate between these structures are effectively underwriting a bet on AI demand continuing uninterrupted, rather than on the specific economics of the asset in front of them.

There is also a concentration risk building in bond markets. As hyperscalers absorb a growing share of available credit, yield-insensitive and spread-insensitive by their own admission, they can crowd out other borrowers and distort pricing signals that typically help investors assess risk. When the largest, best-capitalized players in a sector are willing to pay almost any rate to secure financing, it becomes harder for smaller or more leveraged data center operators to compete on equal terms, pushing some toward private credit structures that carry less transparency than public bond markets.

Rising sovereign yields compound the problem. Higher government borrowing costs tend to ripple into corporate credit spreads over time. A sector already absorbing trillions in new debt issuance has less room to absorb a broader repricing of risk if macro conditions tighten further.

Investment Implications

Elhedery's framing suggests a bifurcated market ahead, strong assets with quality tenants on one side, weaker structures vulnerable to a credit event on the other. Investors with exposure to data center debt, whether through direct lending, private credit vehicles, or hyperscaler bonds, should treat tenancy quality and contract pricing as the primary diligence filters, not the scale of the AI narrative attached to the deal.

The $4.1 trillion figure from JPMorgan is a useful anchor for sizing the opportunity, but it also underscores how much capacity for mispricing exists within that number. Not all of it will perform equally. Watching how spreads on hyperscaler debt behave relative to sovereign yields over the next several quarters will offer an early signal of whether the market is starting to price AI infrastructure risk with the discernment Elhedery is calling for, or whether it continues to treat the sector as a uniform bet on the technology's success.