Data centre infrastructure investment - AI bubble, boom or bust?

Data centre financing is becoming increasingly complex.

Thanks to AI it is also a topic that also has become regularly covered in mainstream business media (and even some dinner parties). 

The spark for this was talk of an AI bubble which began in the early part of Q425. 

It was still being discussed at the World Economic Forum in DAVOS in mid-January 2026.

A fireside chat between Blackrock CEO Larry Fink and Jensen Huang CEO of NVIDIA cited the trillions of dollars needed for the 'largest infrastructure buildout in human history in computing and energy infrastructure.’ 

As reported in an NVIDIA blog: “Fink summarized the discussion by saying that what he heard suggested the world is far from an AI bubble. Instead, he posed a different question: Are we investing enough?”

Boosters v Doomsters

The late 2025 doomster v booster stand offs started with ripples of disagreement across finance providers about whether an AI bubble exists, if it were about to burst and what the fallout for data centres might be.

At the time this spiked the level of mainstream media interest and social media activity.

On 5th Nov Blackstone president Jon Grey said in a Linkedin video: “AI is the main thing. And to make this AI a reality, we've got to build a lot of infrastructure. We've got to build a bunch of data centers and a ton of power. And that requires enormous amounts of capital. Then that will lead to a huge step function increase in productivity.”

The November 15th edition of the Economist magazine cover story led with “How markets could topple the global economy – If the AI bubble bursts, an unusual recession could follow”. 

In the online and broadcast world the BBC’s economics editor Faisal Islam conducted a face-to-face interview with Google CEO Sundar Pichai. This was first published on Nov 18th with a headline: “No company is going to be immune if AI bubble bursts.” 

Pichai went on to say that while there would be overshoots in terms of valuations the fundamentals of investment and demand are strong. He said even a company operating at the scale of google is having to work hard to keep up the research, product development and consumer and business demand for AI services. Google invested $90bn in data centre infrastructure in 2025. The number for 2026 is expected to exceed this. 

First the bad views…

However, public market volatility remained throughout the remainder of 2025.

For those in private finance seeking and providing long term investment, the ebbs and flows of what was happening in public financial markets became increasingly important. 

Up to this point in November, AI, financial, investment and economic headlines reflected a war of words between equity market Doomsters and Boosters with some predicting a major stock market correction while others remained confident in the fundamentals of data centre infrastructure investment.  

AI doomsters expressed numerous concerns. 

These included public stock markets under-pricing risk; So called ‘prisoner’s dilemma’ (consumer’s investing heavily in big AI for fear of losing out ‘even if near term returns are unclear’); Digital infrastructure equity and loan investment deal structure circularity between big tech and AI players; Growing debt levels in the market being bundled into increasingly complex collateralized instruments. 

These views pointed to an AI stock market valuation bubble that is said to be shoring up the US economy, resulting in substantial concentration of risk – where if/when there is a rush for the door would result in an existential risk to the world economy.

For example, one report said  “capital spending contributed nearly half of U.S. GDP growth in the first half of 2025”. According to a JPMorgan team it can be estimated that U.S. households have gained $5.2 trillion in wealth over the last year from the appreciation of a basket of just 30 stocks tied to AI.’ 

Now the good views…Strength in Fundamentals

If there are two sides to every story the positive view is that data centre infrastructure investment is just starting and will remain fundamentally strong.

NVIDIA is still the world’s most valuable company, Alphabet has doubled in value to over £3trn, OpenAI is (probably) planning a $1trn IPO which could mean $1trn in infrastructure spending?

The current investment commitments into AI infrastructure are already north of a trillion dollars. Behind much of this are the tech giants who are deploying capital at an extraordinary rate. The big four, Google, Microsoft, Meta and Amazon are expected to spend $450bn in 2026 alone.

Private capital is being invested as unprecedented levels. As referenced above Blackstone President Jon Gray offered to ‘gently pop the idea of ‘the AI bubble, private credit bubble and the stock market bubble,’ in a Linkedin video post.

He said: “On Private credit blow-ups: the facts on the ground are so compelling because what is private credit at the end of the day? It's basically taking an investor and bringing them right up to a borrower…what we continue to see is lending to high quality large scale businesses. You can earn a meaningful premium. And so, to me that's going to continue to make private credit quite attractive.”

Meet in the middle… Changing Risk Profiles

There can be little doubt that talk of an AI bubble has affected sentiment towards big tech in terms of cost of capital for its infrastructure spending plans.

For example, there is some evidence that data centre investor clubs are seeking a lower risk profile on ABS (Asset Backed Securitization) deals by demanding higher yields through wider spreads.

Though not the direct result of fear of an AI bubble, data centre companies are issuing more ABS and CMBS (Commercial Mortgage Backed Security) paper than in previous cycles. 2025 saw some large colos using ABS and CMBS for the first time to refinance.

For example, a report titled: “Financing a Digital Future, Opportunity, Risk and Reward in the US” from market information and advisory firm Global Capital said: “In 2025, a record year for data center ABS issuance in the US, $11.5bn of paper from the sector has hit the market across 23 series of notes. Another $7.14bn of data center paper was issued in the CMBS market. The issuance tally could double in 2026 as Bank of America predicts the data center ABS and CMBS markets could collectively top $35bn across new issues and refinancings.”

Fuelling issuance is huge capital expenditure from hyperscalers. Amazon, Apple, Google, Meta, Microsoft and Oracle, collectively, are due to spend $382bn on capex this year, $432bn next year and $539bn in 2029, according to Citi, it said.

As deal sizes growth the fundamentals for lenders to colocation providers remain favourable towards developers able to show long-term tenant leases (seven years and more) with secured recurring colocation revenue from investment grade user contracts.

But big tech is not the only recurring revenue and net operating income story. It appears investors are taking note where colocation sole reliance for revenue is on big tech customers. To appear more attractive to lenders some issuers are choosing to emphasise their diversity of customer that includes Hyperscalers with a mixed enterprise customer base.

More broadly investors and lenders are taking a more detailed view of the data centre sector.

There are reports of widening spreads for issuers resulting in higher debt costs. Yields on bonds issued by big tech are also widening with investors demanding higher premiums due to changes in perceived risk.

Debt levels rising

There also appears to be concern about increasing complexity in the debt financing market for data centre developments.

Hyperscale borrowing amounts and structure announcements in September and October 2025 are highlighted in a Reuters report from Nov 5th which said, ‘Investment Grade Borrowing by big AI tech firms spiked in September and October 25 after years of steady annual issuance,’. 

Who data centres serve…the growing AI ecosystem 

It is clear that talk of an AI bubble is not going to subside and the adoption curve will not be smooth. 

Like previous disruptive technology inflection points there will be friction in the market amid successful and less successful projects, take up and ROI.

Julius Baer, Swiss wealth management firm said about AI that: “AI remains the fulcrum of market narratives and corporate investment, as AI hyperscalers have led US equity markets to new highs and have committed gigantic sums to AI infrastructure, with power and grid capacity emerging as binding constraints. While AI’s disruptive potential may surpass that of the internet, its payoff is anything but straightforward. Generative AI tools are everywhere, yet widespread adoption does not guarantee business transformation or bottom-line impact.”

Others point out that as with earlier tech shifts much GenAI adoption is happening relatively quietly within sectors such as fintech (high speed traders being a prime example) and in large financial services and investment houses – i.e. companies that are technology driven traditional early adopters. Similarly in consumer banking and many other b2c sectors Agentic AI is clearly on the rise.

Then there is the tech investment sector itself which is moving fast on AI.

In terms of different types of production and inference AI use cases the start-up and growth capital investment landscape for AI firms provides perspective on confidence in the sector. In the US Total U.S. startup funding reached about $162.8 billion in H1 2025, with AI accounting for a large share of that growth. Arion Research LLC reported that “Artificial intelligence startups remained the primary beneficiaries of venture investment, with several blockbuster deals defining the quarter. Meta's $14.3 billion June investment in Scale AI led the charge, following SoftBank's record-breaking $40 billion March financing for OpenAI. Other major AI-focused deals included: $10 billion financing for xAI; $3.5 billion financing for Anthropic; $2 billion financing for Thinking Machines Lab.”

In 2024, U.S. AI startups were key to overall VC growth, with AI-oriented companies capturing a record ~46.4% of all U.S. VC dollars, part of a total of $209 billion in VC investment in the U.S. for the year. Reuters

According to UKtechnews AI has generated £2.4bn of VC capital going to UK AI firms which represents a record 30% share of total VC investment in H125.

The Data Centre Investment Community View

The view within the data centre investment community is that strength and depth of demand will remain for the right types of data centre in the right locations.

Richard Ludwig, director of investment banking at Alantra issued a note which included: “EBITDA multiples of 30x+ are achievable - but only when backed by tangible growth plans & solid fundamentals: the right location, customers & power strategy. The strongest platforms combine organic expansion (available, built capacity), M&A (national & international), & greenfield development, creating scalability & investor confidence.”

Zahl Limbuwala, is a partner at DCTP (Digital Transformation Capital Partners).

He said: “The exam question is how much speculative build is occurring? While speculative deployment of capex remains low then risk of oversupply remains low and you just end up with a glut of announced but unrealised projects. In respect to a “bubble” this normally refers to over inflated valuations and the “bust” part is when a valuation correction happens. While supply/demand of data center capacity is a factor in a data center company’s valuation the so called bubble most are talking about these days is the valuations we are seeing on AI related companies (both software and hardware) based on the current value delivered by AI. Of course, if there is a valuation bubble and that bubble bursts then everyone including the data center operator will feel some impact because these AI companies and those that lease GPUs to them are becoming a larger proportion of the data center customer base. However, unless you are overexposed to this client type the rest of the customers and demand is still based on almost 30yrs of stable growth of the digitalisation of everything.”

Conclusion

Data centre infrastructure financing being inextricably and directly linked to AI driven global economic outlook is a new landscape for issuers and lenders.

The scale of AI infrastructure investments required mean any perceived slowdown in AI use among consumers or pauses in corporate enterprise AI projects and investments – reports of which have been mixed in Q425 - are a barometer against which the risk profile of digital infrastructure lending and investing are now measured.

On the infrastructure side both private and public market investors in data centre and AI digital infrastructure are keen to point out that, as with any fundamental digital technology shift (such as mobile, the internet, cloud), there will be those whose gains are game changing amid some for whom the losses in the disruption could be existential.

AI could be the greatest example yet.