Dealmaking in AI infrastructure

Investment in digital infrastructure has grown year-on-year since the turn of the century, but the recent AI revolution has catalysed exponential growth, driving M&A activity to an all-time high. We examine whether this is a bubble or a sustained boom. 

What do we actually mean by a “bubble”?  

The best way to think of a “bubble” is an overheated sector where asset valuations are increasingly divorced from analytical means of assessing fundamental value – in other words, valuations are driven by “hype” rather than intrinsic value. The surge of investment in AI infrastructure shares some all too familiar traits with previous bubbles: soaring valuations for not-yet profitable assets, aggressive capital inflows and highly leveraged financing structures. The crucial difference though is demand: while a lack of consumer demand ultimately exposed fundamental flaws in business models during the 1990s dot-com bubble for example, demand for AI infrastructure continues to surge. Early and development-stage infrastructure investors are also used to a higher risk and return profile and investment horizons are often longer. 

Will power cause a pop? 

While demand seems unlikely to slow, the risk of a market correction remains and infrastructure bottlenecks, particularly related to resource capacity, are a likely potential trigger for that correction. Power grids are already constrained, and we are increasingly seeing infrastructure investors relying on innovative “off grid” or self-sufficient energy solutions to power their AI infrastructure. While a grid connection remains the optimal solution, these come with significant lead-times and increasing regulatory hurdles. Nevertheless, these self-sufficient “island” solutions may allow supply to keep up with demand. Cross-sector partnerships with power providers can allow the parties to share capital costs and operational risks while leveraging synergies: investors bring demand certainty while energy providers deliver generation capacity, but making the economics work can be challenging given the demand energy providers currently have. Long-term offtake agreements, ensuring predictable revenue streams can help manage this.  

Is it all AI, or is cloud still carrying the load?  

While AI dominates the narrative, the digital infrastructure boom has also been powered by strong demand for cloud computing. Enterprise migration, the rise of Software-as-a-Service, and the growth of edge computing all drive data centre investment. AI and cloud workloads are increasingly interdependent, with hyperscale data centres designed for both and investment strategies reflecting this convergence. This dual demand base adds resilience, supporting sustained deal activity even as market conditions shift. For dealmakers, understanding the interplay between AI and cloud is key to assessing long-term value. Ultimately, it is the combined momentum of AI and cloud that underpins the current wave of dealmaking, suggesting the boom is grounded in broad structural shifts rather than short-term hype.  

Creative deal and platform structuring  

To manage risk and attract diverse capital, maintaining the necessary investment in the sector, creative structuring such as DevCo/YieldCo structures are being adopted. Construction-stage projects carry significant uncertainty, including risks of cost overruns and delays, which certain investors traditionally avoid. These structures separate development assets from operational ones, enabling investors such as infrastructure funds to invest in the higher-risk DevCo while other investors such as pension and real estate funds focus on the YieldCo, which holds stabilised assets with contracted revenues. Completed projects can be sold from DevCo to YieldCo, creating a self-sustaining pipeline of developments without excessive leverage. This structure supports efficient capital allocation: DevCos typically use mezzanine debt or construction loans with higher margins, while YieldCos secure long-term, lower-cost financing backed by predictable cash flows. More sceptical commentators point to circular financing structures as an early warning sign of an investment bubble, but when implemented well these deal structures actually channel investment into growth without forcing risk-averse investors into speculative positions, ultimately helping to temper volatility. 

Alternatively, HoldCo structures can also be popular, consolidating development and stabilised assets under a single corporate entity. In some ways the opposite of a YieldCo/DevCo, this allows equity investors looking for portfolio diversification with some a mixed risk profile to invest, while enabling developers to raise corporate-style debt facilities backed by the operating portfolio, avoiding the complexity of asset-level financing. By offering exposure to both growth and stability within one vehicle, HoldCos attract capital from diverse sources and strengthen resilience against market fluctuations. In an environment where AI-driven demand is fuelling rapid expansion, these structures provide some stability, helping developers secure long-term funding and reducing systemic risk in the sector. 

Rethinking exit strategies  

As private capital pours into the AI infrastructure sector, exit horizons are becoming a key consideration in M&A deals. Platforms are becoming significantly larger, major investors are increasingly developing their own platforms and public markets remain challenging. This creates significant difficulties in exit which in turn tempers appetite for initial investment. Again, creativity is being deployed to find solutions. For example, we are increasingly seeing investors use securitisation structures to realise investments without disposing of the underlying assets. While features like amortisation and long-term leases with strong guarantees can bolster confidence, whole-campus securitisations may be too large for the market, requiring smaller, phased securitisation or indeed site by site disposals. Rather than creating a barrier to growth, these evolving exit strategies, including securitisation and phased transactions, are allowing investors to recycle capital and reinvest in new opportunities. By anchoring valuations to contracted revenues and supporting steady asset turnover, these mechanisms are helping to sustain deal flow and underpin the continued boom in AI infrastructure investment. 

So is it a bubble or a boom? 

While the current frenzy of investment and dealmaking in the sector is revealing some symptoms of an investment bubble, the pace of investment is, crucially, matched by an exponential increase in demand. That demand is, in turn, supported by structural shifts in technology and enterprise adoption, pointing towards long-term utility rather than short-lived hype. Unprecedented levels of dealmaking bring novel challenges, and we expect to continue to see bespoke structures and solutions to sustain investment over the long term – but our view is that investment will sustain. The AI boom (or bubble) is now too big and too intrinsic to the future to pop.