Investor interest in initial public offerings (IPOs) of data centre companies appears to be waning, as several high-profile listings face delays or cancellations. In the past month, three major data centre operators—AirTrunk, Firmus, and SB Energy—have either postponed or withdrawn plans for public listings in Singapore, Sydney, and New York, respectively.
AirTrunk, a Blackstone-backed data centre operator, postponed its $1.5 billion IPO scheduled in Singapore. Shortly before, Firmus, supported by Nvidia, withdrew a planned $5 billion float in Sydney, and SB Energy, backed by SoftBank, delayed its New York IPO. Collectively, these firms sought around $12 billion through their offerings, a small fraction of the estimated $5.2 trillion investment McKinsey forecasts will be required for the data centre sector by 2030.
Despite the sizable capital need driven partly by demand for artificial intelligence (AI) infrastructure, investors appear cautious about public market offerings. Several factors contribute to this hesitancy. For instance, AirTrunk’s structure as a real estate investment trust (REIT) makes its stock sensitive to rising interest rates, which have negatively impacted similar securities. Firmus, having drastically increased its valuation since a private funding round three months prior, and SB Energy both report low revenue and continue to operate at a loss, complicating traditional IPO valuation methods.
Nevertheless, companies remain interested in equity financing as it provides flexible capital that can be allocated across projects without the restrictions often associated with debt financing. A public listing also creates a platform for potential future capital raises. However, there is growing recognition that data centres’ hybrid nature—combining real estate, infrastructure, and fast-evolving technology—may be better financed through alternative methods.
Project finance, where debt is secured against the cash flow of specific assets, allows investors to choose projects individually, reducing exposure to broader company risks. Firms that are early-stage and revenue-light often find more alignment with venture capital funding. Additionally, lenders may be less vulnerable to rapid technological obsolescence compared to equity investors, who bear initial losses if server equipment and chips become outdated sooner than expected. Rating agency S&P assumes data centre assets have a lifespan of up to 25 years, which is about half the duration considered for other asset types.
The projected scale of investment needed to support AI computing infrastructure remains uncertain. While McKinsey’s $5.2 trillion estimate encompasses broader data-related developments beyond AI, Moody’s projects a $3 trillion requirement through 2031 specifically for AI-related capacity. Regardless of the precise figure, industry observers agree that the capital demands are significant, but public stock markets may not be the most suitable channel to meet these funding needs.
