Global investment in data centre infrastructure is expected to reach $31.6 trillion by 2050, driven in large part by the escalating demand for artificial intelligence (AI) capabilities. However, the sector faces significant challenges as developers race to meet tight deadlines and manage escalating costs, raising questions about whether the industry will avoid the common pitfalls of large-scale construction projects.

Danish geographer Bent Flyvbjerg has long argued that major infrastructure efforts frequently run over budget and experience delays. This "iron law" may also apply to the ambitious AI data centre boom. SB Energy, a subsidiary of Japan’s SoftBank Group, exemplifies both the promise and risks inherent in this market. The company has filed for an initial public offering in New York, positioning itself as a major player with an extensive pipeline of projects, including the massive PORTSpike facility in Ohio.

Unlike companies that supply the chips and servers powering AI, SB Energy focuses on building the data centre facilities themselves and related systems. This strategic focus shields it from the volatility seen in hardware prices but exposes it to construction-related risks, including materials shortages, workforce constraints, permitting delays, and access to reliable power. Securing interconnection—the process of gaining permission to connect to the electrical grid—can take years and varies widely by state, sometimes stretching to a decade.

SB Energy plans to co-locate many of its facilities alongside new power plants, a strategy that presents additional challenges related to plant construction and regulatory approval. The company also faces growing public opposition in the United States. According to a recent Gallup poll, 71% of Americans somewhat or strongly oppose the local construction of large AI data centres, reflecting concerns about environmental and community impacts. This social resistance contrasts with former President Donald Trump’s dismissive remarks about opposition to such projects.

Despite having no direct experience building data centres, SB Energy has acquired a consultancy with completed projects under its belt and has filed for more electrical interconnection agreements than it currently needs. The company acknowledges in its filings that "growing public resistance" to AI data centres is an ongoing risk that could affect project timelines and costs.

Financially, the stakes are high. SB Energy has secured long-term contracts with tenants including OpenAI, which has committed to partnerships extending beyond a decade. While this tenant lineup reduces some uncertainty, it also imposes strict deadlines. Delays could force SB Energy to offer discounted rates or result in customers buying out projects at lower-than-anticipated prices, impacting cash flow and profitability.

Currently, SB Energy seeks a valuation near $50 billion through its proposed IPO, implying a multiple of approximately 400 times its earnings before interest, taxes, depreciation, and amortization (EBITDA) for the first half of 2026—a reflection of the market’s optimism and a sign of the high-risk, high-reward nature of the business. OpenAI holds warrants that could boost the company’s valuation to $80 billion, indicating confidence in SB Energy’s growth prospects.

The success of these ventures is critical not only to the companies involved but also to the broader economic future, as robust AI infrastructure underpins technological progress and productivity gains. The U.S. shale boom has demonstrated the country’s capability to execute massive infrastructure projects, though it has also highlighted the risks when such ventures underperform or falter. As the race to build out AI data centres intensifies, stakeholders must navigate a complex landscape of technical, regulatory, financial, and social hurdles to avoid repeating past mistakes.