Bain & Co. projects that the worldwide AI sector would need to reach roughly $6 trillion in yearly revenue by 2031 to justify the volume of investment flowing into data centre infrastructure, according to the firm's annual technology report published Tuesday.

Current consumer and enterprise AI offerings may account for roughly $1.8 trillion of that projected total, leaving a gap of approximately $4.2 trillion in revenue the sector has yet to produce, according to Bain. Industry analysts noted that bridging that gap would depend on expansion across sectors still in early development, including autonomous systems, robotics, pharmaceutical research, mental health technology, and power generation.

David Crawford, lead author of the report and chair of Bain's Global Technology, Media, and Telecommunications practice, argued that the industry requires a wave of innovation surpassing what mobile and cloud computing unlocked, and that AI infrastructure is being built well ahead of actual demand. Crawford further argued that sustainable funding would require adding roughly one percentage point to the annual global GDP growth rate.

Bain's forecasts put cumulative data centre investment at between $5 trillion and $6.5 trillion through 2030, a level of expenditure that would bring roughly 150 gigawatts of additional compute capacity online and come close to tripling the worldwide total within five years. Annual spending on AI infrastructure — encompassing data centres, processing power, and hardware at the chip level — may reach $1.5 trillion by 2031, according to the company's projections.

According to the report, construction is advancing at a rate that has few parallels in recent history. Five years ago, a 50-megawatt facility would have been viewed as a large-scale installation. According to Bain, large-scale technology infrastructure operators are now developing facilities with capacities exceeding 5 gigawatts, with projected expenditures ranging from $150 billion to $200 billion per site.

According to the firm, domestic investment in data centre infrastructure, measured as a proportion of GDP, has reached levels comparable to the spending once devoted to building out the country's rail, telephone, and power networks. Media reports indicate that property developers are facing constraints across electricity supply, machinery procurement, and workforce availability. One media outlet reported that community resistance stalled or halted an estimated $68 billion in infrastructure and development ventures across the country during the second quarter.

Questions about revenue generation from AI form a central part of wider discussions about the technology's financial viability. Bain notes that although much of the conversation around AI has centred on workplace productivity gains, the capital outlay involved in developing AI infrastructure points to a revenue requirement running into the trillions — a threshold that operational efficiencies alone are unlikely to meet, the firm states.

PricewaterhouseCoopers projects that worldwide data centre investment will total $31.6 trillion between now and 2050, with artificial intelligence driving yearly capital outlays from approximately $800 billion at present to $1.8 trillion by mid-century. Questions about whether AI computing demand can sustain unconventional infrastructure arrangements — among them data centres placed in orbit — have intensified alongside growing uncertainty over long-term revenue prospects and efficiency improvements that are steadily lowering the expense of operating AI systems.