Supporting the growth of artificial intelligence (AI) will require enormous investment in physical infrastructure, according to Citi, which projects global AI-related capex to grow to US$3.8 trillion by 2030, up from US$1 trillion this year.
A new Citi Research report, from a team led by global head of technology & communications research Heath Terry, outlines how large financing partnerships are increasingly emerging to help fund these projects, reflecting confidence in the long-term demand outlook.
“The scale of planned construction is equally notable. New AI-related power capacity is expected to reach 19 gigawatts in 2026, 29 gigawatts in 2027, and 45 gigawatts in 2028,” Terry said.
Citi notes that even as challenges emerge, including limited power availability, shortages of skilled workers and regulatory hurdles, companies continue to invest aggressively because demand remains strong.
“Importantly, these investments are producing attractive economic returns. Hyperscalers are generating cash returns of roughly 30 per cent on invested capital from their infrastructure spending. That profitability helps justify continued expansion and reinforces the connection between AI demand and infrastructure investment.”
But the next phase of AI adoption is increasingly focused on measurable business results, according to the financial giant, with investors seeking evidence of real-world adoption, monetisation and returns on investment.
These and other themes will be explored at Citi’s flagship Global TMT Conference, on 8-10 September in New York City. The conference is Citi’s largest globally, attended by over 2000 institutional investors, venture capitalists, sovereign wealth investors, and corporate management teams.
“During the conference, we will be looking for further evidence that enterprise adoption is accelerating. Areas of focus include actual AI spending levels, customer usage rates, measurable business benefits, and signs of how companies are allocating IT budgets for 2027,” Terry said.
“We will also monitor trends such as growing use of AI agents, adoption of open models, and decisions about whether AI workloads run in public cloud environments, private facilities, or hybrid configurations.”
The overall picture, according to Citi, continues to suggest that enterprise AI adoption remains in its early stages, with many large organisations currently allocating only a small portion of technology budgets to AI initiatives, while a smaller group of leading adopters has already committed substantially larger shares.
“That gap implies significant room for future spending growth if AI projects continue delivering measurable returns.”
Across the technology supply chain, the report highlights that this demand is likely to benefit chip makers, networking providers, data-centre operators, hardware suppliers, enterprise-software companies and communications-infrastructure providers.
Terry said it also helps explain why investors remain focused on the durability of AI spending and the pace at which businesses move from experimentation to broad deployment.
A major focus across software companies is whether AI is moving beyond pilot projects and becoming embedded in everyday business operations. According to Citi, investors increasingly want proof that AI is driving customer adoption, revenue growth, workflow expansion and measurable business outcomes.
“The key debate is no longer whether companies will spend on AI, but which vendors will capture that spending and establish durable competitive advantages. Conference discussions are likely to centre on evidence of production deployments, successful monetisation, and whether agentic AI strengthens established software platforms or lowers barriers to entry for new competitors.”
The most important question for the AI ecosystem remains straightforward, according to Terry.
“How quickly will businesses adopt AI at scale? The answer will shape demand for computing infrastructure, data centers, networking equipment, software, and many of the companies participating across the broader technology landscape,” the September report said.
Terry added that one of the clearest signs of accelerating adoption is the rapid growth in orders placed with hyperscalers.
“Combined backlog among these companies reached approximately US$1.7 trillion at the end of the second quarter, up 151 per cent from a year earlier. That suggests customers are committing to significant future spending as they prepare to deploy AI more broadly.”
At the same time, the AI market is evolving, according to Citi. While access to the most advanced chips remains constrained and some next-generation AI models have been slower to reach the market, the growing use of open-weight models has expanded the range of options available to enterprises.
“Rather than viewing open and proprietary models as direct competitors in a winner-take-all market, we see room for multiple approaches to succeed.”
Improving economics and software are increasing the amount of AI work that can be processed in a given period and making the cost of completing tasks cheaper. Citi said businesses are likely to increase overall demand for AI.
“This dynamic could support a more diverse AI ecosystem that includes leading frontier-model developers, providers of smaller specialised models, and organisations that customise models for specific business needs. Lower costs can expand the market by making AI practical for a wider set of use cases across industries.”
Regulation remains an important factor across the technology sector, the report notes, while mergers and acquisitions are expected to remain a key consideration.
“Conference participants will be watching for developments that could affect data-centre construction, software adoption, communications services, satellite networks, and overall technology investment. Changes in policy could influence both demand trends and competitive dynamics across multiple industries,” Terry said.
“Investors will be looking for updates on announced transactions, potential consolidation opportunities, and signs that strategic deals could help companies strengthen competitive positions or unlock additional value.”