AI Infrastructure Expansion May Need More Equity as Debt Concentration Risks Grow

Table of Contents

• You might want to know


• Main Topic


• Key Insights Table


• Afterwards...

You might want to know

• Why could the rapid expansion of artificial intelligence infrastructure make equity financing more important than debt?


• How might the scale of projected AI spending affect lenders, investors, and the companies building the infrastructure?

Main Topic

The construction of infrastructure to support artificial intelligence (AI) is demanding extraordinary amounts of capital. Data centers, computing equipment, power connections, and related facilities all require substantial investment. As financing needs grow, Blackstone President Jon Gray has argued that the sector will need more equity funding, particularly as lenders become increasingly exposed to the companies driving the buildout.

Gray made the comments Tuesday at the Hong Kong Monetary Authority’s Global Financial Leaders’ Investment Summit. He said the scale of capital required for AI is creating concentration risks for debt investors. These risks can arise across the sector as a whole, as well as when a lender has substantial exposure to an individual company. If many borrowers depend on similar business conditions or a small group of major customers, the effects of a downturn could be concentrated rather than widely distributed.

The difference between equity and debt helps explain Gray’s concern. Shareholders may benefit if a company grows in value, because their potential returns are linked to its success. Lenders, by contrast, generally receive interest and repayment of principal. Their potential upside is therefore more limited, even when a borrower’s business expands rapidly. When financing needs are immense, relying heavily on debt can leave lenders carrying significant risk without sharing proportionately in the gains.

Emmanuel Roman, CEO of Pimco, raised a related concern during the same panel. He cautioned that the volume of financing associated with AI might eventually be too much for markets to absorb. His warning was not that investment in the technology would necessarily stop, but that the amount of capital required could test the capacity of investors and financing markets.

Roman cited an estimate from Bill Ford, CEO of General Atlantic, that AI-related capital spending could reach around $10 trillion over the next five years. He also said about $550 billion in AI debt had been issued in various forms so far this year. These figures point to both the potential scale of future investment and the financing activity already taking place. They also raise questions about how debt issuance may develop if companies continue to expand their infrastructure plans.

Even if companies use a mix of equity and debt, Roman said the total funding requirement would be “gigantic.” He compared it with a global corporate bond market of roughly $20 trillion, emphasizing the scale of the potential demand relative to an established source of financing. “At some point in time, it’s going to be too much,” he said. The comparison does not mean that every AI project will be financed through corporate bonds; rather, it illustrates why the overall funding challenge is drawing attention from investors and financial institutions.

Separate estimates underscore the size of the infrastructure task. Goldman Sachs wrote in a report Monday that global data center demand will require about $8.5 trillion in capital spending over the next five years. Data centers are central to the expansion of AI because they house computing equipment and provide the capacity needed to develop and operate digital services. Meeting projected demand involves more than purchasing technology: developers also need suitable locations, approvals, reliable electricity, and access to financing.

Gray described a broad financing mix as necessary. Public markets, banks, and private capital are all likely to contribute, while companies may increasingly use structured transactions to manage how much debt they take on. Structured arrangements can help allocate funding and risk in different ways, though they do not eliminate the underlying need to finance construction. Cash generated by hyperscalers—large companies that operate extensive cloud and computing services—could also support part of the spending, he said.

That range of funding sources reflects the scale of the undertaking. Equity can provide companies with capital without creating the same fixed repayment obligations as borrowing, but issuing equity can affect existing shareholders’ ownership. Debt may allow a business to raise funds while retaining ownership structure, but it creates obligations to repay principal and interest. In practice, companies and investors must consider the costs, risks, and availability of each approach, rather than assuming that one source can meet all requirements.

“The creativity is needed because the scale of capital is enormous,” Gray said. His comment captures a practical challenge for the sector: financing structures may need to adapt as projects grow, timelines vary, and risks differ among businesses. A data center project, for example, involves long-term assets and depends on the availability of power and other essential inputs. The ability to raise funds is therefore connected to whether projects can secure the necessary sites, equipment, and operating resources.

Despite the financing concerns, Gray remained positive about the outlook for AI infrastructure. He said demand for computing power was “running well ahead of supply.” That imbalance suggests that companies and investors may continue to see strong reasons to build, even as the capital required becomes more difficult to organize. Strong demand, however, does not remove execution risks. Projects still have to be built and supplied before capacity can be put to use.

Evidence of business interest is also appearing in spending on AI tools. Gray said spending on AI models across Blackstone’s portfolio companies had risen 20-fold over the past year. Businesses were seeing financial benefits from applications such as coding, customer engagement, and content creation. These examples show why companies may be willing to invest in AI: tools can support different functions across an organization, rather than serving only a narrow technical purpose.

At the same time, greater use of AI creates additional demand for the infrastructure that enables it. Companies need computing capacity to develop models and run applications, while service providers need sufficient data center resources to meet customer requirements. This connection helps explain why investment plans can grow quickly: adoption across businesses may translate into increased demand for processing, storage, and power. Yet the benefits of applications and the availability of infrastructure do not necessarily advance at the same speed.

Gray said that meeting demand remains difficult because securing sites, permits, chips, power, and capital is becoming more challenging. Each constraint can slow a project or affect its cost and timing. A site may be difficult to obtain; required permits may take time; chips and electricity must be available; and financing must be assembled on suitable terms. These factors are interconnected, so resolving one bottleneck does not guarantee that a project can proceed smoothly if another remains.

The comments from Gray and Roman therefore point to two aspects of the same market. On one side, demand for AI computing is rising, and businesses are reporting practical uses for AI models. On the other, financing and construction require resources on a scale that may expose lenders to concentrated risks and test the capacity of capital markets. The central issue is not simply whether investors are interested, but how the required investment can be shared among different funding sources and matched to the risks of individual projects.

Equity financing may become more important as companies and financiers seek to avoid excessive dependence on debt. A balanced approach could involve public equity, borrowing, private investment, structured transactions, and operating cash flows. The appropriate combination will depend on the company, the project, and market conditions. No single financing method removes uncertainty, but spreading capital needs across multiple sources may help address the limits and risks associated with relying too heavily on one channel.

Key Insights Table

Afterwards...

The next stage of AI infrastructure development will depend not only on how much capital investors can provide, but also on how effectively that funding can be connected to deliverable projects. Investors, companies, and policymakers will need to understand the relationship between projected demand, actual use, construction timelines, and the availability of essential inputs. Better information about these factors could help markets distinguish between plans that are feasible and those that face substantial execution challenges.

Further work on financing structures will also be important. Research into how equity, bank lending, private capital, and structured transactions distribute risk could help clarify which approaches are appropriate for different types of projects. Greater transparency around debt exposure and funding commitments may also help investors assess concentration risks, while careful analysis can show whether individual projects and companies can meet their repayment obligations.

Infrastructure planning should give equal attention to power, permitting, land, and technology supply. Advances in energy systems, efficient computing, and data center design could help ease some constraints, although they cannot substitute for realistic project planning. A durable AI buildout will require financial innovation alongside dependable physical infrastructure and clear assessments of risk. As AI adoption expands, continued study of its economic benefits, capital requirements, and resource demands can help ensure that investment decisions remain grounded in both opportunity and practical capacity.