Wall Street’s enthusiasm for the physical infrastructure behind the artificial intelligence boom is facing a more complicated test. Several companies tied to data centers, power generation and AI infrastructure have delayed or slowed plans to access public equity markets, exposing a growing divide between the scale of expected AI investment and investors’ willingness to fund long-duration projects at aggressive valuations. The shift does not signal that demand for computing infrastructure has disappeared, but it does suggest that public investors are examining how quickly projects can generate operating assets, contracted cash flows and visible returns. That distinction matters because the next phase of AI infrastructure requires enormous amounts of capital before many facilities can begin producing revenue. PIMCO estimates that the broader AI infrastructure buildout could require more than $5 trillion through 2030, including data centers, power, chips and related infrastructure.

AI Infrastructure Meets a Different Public Market

The latest hesitation is emerging around companies whose valuations depend heavily on infrastructure that remains under development rather than fully operational. SB Energy, backed by SoftBank, filed for a US IPO earlier this month with a proposed valuation of about $50 billion, while its development pipeline remains heavily linked to future data center capacity and AI demand. Recent reporting indicates that SB Energy has faced investor scrutiny over its proposed valuation and the timetable required to turn its development portfolio into operating infrastructure, with the Financial Times reporting on September 22 that the company had delayed its IPO amid those concerns. The situation places an unusual burden on the equity story because investors must assess not only demand for AI compute but also construction, power availability, financing, customer commitments and the timing of commercial operations.

SB Energy Exposes the Valuation Problem

SB Energy’s proposed offering illustrates why public-market scrutiny can intensify when an infrastructure company reaches the market before its physical assets reach commercial operation. The company’s Ohio development has attracted major technology-sector commitments, but the scale of those commitments also makes execution central to the investment case. NVIDIA disclosed a $1.5 billion investment in SB Energy and guarantees capped at $105 billion to provide credit support for the land, power and shell buildout associated with approximately 4.25 GW of IT load at SB Energy’s PORTS-Pike campus, with the guarantees taking effect in phases and subject to specified conditions. The project therefore sits at the intersection of semiconductor demand, AI model deployment, power generation, data center construction and long-term lease obligations. Investors evaluating the proposed valuation must consequently consider several linked dependencies rather than treating the business as a conventional operating data center company.

The capital structure adds another layer to the question facing investors. SB Energy’s business model requires substantial investment before infrastructure can produce its expected long-term returns, making access to debt and equity markets an important part of the development equation. The company has disclosed extensive future capital requirements, while reporting around the IPO has highlighted investor concerns about the distance between current operations and projected infrastructure revenues. That distance changes how public markets evaluate risk because construction delays, power constraints or changes in customer requirements can affect projects long before an operator reaches steady-state utilization. The issue is not simply whether AI demand exists, but whether investors can establish a sufficiently clear relationship between that demand and the timing of infrastructure cash flows.

Holtec Connects Nuclear Ambitions to Data Center Demand

Holtec’s decision to postpone its IPO brings the power side of the AI infrastructure equation into sharper focus. The nuclear company announced on September 17 that it had postponed its initial public offering following discussions with its banking syndicate and would keep its registration statement on file with the Securities and Exchange Commission. Holtec attributed the decision to deteriorating investor confidence across equity markets and the nuclear sector, identifying uncertainty around data center development as a primary driver alongside higher energy costs, trade tensions, military conflicts and inflation concerns. The company had been preparing an offering of 50 million Class A shares at a proposed price range of $15 to $18 per share, according to its preliminary prospectus. Its experience shows how uncertainty around future electricity demand can reach beyond data center developers and affect companies positioning themselves as suppliers of the power systems needed for AI growth.

Holtec’s own explanation is unusually direct about the connection. The company said the postponement followed “the unusual confluence of developments that has impaired investor confidence in the market for new public offerings.” It also said the nuclear sector’s retreat was driven primarily by “uncertainty over data center development,” alongside other macroeconomic pressures. Those statements matter because they place AI infrastructure expectations inside a wider investment chain rather than treating data centers as an isolated real estate or technology story. Nuclear developers, utilities, equipment suppliers and power infrastructure companies increasingly depend on expectations about where large computing loads will emerge and when those loads will actually connect to the grid. Meanwhile, public-market investors can challenge those assumptions before developers commit capital to the most expensive stages of construction.

The Public Market Is Testing Execution, Not AI Demand

The emerging caution should not be confused with a collapse in infrastructure demand. Major technology companies continue to plan substantial investments in AI infrastructure, while financing markets are developing structures designed to accommodate projects with long construction cycles and significant physical collateral. PIMCO has argued that investors can approach the AI infrastructure opportunity through secured financing and assets supported by enforceable contracts, while multiple researches highlighted the expanding role of private markets in funding the buildout. That distinction may become increasingly important as public equity investors demand greater visibility into operational assets and cash generation. A project can remain strategically essential to the AI economy while still proving difficult to package into a public-equity valuation that investors accept. The financing question is therefore moving from “How much AI infrastructure will the world need?” toward “Which projects can demonstrate enough certainty to attract capital on acceptable terms?”

That shift also changes the competitive landscape between financing channels. Public equity offers visibility and access to a broad investor base, but listed companies must continually defend valuations against changing expectations and quarterly market scrutiny. Private credit, infrastructure funds, strategic investors and project finance can instead structure capital around individual assets, contracts or counterparties, potentially making them better suited to projects with long construction horizons. Industry expects private-market financing to play an increasingly important role as hyperscalers pursue trillions of dollars in AI and data center investment through 2030. Yet, private financing does not eliminate infrastructure risk; it changes how that risk gets allocated, priced and secured.

The New Question Behind AI Infrastructure Capital

The latest IPO developments show how public-market investors are scrutinizing the relationship between AI infrastructure expectations, project execution and the timing of future returns. Investor scrutiny surrounding SB Energy’s proposed offering and Holtec’s confirmed IPO postponement show how questions around data center development can affect companies across the infrastructure stack. At the same time, the projected requirement for trillions of dollars in AI infrastructure investment means the industry cannot rely on a single financing channel.

Public equity, private credit, infrastructure funds, strategic capital and project financing can each support different stages of power and compute development, particularly where projects require substantial capital before commercial operations begin. The critical issue for the next phase of the market will be whether individual projects can convert enormous expectations for AI demand into infrastructure that reaches customers, connects to power and generates predictable economic returns. That is the test now confronting data center IPOs, and it could shape how the global AI infrastructure boom gets financed well beyond Wall Street.