Wall Street Turns Data Center Power Into a Traded Bond Market

Wall Street has found a new asset to package and sell: the electricity that powers AI data centers. As artificial intelligence workloads drive up demand for computing, the cost and availability of power have become the core value proposition for data center operators-and now, for bond investors. The financial industry is transforming megawatts into tradable securities, with $61 billion in outstanding data center-backed bonds projected by mid-2026.

Megawatts as Collateral

Every AI query triggers a cascade of server activity, cooling, and network traffic inside a data center, all of which consume electricity. For operators, the electric bill is now a defining expense, and the ability to secure enough power often determines how much revenue a facility can generate. Wall Street is capitalizing on this by structuring bonds where the underlying collateral is not just real estate, but the entire business of delivering reliable electricity to high-value tenants.

Once a data center is operational and has signed customers, its owner can move the facility and its contracts into a special-purpose entity that issues debt. Investors are repaid from customer payments-mainly rent and service fees-after covering electricity, maintenance, taxes, and insurance. The bond's risk profile is shaped as much by power prices and deliverable megawatts as by the creditworthiness of the tenants, which are often large cloud or AI companies leasing capacity in blocks measured by the megawatt.

Structuring the AI Bond

Data center securitizations have grown rapidly, expanding from about $4 billion in 2020 to $61 billion expected by July 2026, according to research from the Structured Finance Association using Barclays data. In February, S&P assigned an A(sf) rating to Sabey Data Center Issuer's $475 million 2026-1 notes, which are backed by both real estate and tenant lease payments. The structure typically starts with debt at no more than 70% of appraised asset value, with the remainder as sponsor equity. Bonds often have an expected repayment point around five years, but legal maturities can stretch to 25 or 30 years, exposing investors to refinancing risk if market conditions shift or facilities require costly upgrades to support denser, hotter AI chips.

Wall Street divides these bonds into tranches with different risk and return profiles. Senior tranches get paid first and carry lower yields, while junior tranches absorb losses sooner but offer higher interest. The average data center asset-backed security (ABS) issuance is about $600 million, while commercial mortgage-backed securities (CMBS) deals average $1.2 billion. Despite the growth, these figures are small compared to Morgan Stanley's estimate of $2.9 trillion in global data center spending through 2028, with only about $150 billion expected to come from securitizations and mortgage bonds.

Regulatory and Market Shifts

The U.S. Securities and Exchange Commission (SEC) recently clarified that many data center securitizations do not fall under the Exchange Act's definition of asset-backed securities, provided they match certain structural criteria. This distinction allows issuers to avoid some ABS-specific rules, such as mandatory risk retention and certain disclosure requirements, potentially lowering issuance costs and encouraging more facilities to enter the bond market. Still, federal antifraud laws and offering regulations remain in force, and investors must scrutinize deal documents for details on power contracts, tenant concentration, and asset condition.

For bondholders, the risks are tightly linked to the physical realities of data center operations. Delays in grid connections can postpone lease revenue, while concentrated tenant bases mean that a single large customer's departure could destabilize cash flows. Rising electricity costs or the need for expensive retrofits to accommodate new AI hardware can further erode margins. If refinancing becomes difficult near the five-year mark, issuers may be forced to seek new lenders just as older facilities lose their competitive edge.

AI Demand and Power Constraints

Lawrence Berkeley National Laboratory projects that U.S. data centers could consume 649 terawatt-hours of electricity in 2030, or nearly 12% of total U.S. usage, with a possible range from 9.5% to 15.3% depending on technology adoption and efficiency. This surge is driven by the proliferation of AI chips, which increase both revenue potential and infrastructure demands. Even with long-term customer contracts, data centers may need to retrofit power and cooling systems to keep up with evolving hardware requirements.

Some operators are experimenting with flexible computing strategies-similar to those used by Bitcoin miners-to reduce power draw during grid stress, but the bond market ultimately prices in the risk that a facility's physical limitations could undermine its long-term value. The entire securitization model depends on the ability to deliver reliable electricity to tenants who can pay, and on the assumption that those tenants will continue to need the same type of infrastructure as AI workloads evolve.

According to the Structured Finance Association, data center ABS is expected to account for 12% of the esoteric ABS market by 2026, up from 3% in 2020, while data center CMBS will represent about 6% of single-asset, single-borrower CMBS. Barclays projects that outstanding data center securitizations could reach $180 billion by the end of 2028. For context, the SEC's recent move to allow regulated transfer agents to use public blockchains for shareholder records, as reported earlier, signals a broader willingness to adapt legacy finance to new forms of digital and physical infrastructure.

Wall Street's embrace of data center power as a financial asset is a tactical response to the realities of AI's energy appetite. Investors are no longer just betting on real estate or technology-they are underwriting the physical flow of electricity to the world's most demanding computing customers. The market's future will be shaped by how well these facilities can adapt to shifting power constraints, tenant needs, and refinancing cycles. For now, every coupon paid to bondholders is a direct function of the megawatts delivered to AI's insatiable engines, and the risks are as tangible as the substations and server racks that anchor this new asset class.

Unlike traditional real estate, where square footage and location drive value, data center bonds are fundamentally tied to the ability to deliver and manage power at scale. This shift forces both issuers and investors to grapple with operational, technological, and regulatory risks that are unique to the intersection of AI, infrastructure, and finance. As the market grows, transparency around power contracts, tenant concentration, and facility upgrades will become critical for pricing risk and protecting investor capital.