How Credit Markets Are Powering AI Capex Financing
Artificial Intelligence (AI) development is at a pivotal stage. Generative AI and advanced machine learning are moving from research labs into real-world applications, driving massive demand for compute infrastructure, which includes the necessary hardware, software, and networking, as well as storage infrastructure, including data centers and Graphic Processing Units (GPUs). AI development is shifting from a software-driven innovation to epic-scale ecosystem buildouts requiring significant capital expenditures (capex). For example, hyperscalers, companies that offer huge, scalable, on-demand computing, networking, and storage services, gradually evolved from asset-lite to more asset-heavy business models. This massive shift has created real bottlenecks in the power, land, fiber, cooling systems, and skilled labor required. However, because data centers also have become mission-critical infrastructure, they benefit from the policy tailwind of national security priority, which in turn is shaping industrial policy and energy investment. This “arms race” to build out AI infrastructure has created a growing financing opportunity for credit markets, particularly data center asset-backed securities.
AI Capex Boom Transforming Credit Markets
The AI capex boom is at an inflection point that marks not only a technological revolution but also a fundamental transformation in how it is financed through capital markets. Hyperscalers once financed capex buildouts almost entirely through internal free cash flow. Although AI capex remains well funded by free cash flows in the medium term for the majority of those firms, heavy capex investment is a growing piece of the cash flow pie (Exhibit 1). Cheap debt financing allows companies to tap external capital from multiple financing channels. Hence, fixed income markets are now central to AI infrastructure deployment. Issuance across investment grade corporate debt and data center asset-backed securities (ABS) and commercial mortgage-backed securities (CMBS) is accelerating and emerging as a reliable, scalable financing channel.
Credit Markets Filling the Funding Gap
Morgan Stanley research estimates that under the $2.9 trillion global capex needed by data centers from 2025 to 2028, there is a $1.5 trillion funding shortfall. Credit markets will be critical to filling this funding gap, with the estimated split encompassing $800 billion in private credit, $200 billion in corporate debt, $150 billion in ABS and CMBS, and $350 billion across private equity, venture capital, bank lending, and others (Exhibit 2).
Assessing Various Credit Financing Channels
The decision to finance via corporate bonds, ABS, or CMBS depends on the issuer’s familiarity with each asset class and the characteristics of the underlying assets, cash flow stability, tenant quality, operational flexibility, financing cost, and the stage of asset development.
- Corporate Bonds: With plenty of cash, low leverage, and high credit ratings, hyperscalers have large capacity to issue bonds. However, they may take precautions given the uncertainty of the monetization pace and instead tap the corporate bond market opportunistically. At the early stage of building data centers, they can resort to construction loans and leases from developers that build and eventually finance the stabilized assets via ABS and CMBS. Bank of America (BofA) data shows there was around $75 billion of US investment grade debt issued by AI-focused hyperscalers in September and October 2025 alone, more than double the sector's average annual issuance of $32 billion between 2015 and 2024. They expect this number to grow to over $100 billion in 2026. The BofA data also highlights the market reaction in September when hyperscaler investment grade bond spreads widened toward 80 basis points (bps) from 50bps. It remains to be seen how the market will receive a further deluge in issuance.
- Securitized Credit: For data center ABS and CMBS, total new issuance has exceeded $20 billion in 2025 year to date, eclipsing the 2024 full-year total of $11.6 billion by 50%. Over $50bn of data center-backed ABS and CMBS bonds has been raised since 2018 (Exhibit 3). Data center ABS and CMBS, which complement traditional bank lending and corporate bond markets, are generally issued only once data centers have achieved some level of stabilized performance and are operational with tenants in place. The primary difference between data center ABS and CMBS is that ABS is backed by real estate ownership and lease cash flows whereas CMBS is secured by the commercial mortgage loans. The financing raised by these issuances is often used to refinance the developers' existing debt, including the original project finance or construction loans necessary to begin the buildout. According to Morgan Stanley research, the current securitization rate of data center capacity is around 10%, and they expect to see an annual securitization rate of up to 25% by 2028, indicating significant growth potential.
Financing the Mission Critical with Stable Fundamentals
- Secular Strength in Data Center Fundamentals
The supply of data centers, the backbone of AI infrastructure, remains constrained by power availability, site selection challenges due to land and environmental issues, and regulatory and community hurdles, such as zoning, permitting, and water/noise concerns. Meanwhile, demand is very strong. According to JLL Research1, 73% of data center construction is preleased, and vacancy has been declining to around 2.3%. Tenants usually sign multi-year contracts, which typically makes them sticky with very low churn rates because switching costs are meaningful. These characteristics create strong forward cash flow stability and limited cyclical sensitivity, supporting attractive risk-adjusted yield for fixed income portfolios. - Underwriting the Credit Fundamentals
Given the exponential growth of data center financing in fixed income markets, investors need to be prudent and approach this young asset class with thorough, disciplined underwriting. Assessing the credit worthiness of underlying data center collateral requires an understanding of many complex factors, including: geographic location, age, and purpose of facility; power availability, cost, and usage efficiency; tenant quality, concentration, and churn rate; operator execution capability; lease length and structure; and other considerations. - Relative Value in ABS versus Other Credit Sectors
Data center ABS has transitioned from an esoteric niche to a sector too big to ignore. Liquidity has improved, and larger mezzanine tranches now offer scalable exposure and attractive relative value within structured credit. These securities, which are backed by real estate ownership and cash flows from stable leases and service contracts, can offer pricing advantages, moderate operational flexibility to add or remove collateral, and structural protection. The latter may include credit subordination, excess spread, or triggers to divert cash and accelerate payment when needed.
Rating agencies cap the rating of data center ABS senior tranches at single A, due to potential obsolescence risk and short history. Compared to single A-rated consumer ABS, like subprime auto and rental car ABS, and similarly rated investment grade corporate bonds, data center ABS may offer lower cyclicality, stable contractual revenue, and exposure to the tailwind of secular demand. The spread of single A-rated 5-year data center ABS offers approximately 40bps concession to 3-year subprime and auto ABS of the same rating and 75bps concession to investment grade corporate bonds (Exhibit 4). These ABS are attractive relative to investment grade corporate bonds for their higher yield and high-quality rating.
Risks to Monitor: Further into the Future
On the macro level, an AI bubble or potential crash from the current lofty valuations of hyperscalers is not our base case; we deem either outcome as premature. However, heavy stock market concentration in “Magnificent Seven” companies and lack of market breadth warrant precautions.
The primary risk is whether the data centers have sustainable access to power, network adjacency, and hyperscaler tenancy. Technological disruption, like the “DeepSeek Moment” that sent ripples through the industry in early 2025, could change compute needs if efficiency leaps accelerate. As a result, risks from collateral valuation uncertainty and technological obsolescence are front and center. However, bottlenecks of energy, fiber, and grid deployments will constrain the supply, offering some resilience. Monetization risk, which stems from when those investments can be paid back, is another major risk that has come under further pressure from debt financing. Refinancing risk and the maturity wall will become more acute if the macro market destabilizes. Lastly, tenant concentration and operator risk are also critical to ABS.
Policy Tailwinds and Market Technicals
Despite lingering risks around AI monetization, national security priority status provides support for AI infrastructure funding. In addition, elevated long-duration investment grade corporate bond issuance increases pressure on long-end Treasury rates, enhancing relative value in shorter ABS structures.
Conclusion: Owning the Backbone of AI Capex
Financing AI capex is becoming a defining component of the fixed income opportunity set. Both corporate and securitized credit, including investment grade bonds, ABS, and CMBS, have emerged as alternatives to fill the AI infrastructure financing gap and complement traditional bank lending. However, data center ABS, in particular, sits at the intersection of technological revolution and stable data center credit fundamentals. Among their potential advantages, they offer secular growth exposure, strong structural protection, scalable investment in mezzanine tranches, improving liquidity, cheaper relative valuations, and relative insulation from consumer fragility. Thoughtful, research-driven exposure to data center securitization is not just a tactical trade—we believe it is a blueprint for how the fixed income market adapts to transformative technology breakthroughs.
Note:
Tracy Chen, CFA, CAIA
Portfolio Manager
Key Takeaways
- The race to build out artificial intelligence (AI) infrastructure has become a national security priority, which in turn is shaping industrial policy and energy investment.
- Hyperscalers financed capex buildouts almost entirely through internal free cash flow. However, cheap debt financing allows companies to tap external capital from multiple financing channels.
- Fixed income markets are now central to AI infrastructure deployment, creating a growing financing opportunity for credit markets.
- Issuance across investment grade corporate debt and data center asset-backed securities (ABS) and commercial mortgage-backed securities (CMBS) is accelerating and emerging as a reliable, scalable financing channel.
- From 2025 to 2028, there may be an estimated $1.5 trillion funding shortfall in global capex needed by data centers. Credit markets will be critical to filling this funding gap.
- Data center ABS offer potential advantages, including secular growth exposure, strong structural protection, scalable investment in mezzanine tranches, improving liquidity, and cheaper relative valuations.
Share this article: 
✉
Subscribe to Around the Curve and receive our latest global macroeconomic, fixed income, and equity views directly to your inbox.
Groupthink is bad, especially at investment management firms. Brandywine Global therefore takes special care to ensure our corporate culture and investment processes support the articulation of diverse viewpoints. This blog is no different. The opinions expressed by our bloggers may sometimes challenge active positioning within one or more of our strategies. Each blogger represents one market view amongst many expressed at Brandywine Global. Although individual opinions will differ, our investment process and macro outlook will remain driven by a team approach.
