Key Takeaways

  • BlackRock is marketing high-grade bonds due in 2048 to finance a 1-gigawatt AI data center campus in El Paso, Texas.
  • Initial pricing reflects a wider risk premium as investors assess rapidly expanding AI infrastructure debt.
  • The offering could indicate how much long-duration funding markets can absorb as technology companies increase capital spending.

BlackRock Inc. has begun marketing $12.3 billion of high-grade bonds for a Meta Platforms Inc. data center project, putting one of the largest AI infrastructure financings in front of investors during a more cautious period for technology credit. The transaction arrives after an AI-related debt selloff increased scrutiny of spending commitments, project economics and the volume of new bonds entering the market.

Sopaipilla Investor, a holding company tied to BlackRock, is offering the financing as a single tranche of notes due in 2048, according to Bloomberg. Initial price talk indicated a premium of about 2.875% over comparable Treasuries. JPMorgan Chase & Co. and Morgan Stanley are managing the offering.

The proceeds are intended to finance a data center campus in El Paso, Texas, with as much as 1 gigawatt of computing capacity for artificial intelligence workloads. BlackRock subsidiaries Global Infrastructure Management and HPS Investment Partners co-own the project via a dedicated holdings vehicle alongside Meta Platforms Inc. That ownership structure places most of the capital exposure with BlackRock’s investment businesses while keeping Meta directly involved in an asset supporting its computing requirements.

A 2048 maturity requires bondholders to evaluate far more than near-term demand for AI processing. They also have to consider operating costs, equipment refresh cycles, power availability, cooling requirements and the possibility that computing architectures will change over the life of the debt. A data center building may operate for decades, but servers and accelerators can become economically dated much sooner.

That helps explain the 2.875% spread under discussion. High-grade status can make the debt accessible to a broad institutional audience, yet investors still appear to be seeking additional compensation for duration and project risk. The pricing process will show whether buyers view the Meta Platforms Inc. relationship and BlackRock Inc. sponsorship as sufficient support, or whether the recent flood of technology issuance has reduced available capacity.

Supply is becoming harder to ignore. Since early 2025, AI hyperscalers including Alphabet, Meta, Amazon and Oracle have issued more than $300 billion in bonds to finance AI and data center expansion, according to Bloomberg data. Barclays found that AA and A-rated bonds represent 52% of the US investment-grade index in 2026, up from about 46% in 2021, with large-cap technology issuance contributing to the change. In practical terms, portfolio managers are being asked to allocate more of their credit exposure to the same broad investment cycle.

How much AI infrastructure debt can the market absorb before even strong issuers face higher funding costs? There is no single threshold. Investor appetite can shift with Treasury yields, construction milestones, power contracts and expectations for AI revenue. Still, BlackRock’s offering provides a visible test because of its size, long maturity and concentration in one project vehicle.

Physical performance will matter too. Investors evaluating large campuses commonly examine resilience through systems such as the Uptime Institute Tier Classification, which assesses data center infrastructure reliability and redundancy. Energy performance can also be reviewed through the ISO ISO/IEC 30134 family of data center efficiency metrics. These standards do not remove financial risk, but they can give lenders and operators a more consistent basis for examining availability, energy use and operating design.

That said, the project also illustrates why infrastructure investors remain interested. A 1-gigawatt campus is a scarce, capital-intensive asset at a time when power and computing capacity are central constraints on AI deployment. If the bonds clear near the indicated level, the sale could support further project-finance structures involving hyperscalers and infrastructure funds. If investors demand wider spreads, future AI campuses may require more equity, stronger contractual protections or phased financing rather than one very large, long-dated transaction.