Oracle’s $16 Billion Michigan Data Center Financing Signals Tightening Credit Conditions for AI Infrastructure
PIMCO-backed deal closes after months of delays, revealing institutional skepticism even as hyperscalers commit $690 billion to AI capex in 2026.
Oracle closed $16 billion in financing for a Michigan data center on 25 April 2026, ending a protracted negotiation that saw one lender withdraw entirely and borrowing spreads widen to non-investment grade levels—a signal that even mega-cap tech infrastructure faces mounting institutional scrutiny despite insatiable AI compute demand.
The deal—structured as $14 billion in bonds from Bloomberg-reported PIMCO-managed funds and Bank of America, plus $2 billion in equity from Blackstone and Related Digital—took five months longer than initially projected. Blue Owl Capital withdrew in December 2025, while Blackstone paused talks in September 2025 until local zoning disputes were resolved, according to Bisnow.
Lenders demanded changes to Oracle’s lease terms to guarantee payments regardless of whether all data center capacity is utilized, per Yahoo Finance. The bond yield reached approximately 7.5% in early April—roughly 100 basis points above Oracle’s existing 2040 debt—while TD Cowen analysts noted in January that borrowing cost spreads for Oracle-linked projects had widened to non-investment grade levels.
The Michigan Site: Strategic Compute Positioning
The 250-acre Saline Township facility will deliver 1+ gigawatt capacity across three buildings with closed-loop cooling and LEED certification, according to Related Digital. Construction began in February 2026, with delivery expected through late 2026 into 2027. The Michigan Public Service Commission approved power supply agreements on 18 December 2025, authorizing DTE Energy to provide up to 1.4 gigawatts.
“Demand for digital infrastructure continues at a breathtaking pace, driven by AI and the broader digitalization of the economy. This investment is another way we are capitalizing on this generational opportunity and helping provide much needed compute capacity.”
— Nadeem Meghji, Global Head of Blackstone Real Estate
Oracle funds 100% of power costs including battery storage, generating an estimated $300 million in savings for DTE’s existing customers via fixed cost distribution. The arrangement reflects utilities’ willingness to prioritize hyperscaler demand when infrastructure costs are externalized—a model likely to proliferate as Tech Insider reports US utilities plan $1.4 trillion in capital spending through 2030, up 27% from the prior year’s $1.1 trillion projection.
Capital Intensity Meets Institutional Skepticism
The Michigan deal represents one slice of Oracle’s unprecedented infrastructure campaign. The company announced plans to raise $45-50 billion in 2026—half through equity, half through senior unsecured bonds—to fund AI buildout, according to Oracle Investor Relations. Additional mega-deals include $38 billion for Texas and Wisconsin facilities and $18 billion for New Mexico.
Oracle’s Q4 2025 capex reached $12 billion, with the majority directed to data centers and equipment. The company is currently operating with negative free cash flow as it prioritizes infrastructure expansion over near-term profitability—a posture enabled by its investment-grade balance sheet but increasingly scrutinized by bond investors.
The protracted financing timeline suggests institutional capital remains available for AI Infrastructure at scale, but at a price. PIMCO’s willingness to anchor $14 billion in debt demonstrates appetite for lease-backed data center exposure, yet the spread premium and covenant demands reflect genuine concern about Oracle’s debt trajectory. When even a company maintaining investment-grade status faces borrowing costs comparable to junk-rated issuers, the era of frictionless mega-cap tech financing has clearly ended.
The Broader AI Capex Sprint
Oracle’s $16 billion Michigan project sits within a $690 billion Hyperscaler Capex wave in 2026. Amazon, Google, Meta, and Microsoft plan approximately $630 billion in combined spending, with Oracle adding another $50 billion, per Futurum Group analysis. Goldman Sachs estimates the US faces an 11-gigawatt data center capacity shortfall currently, expected to exceed 40 gigawatts by 2028.
| Company/Group | Estimated Spend |
|---|---|
| Big Four (AMZN, GOOGL, META, MSFT) | ~$630B |
| Oracle | $45-50B |
| US Utilities (through 2030) | $1.4T |
Michigan’s selection reflects both OpenAI’s Stargate initiative to build domestic AI supply chains and DTE Energy’s capacity to source sufficient power with Oracle bearing full infrastructure costs. “This project will help ensure Michigan is a key part of building the AI infrastructure that will power the next generation of American innovation,” said Peter Hoeschele, vice president of industrial compute at OpenAI, per Data Center Dynamics.
What to Watch
Oracle’s remaining 2026 financing pipeline—$56 billion across Texas, Wisconsin, and New Mexico—will test whether the Michigan deal’s terms represent a new baseline or an outlier. If PIMCO and other institutional investors demand similar spreads and covenant protections for subsequent projects, Oracle’s effective cost of capital could rise materially even as its nominal investment-grade rating holds.
- Institutional bond investors remain willing to fund mega-scale AI infrastructure, but with risk premiums and covenant demands that signal heightened scrutiny of Oracle’s leverage.
- Energy Grid capacity—not capital availability—is emerging as the binding constraint on hyperscaler buildout, driving $1.4 trillion in utility spending through 2030.
- Oracle’s willingness to fund 100% of power infrastructure costs creates a replicable model for utilities to offload capital risk while prioritizing hyperscaler demand.
- The five-month financing delay and lender withdrawals suggest even investment-grade tech borrowers face execution risk when deploying capital at unprecedented scale.
The bifurcation is clear: capital remains abundant for AI infrastructure, but increasingly expensive and conditional. Meanwhile, the physical constraints—power supply, grid upgrades, equipment lead times—may prove more binding than balance sheet capacity. Oracle’s Michigan deal closed, but the terms reveal a market recalibrating risk even as demand remains insatiable.