"Hyperscalers are spending upwards of $400 billion this year alone on data centers, and investors are desperate to own any piece of that growth. Two names have ridden the wave especially hard: Oracle ( ) and Digital Realty Trust ( ). Oracle is positioning itself as the cloud backbone for frontier AI models through massive OpenAI contracts and a flood of new data-center debt."
"Data-center construction is running at record pace, but power delivery is not. Amazon ( NASDAQ:AMZN ) has multiple "zombie" campuses in Oregon sitting dark because it can't make grid connections . Northern Virginia, the world's largest data-center market, faces multi-year delays for new substations. Globally, the industry needs an additional 30 to 50 gigawatts (GW) of power by 2030, yet utilities are adding only single-digit gigawatts per year. Oracle and Digital Realty are among the most aggressive builders in exactly these constrained regions."
"Oracle has issued $18 billion in new bonds this year to fund AI-optimized cloud regions. Its headline deal with OpenAI is a five-year, $300 billion commitment, but internal documents show Oracle lost $100 million in the most recent quarter on that contract alone because utilization is far below plan and depreciation is brutal. Leverage is now approaching 4× EBITDA - a level credit markets"
AI spending on data centers exceeds $400 billion this year, driving record construction activity and intense investor demand. Oracle and Digital Realty are among the largest beneficiaries, with Oracle securing large OpenAI contracts and issuing $18 billion of bonds to fund AI-optimized cloud regions. Digital Realty is rapidly expanding rack capacity to house GPUs. Power delivery is lagging construction, creating "zombie" campuses and multi-year grid delays in key markets like Oregon and Northern Virginia. The industry needs 30–50 GW more capacity by 2030 while utilities add only single-digit GW annually. Low utilization, heavy depreciation and rising leverage have eroded returns and increased credit risk.
Read at 24/7 Wall St.
Unable to calculate read time
Collection
[
|
...
]