Even Oracle can’t get all the money it wants for AI. The speculation has been building since late last week, kicked off by a report from financial analyst TD Cowen that percolated through financial media first. It centers on the five-year, $300 billion contract Oracle has with OpenAI (ChatGPT) that requires extraordinary financing to fulfill. According to the report (not publicly available), the capital expenditure needed is estimated at $156 billion to build or lease datacenters. To raise this, TD Cowen projects that the former Cerner, now Oracle Health, would have to be sold, as well as a potential cut of 20-30,000 jobs, about 10-15% or more of the current workforce, saving $8 billion to $10 billion in cash flow.
Why doesn’t Oracle go to the markets and banks, hold out the cup to finance this capital expenditure, and let them fill it? US lenders are apparently growing shy on lending for more AI datacenters, while Asian lenders are willing to lend funds albeit at premium rates. In TD Cowen’s analysis, Oracle is having great trouble financing this massive buildout. Investor nervousness shows in Oracle’s credit default swap (CDS) spreads tripling, as well as pressure on their stocks and bonds.
Let’s look at some factors why across several reports:
- Oracle’s already raised $58 billion in the past two months: $38 billion for facilities in Texas and Wisconsin, and $20 billion for New Mexico.
- Since September, lenders have doubled their interest rates on these Oracle projects to near non-investment grade levels.
- Oracle’s credit default swap (CDS) spreads have tripled,
- Private datacenter operators can’t get financing, so Oracle can’t fill the gap with leases, even though a few months ago Oracle was able to lease 5.2GW of US data-center capacity across Texas, Wisconsin, Michigan, and New Mexico specifically for OpenAI.
- On top of OpenAI, Oracle is also building datacenters for Meta and Nvidia in a $523 billion total commitment
Oracle is tightening the purse strings to reduce capital needs. They now require for new customers 40% upfront on building infrastructure. Another strategy being explored: requiring customers to buy their own hardware in BYOC (bring your own chip) arrangements which moves that expense off Oracle’s books and onto the customer’s.
It appears to this Editor that Oracle is caught in a squeeze play. If the company doesn’t build the datacenters, it risks falling behind its massive strategy to dominate the AI datacenter business. Yet the price of this is to abandon its massive investment in healthcare, a linchpin strategy, and the customers there. And there are Federal consequences: the completely incomplete VA implementation scheduled to resume this year and the complete, but still in progress, Department of Defense system.
Let’s look at what the effect may be on Oracle Health. Oracle bought Cerner back in December 2021 for $28.3 billion–after Cerner’s troubles with the VA EHR implementation and in the midst of the Department of Defense rollout. Oracle now is a fading number 2 in health system EHR implementations. It was all Epic, all the time for the health systems attending HealthIMPACT, the conference this Editor attended over the past two days. If Oracle Health is sold, it represents a major strategic reversal for Oracle and personally for CEO Larry Ellison. Both have pumped and promoted changing healthcare through data and systems integration for the past five years. Perhaps Oracle and Ellison have gotten the data “milk” they want and will sell the ‘cow’ now that Old Bossy is not in great shape. Can the ‘cow’ on the block even get the $28 billion paid for it?
Sell, but to whom? Microsoft is the most probable since it is massively integrated into healthcare in multiple systems. The other two under speculation are Google and Amazon. What they have in common are recent and money-losing experiences with healthcare, closing down and selling various ventures. Google Health was shut down and scattered to the winds in August 2021, and Alphabet’s Verily, pivoting like Nureyev over a decade, is now a ‘precision health’ entity. Amazon, in the midst of layoffs, has done well with Pharmacy, but One Medical. bought during the 2022 Practice Gold Rush, remains a lump of undigested matter in Amazon’s e-commerce digestive tract. Their agita with integrations such as with kiosks for Pharmacy [TTA 9 Oct 2025] and with Prime deserve Pepto-Bismol. Unlikely: any health insurer, for both cash and regulatory reasons. Spinoff? That won’t raise cash. To be continued….
Sources for this article: CIO, The Register, and Fred Pennic at HIT Consultant, who broke it in the health tech press.

















Most Recent Comments