Oracle (ORCL) has cited force majeure to shield itself from problems tied to a data center build. I hold a long position in Oracle, so this view is biased, but the reasoning stands on its own.
What force majeure means here
Force majeure is a contract clause meant for major events outside a company's control - natural disasters, hurricanes, and the like. Seeing it used in this deal is odd, since it usually covers acts of nature, not construction snags. My first worry was whether some outside event, like a hurricane in Texas, had hit. The real trigger is different: delays in getting power to the site.
How Oracle structured the build
Oracle brought in outside financiers to hold the debt and officially own the data center facilities during construction, then hand the keys to Oracle once built. Oracle earns no revenue until a facility is switched on. Oracle argues to Wall Street that this is good - it does not have to show expenses and revenues until both happen at the same moment. That structure applies to many of its deals, including Project Jupiter in New Mexico, a huge project.
The pipeline problem
The delay traces to a pipeline that Energy Transfer is trying to build. Pipeline builders always want the shortest route, but sometimes the land between point A and point B does not want a pipeline on it. New Mexico is fighting the route Energy Transfer picked. The delay was already known. Today's news is the first sign from Oracle that the project may not finish when Oracle wants it done.
The key point on cost: if the pipeline has to be built longer to go around sensitive areas New Mexico wants protected, that physical expense is Energy Transfer's problem, and maybe Blue Owl's. Oracle is far removed from paying for a longer pipeline. A longer timeline is an Oracle problem, but the financial cost of the pipeline is not.
The bet: whose contract is stronger
The financiers here are Blue Owl, which agreed to hold and sell the debt because it is good at that. Oracle also has its own finance people, arguably the best in software, maybe in all of technology. Oracle contracts are legendary in Silicon Valley for being strong and carefully worded. The bet an investor had to make this morning: is Oracle's contract better for Oracle than it is for Blue Owl? Historically, betting on the strength of an Oracle contract is a good bet. That is the bet I have taken.
By citing force majeure, Oracle told Blue Owl in advance that if a force majeure event happens, Oracle will not be stuck holding the bag - Blue Owl will. To me this shows Oracle outsmarting the financiers again.
Is this a warning sign for the wider market?
The whole industry will now study Oracle's contracts to see if their own contracts are as strong. There will be problems in these builds. One of Oracle's CEOs, Clay Magouyrk, said on the last earnings call, paraphrased, that if you do not hit problems as you develop a data center, you are not trying hard enough. If you do not push hard, you will not reach the best revenues or biggest businesses. Delays are expected when building the biggest things ever built in America.
Yes, there is a known problem getting power to these data centers, and bigger problems will come as more are built. But this is construction, so delays are normal. My view is that this delay is actually an advantage to Oracle. If power problems in New Mexico cause delays that are not catastrophic, Oracle gets to stretch its capital spending over a longer period. If the timeline slips from 2028 to 2029, that spreading of spending helps.
It also helps OpenAI, which would pay Oracle later for the data center, easing OpenAI's cash position. (There is no Anthropic in this deal, only OpenAI.)
IPO timing
Does a delayed data center change IPO timing? If OpenAI does not need the money as fast as thought, it could push out an IPO. My suspicion is OpenAI will keep pushing toward a November IPO. Both an Anthropic IPO and an OpenAI IPO are already seen as possibly delayed into next year. SB Energy, backed by SoftBank like OpenAI, faced what looks like a valuation issue, not a business problem.
The bull and bear case
These businesses are racing to build data centers as fast as possible. I watch infrastructure spending closely because I think those are still the best opportunities - semiconductors, photonics, and fiber connecting data centers to each other. Spending in those areas is running as fast as builders can move.
The bears on AI say OpenAI and AI use are not boosting company productivity, that AI is overhyped, and these tools will never get used. They point to Tesla and xAI: Grok was a disaster, xAI built too much data center capacity, then had to rent out the unused space. Meta has been looked at for the same thing. But when xAI went to rent out its data centers, it found it could earn more revenue from renting than from its own products, and prices were rising. Talking with dozens of VCs every week, all their AI startups are desperate for more GPU access and data center capacity. The demand seems real.
The open question with any delay is who is left holding the bag. I am not calling this a failure, because I do not think it will be. Delays will happen. When they do, Oracle has moved to make sure the problem lands on Blue Owl, not on Oracle.
Bloom Energy fell today. SoftBank, a big investor in OpenAI, dropped sharply, after raising $11 billion through a junk bond sale in dollars and euros. These are the tentacles that worry the market, but the core structure protects Oracle.
