Two AI stories landed this week and they tell the same story from opposite ends. Oracle quietly told the SEC that AI helped it cut 21,000 jobs. Meta spent the same week handing out eye-watering paychecks to pull four researchers out of OpenAI. Same technology, very different week depending on which side of it you were sitting on.
Oracle Cut 21,000 Jobs and Put AI Right in the Filing
Oracle's annual filing dropped on June 22 and the number inside was hard to miss. The company is down 21,000 people over the past year, from 162,000 to 141,000. That is about 13 percent of the staff, gone. And Oracle did not tap dance around the reason. The filing flat out says AI adoption across operations led to workforce reductions, and it warned there is more restructuring coming.
Here is why that matters. Companies have blamed layoffs on macro headwinds and rightsizing for years. Putting AI in a legal document is a different animal. That is the kind of language lawyers argue over for weeks. When it lands in an SEC filing, somebody decided it was true enough to defend in court.
Now here is the part nobody wants to say out loud. Oracle burned negative $23.7 billion in free cash flow and ran capex up 162 percent to $55.7 billion, a big chunk of it chasing that $300 billion OpenAI compute deal. So they are spending like crazy on AI infrastructure and trimming payroll at the same time. The robots did not get cheaper. The humans just got cut to help pay for the robots. Watch whether the AI actually picks up the slack, or whether this is a balance sheet move wearing an AI costume.
Meta Paid a Small Fortune for Four People
Same week, Meta pulled four more researchers out of OpenAI into its new superintelligence group. Names like Trapit Bansal, who helped kick off OpenAI's reasoning work, plus Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai. Zuckerberg has been throwing around comp packages that run into the tens of millions, and reportedly up near $100 million for the right name.
Think about the contrast for a second. Oracle is proving AI can replace thousands of workers. Meta is proving the people who build the AI are worth more than entire departments. Four humans, priced like a small company. That tells you exactly where the power sits right now, and it is not with the average employee.
This is the whole AI economy in one snapshot. The folks who can build a frontier model are getting paid like star quarterbacks. Everybody downstream is getting automated. If you work in tech and you are wondering which pile you are in, that is the question worth losing sleep over. Good news is the skills that put you in the first pile can be learned. Bad news is the window to learn them keeps getting shorter.
The Bottom Line
Two filings, two press cycles, one lesson. AI is not coming. It already showed up, cut the check, and signed the paperwork. Figure out where you add value that a model cannot, and go be undeniable at it.