A pattern keeps showing up this year, and yesterday it got harder to argue with. A profitable company sheds thousands of workers, names AI as a reason in its own paperwork, then turns around and pours billions into AI hardware. Oracle just did exactly that.

Oracle puts AI in writing

Oracle cut about 21,000 jobs over the past year. The number matters, and so does where it came from. The company admitted in its own annual regulatory filing that AI played a part. The filing says AI deployment “may continue to result in reductions to our workforce.”

The cuts brought Oracle’s headcount down to roughly 141,000 from about 162,000 a year earlier. Some teams felt it far more than others.

Revenue and Health Sciences saw reductions near 30%, while cloud and AI teams were mostly spared. Oracle plans to spend around $70 billion this year building AI data centers for customers like OpenAI. The company is cutting people and building machines in the same breath.

Meta’s tracking program springs a leak

We told you yesterday about Meta’s program that records employee keystrokes and mouse movements to train its AI. It seems, though, Meta paused the program after a data leak exposed sensitive worker information to the entire company.

So, Meta did not stop surveilling workers because they objected to being watched. It stopped because the watching went wrong.

Private conversations, performance data, and transcriptions became visible across the company. The program, launched in April, recorded keystrokes, mouse activity, and occasional screen content from US staff. A company that wants to read every click could not keep those clicks private. Workers noticed.

Narrower On-ramps

Here is a story about jobs that vanish before anyone gets laid off. A new report from Opportunity@Work warns that AI could wall off millions of “gateway jobs” for workers without a four-year degree.

Gateway jobs are the entry and mid-skill roles people, typically without college degrees, use to climb into better pay. The report counts up to 11 million of them exposed to AI losses. Read that as the group’s own estimate, since they advocate for these workers.

The report notes that 33 states have agreed to drop degree requirements from public jobs. So, AI could open these doors wider or shut them, depending on how employers choose to use it.

A lot of cuts, not much payoff

Concurrently, tech layoffs are running at roughly 1,100 a day in 2026, close to double last year’s pace. Meta, Oracle, and Block all point to AI.

A Gartner study of 350 firms found that the companies cutting the most showed no improvement in financial returns. So the layoffs are not obviously paying off.

That gap is why some economists use the term “AI washing.” It describes firms blaming AI for cuts they wanted to make anyway. Treat the daily-rate figures as estimates from layoff trackers, not exact counts.

The silent killer is slow hiring

Layoffs grab headlines. The bigger shift for young workers isn’t making the front page of the news. Some economists argue that AI is hurting hiring more than it is driving firings, impacting junior staff the most.

Goldman Sachs research found AI trimmed monthly payroll growth by around 16,000 jobs over the past year. Companies are not always cutting people. They are choosing not to add them.

One organizational psychologist put the trap plainly. “The people who get laid off don’t necessarily get the next set of jobs, because the roles are different.”

India’s AI experiment on steroids

For a wider view, look at India. Infosys, TCS, and Wipro have each rolled Microsoft Copilot out to more than 100,000 workers, over 300,000 seats combined. Wipro says its performance AI review agent cut performance-review effort by nearly 70%.

At the same time, these firms are thinning out without big announcements. They lean on delayed hiring and attrition instead.

The Microsoft rollout reads like a press release, so take it with a grain of sale. Nevertheless, workers get AI tools dropped onto their desks, and the company gets to do more with fewer staff. To their credit, the firms are paying for retraining.

Average learning hours at TCS rose past 120 per employee this year, with Infosys up 58%.

Where this leaves us

So, Oracle wrote AI into the official reason it cut 21,000 people, and a respected study found that the firms cutting hardest got nothing to show for it on the balance sheet. Companies are forming around a bet on AI, and the bet has not paid off yet for many of them.

Workers, meanwhile, are mere lab rats in a grand experiment of capitalism.

Get full access to The AI & Work Chronicle at ailabor.substack.com/subscribe