Meta Is Firing 16,000 People to Fund AI That Can't Beat the Competition
Meta delayed their flagship AI model. Their solution? Fire 16,000 employees to fund the AI that isn't working.
The company just delayed its flagship model. Their solution? Cut 20% of the workforce.
Mark Zuckerberg has a problem. He's spent billions on AI talent, offered $100 million pay packages to researchers, committed $600 billion to infrastructure by 2028, and brought in Scale AI's Alexander Wang as Chief AI Officer.
The result? A flagship AI model that can't beat Google, OpenAI, or Anthropic in basic benchmarks.
And his solution? Fire 16,000 employees to pay for more of the same.
The Double Disaster
This week, two massive stories hit Meta simultaneously. Most outlets covered them separately. They shouldn't have.
Story one: The New York Times reported that Meta's new AI model, codenamed "Avocado," has been delayed until at least May. The model failed internal tests for reasoning, coding, and writing when compared to competitors. It couldn't beat Gemini 3.0, let alone GPT or Claude.
Story two: Reuters broke that Meta is planning layoffs affecting 20% or more of its workforce. With 79,000 employees as of December 31, that's roughly 16,000 people. The explicit reason? "Offset costly artificial intelligence infrastructure bets."
Read those two stories together and a clear picture emerges: Meta is cutting humans to fund AI that isn't working.
The Numbers Tell the Story
Let's look at what Meta has spent on AI:
| Investment | Amount |
|---|---|
| AI infrastructure by 2028 | $600 billion |
| 2026 capital expenditures | $115-135 billion |
| Top AI researcher pay packages | $100+ million each |
| Alexander Wang's reported package | Up to $1.5 billion over 6 years |
| Scale AI stake (49%) | Undisclosed billions |
And what have they gotten?
| Model | Result |
|---|---|
| Llama 4 | Delayed, misleading benchmarks, Behemoth abandoned |
| Avocado | Failed vs GPT, Claude, Gemini 3.0, delayed to May |
Meanwhile, their layoff history keeps growing:
- November 2022: 11,000 employees (13% of workforce)
- March 2023: 10,000 employees
- March 2026: ~16,000 employees (20% of current workforce)
That's 37,000 jobs cut since the "Year of Efficiency" began. And they're still falling behind.
Why This Matters
This isn't just a Meta story. It's the first major tech layoff where a company explicitly tied job cuts to "AI costs mounting." They're not hiding it anymore. The AI arms race is costing jobs, and the companies doing the firing are the same ones failing to win the race.
Zuckerberg said in January he was seeing "projects that used to require big teams now be accomplished by a single very talented person." But that efficiency hasn't translated to competitive AI products. The "very talented person" is still using Claude or GPT because Meta's own models can't keep up.
The Pattern No One's Discussing
Meta has a pattern here that should concern anyone watching the AI industry:
- Announce massive spending ($600B infrastructure, $135B capex)
- Poach expensive talent (Wang, Tulloch, others at $100M+ packages)
- Miss benchmarks (Llama 4, Avocado)
- Delay releases (Behemoth abandoned, Avocado pushed to May)
- Cut employees to fund more of steps 1-4
This cycle has been running for over a year. There's no evidence it's working.
A Meta spokesperson told Gizmodo: "Our next model will be good, but more importantly, show the rapid trajectory we're on." That's almost word for word what they said about Llama 4.
What Happens Next
The Avocado delay is particularly telling. According to sources, the model outperformed Meta's previous models and even beat Gemini 2.5. But it couldn't touch Gemini 3.0, which launched in November.
Here's the problem: by May, when Avocado finally ships, Google, OpenAI, and Anthropic will have moved on. The target keeps moving, and Meta keeps missing it.
Internal discussions have even included temporarily using Google's Gemini models for Meta products. That's the AI equivalent of Ford putting Toyota engines in their trucks.
The Real Question
Companies across tech are using "AI efficiency" as justification for layoffs. Amazon cut 16,000 jobs in January after their AI tools started breaking production. Block cut nearly half its staff, with Jack Dorsey explicitly citing AI tools.
But Meta's situation is different. They're not cutting jobs because AI made workers redundant. They're cutting jobs because their AI bets aren't paying off and they need to cover the costs.
That's a crucial distinction. One is progress. The other is failure.
The question investors and employees should be asking: At what point does "Year of Efficiency" become "Decade of Denial"? How many more layoffs before someone admits the strategy isn't working?
Meta has the capital, the talent, and the infrastructure. What they don't have is a model that can compete. And no amount of layoffs will fix that.
Sources: New York Times, Reuters, The Guardian, Gizmodo
