THE $600 BILLION BET

Picture a mid-level engineer at Meta.

She joined in 2021. Spent three years building recommendation systems. Got promoted. Bought a house in the East Bay. Told her parents she worked at one of the most valuable companies in the world.

This week, her manager was asked to start drawing up a list.

She doesn't know yet. But the list exists.

Mark Zuckerberg is building something extraordinary.

Let's be honest about that first.

According to Reuters, Meta has committed $600 billion to build AI data centers across the United States by 2028. Not over a decade. By 2028. That's roughly the GDP of Sweden — deployed into servers, chips, power grids, and cooling systems in less than three years.

As Fortune reported, Zuckerberg launched a new top-level initiative called Meta Compute, personally naming three executives to lead it. The mandate: build "tens of gigawatts this decade, and hundreds of gigawatts or more over time."

For context — one gigawatt powers roughly 750,000 homes.

Meta isn't building a product. It's building a civilization-scale infrastructure layer.

And to pay for it, according to Reuters, Meta is planning to cut 20% or more of its workforce — potentially 16,000 people — in what would be the largest restructuring in the company's history.

The math is clean, if you're a spreadsheet.

$135 billion in capital expenditure this year alone — nearly double last year's $72 billion. Hundreds of millions in pay packages for AI researchers. A $14.3 billion investment in Scale AI. A $2 billion acquisition of Chinese AI startup Manus. A new AI engineering organization where manager-to-employee ratios run 1:50.

Something has to give. And it's the people.

Here's what Zuckerberg said in January, on an earnings call, to Wall Street:

"Projects that used to require big teams can now be accomplished by a single very talented person."

CNBC reported that investors cheered. Meta shares popped 10% in after-hours trading.

Wall Street heard efficiency. The engineer in the East Bay heard something else.

We talk about the investment numbers. The gigawatts. The benchmark scores. The race to superintelligence.

We don't talk about what "upskilling" actually means in practice.

It means: you have 90 days to learn prompt engineering, or we'll find someone who already knows it.

It means: the Amazon Alexa division — once 800 engineers — is now reportedly down to a skeletal crew of 23, with remaining work handed to contractors using AI coding tools.

It means: workers at Amazon were asked to spend their final weeks meticulously documenting their decision-making processes — their institutional knowledge, their creative workflows — believing they were helping with a "standard transition." They were, in fact, building the training data for the AI agents that replaced them.

That's not upskilling. That's extraction.

It's not the AI researchers getting $200 million packages. They're fine.

It's not the senior engineers who can pivot to prompt architecture and model evaluation. They'll land somewhere.

It's the middle layer — the project managers, the content moderators, the mid-level developers, the operations staff, the people who made the machine run smoothly but never wrote the code that made it think.

Business Insider noted that Meta's planned cuts would "wipe out more jobs than the entire headcount of many midsize tech companies." In absolute terms, 16,000 people is not a rounding error. It's a city.

And Meta is not alone.

Amazon cut 16,000 in January. Block cut nearly half its staff. Atlassian cut 1,600. The pattern is the same everywhere: capital moving from salaries to servers.

Not the corporate version — the press release version, where "we're committed to supporting affected employees through this transition" means a 60-day severance and a LinkedIn post.

A real just transition would look like this:

First — transparency before the list. Tell people the direction of travel 12 months before the cuts, not 12 days. Give them time to actually retrain, not just the appearance of it.

Second — meaningful retraining investment, not token gestures. If you're spending $135 billion on compute, spending $500 million on genuine workforce transition programs isn't charity. It's the cost of doing this with integrity.

Third — acknowledge what's actually being lost. Institutional knowledge. Mentorship chains. The informal networks that make organizations function. These don't show up on a balance sheet. They disappear quietly, and companies spend years wondering why things stopped working.

Fourth — policy. This is bigger than any one company. The US has no federal framework for AI-driven displacement. No portable benefits system. No retraining infrastructure at scale. The "just transition" conversation that happened around coal and manufacturing needs to happen now, before the wave fully arrives — not after.

Zuckerberg isn't wrong about the technology. The efficiency gains are real. The infrastructure bet may well be the right one. In ten years, the $600 billion may look like the most prescient capital allocation in corporate history.

But efficiency and justice are not the same thing.

A system can be extraordinarily efficient at concentrating value at the top while distributing disruption across the bottom. That's not a bug in the model. For some, it's the feature.

In Abundance vs. Scarcity, I wrote about the Recursive Engine — the way AI compounds advantage for those who already have it, while compressing opportunity for those who don't.

Meta's $600 billion bet is the Recursive Engine made visible.

The servers get smarter. The shareholders get richer. And the engineer in the East Bay gets a calendar invite she doesn't know is coming.

I don't write this to condemn Zuckerberg.

I write it because I think most people — most normal people, not the ones on earnings calls — are watching this happen and feeling something they can't quite name.

It's not just fear of losing a job. It's something deeper.

It's the sense that the rules changed while you were playing by the old ones. That the skills you spent years building are being quietly deprecated. That the social contract — work hard, stay loyal, build something — is being rewritten by people who will never feel the consequences of the rewrite.

That feeling is real. And it deserves more than a press release.

The $600 billion bet will be built. The data centers will go up. The models will get smarter.

The question isn't whether the wave comes.

The question is whether we build anything to catch the people it sweeps away.

Precision in a world of noise.

Analysis by Slone Sterling