You Cannot Buy What Can Only Be Built

Meta spent $14.3 billion last year to buy half of a data labeling company.

They made its 28-year-old CEO the head of their Superintelligence Lab. Gave him a title that didn't exist six months before. Put him above researchers who'd been doing AI before he was in high school.

Nine months later, his team reports to other people. A new group controls the data pipelines. The researchers he was supposed to lead are leaving.

And the man who was already there — Yann LeCun, one of three people alive who can legitimately be called a godfather of AI — called the new hire "inexperienced." Then quit. Then started his own company.

$14.3 billion. And the most important person walked out the door for free.

I keep coming back to this story because it's not really about Meta.

It's about a pattern I see everywhere. In tech. In consulting. In the way people build businesses. In the way countries try to buy their way into the future.

The pattern is simple: someone with a lot of money looks at a capability they don't have, and instead of building it, they try to buy it.

It almost never works. And the reason it doesn't work is the same every time.

Mark Zuckerberg wasn't stupid. He looked around in 2024 and saw that Meta was falling behind OpenAI had ChatGPT.

Google had Gemini and a decade of DeepMind research. Anthropic had Claude and a founding team that left OpenAI because they believed something different about how AI should be built.

Meta had Llama. Which was good. Which was popular with developers. But which wasn't frontier.

So Zuckerberg did what people with unlimited capital always do. He went shopping.

$14.3 billion for Scale AI. $100 million signing bonuses to poach researchers from OpenAI. $72 billion in AI capex in a single year. $125 billion planned for this year. The most aggressive infrastructure buildout in corporate history.

The logic made sense on paper. Buy the best data company. Hire the best talent. Build the biggest data centers. Outspend everyone.

Here's what happened instead.

The Llama 4 team got caught faking their benchmark results. Their own chief AI scientist confirmed it — "the results were fudged a little bit." They used different models for different benchmarks to make the numbers look better.

Their flagship model, Behemoth, flopped and the team was disbanded.

Zuckerberg lost confidence in the entire GenAI organization and sidelined it.

Researchers started leaving. LeCun said publicly: "A lot of people have left. A lot of people who haven't yet left will leave."

After all that spending — more than most countries' GDP — Meta is still behind Google, still behind OpenAI, still behind Anthropic in the conversation about who's building the future of AI.

That's not a resource problem. That's a relevance problem.

I think about this a lot because I used to believe the same thing Zuckerberg believes.

When I was 22, I thought the answer to everything was capital. Raise a Series B. Hire the best people. Scale fast. If you have enough money, you can build anything.

That's the story Silicon Valley tells. And for a very small circle — the people who already had the network, the timing, the domain expertise before they had the money — it works.

But for everyone else, money is a magnifier. It makes what you already are bigger. It doesn't make you something you're not.

Meta already had a culture that pushed out its best people. The Instagram founders left because Zuckerberg couldn't handle their product outshining his.

John Carmack, one of the greatest programmers alive, left out of frustration with Meta's leadership on the metaverse. LeCun had a decade of freedom to do esoteric research — until a 28-year-old data labeling CEO was placed above him.

Money didn't fix the culture. Money amplified the culture. And the culture was: we acquire things instead of growing them.

Here's what I've learned watching this pattern play out — in tech, in AI, in my own businesses.

The things that matter most are the things that can only be built from the inside.

Google didn't buy DeepMind and immediately become an AI leader. They bought DeepMind in 2014 and gave it a decade to compound. They let researchers research. They let culture form. They let capability accumulate slowly, unglamorously, without demanding quarterly proof that the $500 million acquisition was paying off.

That's building.

Anthropic didn't raise billions and then go hire the best people. The best people left OpenAI together because they shared a belief about how AI should be developed. The culture existed before the company did. The money came after, to fund what was already there.

That's building.

Meta spent $14.3 billion and put a data labeling CEO in charge of superintelligence research. That's buying.

The difference isn't subtle. And the market can tell.

I see the same pattern in founder mobility.

I spent years in that world. And the most common mistake I saw was people who thought buying a visa meant they'd built a life.

They'd move to Dubai or Canada or the US on an investor visa. Put money into real estate. Enroll their kids in international schools. Check all the boxes.

But they never built the local relationships. Never understood the culture. Never put down the kind of roots that hold when things get hard.

When the world shifted — and it always shifts — they were the ones stuck in airport lobbies, refreshing flight apps that showed nothing.

You can buy access to a country. You can't buy belonging.

You can buy a visa. You can't buy citizenship in the way that matters — the deep sense that this place is yours and you are part of it.

That takes time. That takes presence. That takes building.

The same thing is happening right now with AI adoption across every industry.

Companies are spending millions on AI tools, AI consultants, AI infrastructure. They're buying licenses and hiring prompt engineers and announcing AI strategies at board meetings.

But the companies that are actually getting value from AI aren't the ones spending the most. They're the ones where someone inside the organization actually understands the technology.

Where the culture allows experimentation. Where people have been building with these tools long enough to know what works and what doesn't.

You can't buy AI fluency. You can't acquire AI culture. You can't hire a consultant to make your organization AI-native.

You have to build it. From the inside. Over time. With people who care about more than the quarterly report.

I'm 27. I run a small operation. Four people. No VC. No $14.3 billion.

And honestly, I think that's the advantage.

Because everything I have was built. Not bought. Not acquired. Not poached from a competitor with a $100 million signing bonus.

My understanding of AI and sovereignty — built from years of working in immigration and watching how jurisdictions actually function.

My audience — built one essay at a time, one tweet at a time, by saying things I actually believe.

My team — built on trust that compounds over years, not compensation packages that vest over four.

My thesis — "Own or Be Owned" — built from lived experience, not market research.

None of it is fast. None of it looks impressive on a pitch deck. But none of it can be taken from me, either.

Meta can lose $14.3 billion of capability overnight because the person they bought walks out the door. I can't lose what I built because it's not housed in a single hire or a single investment. It's in the culture. The relationships. The ideas. The years of compounding.

That's the difference between buying and building.

This isn't a story about Meta being dumb.

Zuckerberg is one of the most successful founders alive. He built Facebook from a dorm room into a $1.5 trillion company.

He survived the Cambridge Analytica scandal, the metaverse pivot, and a stock price that lost 70% of its value in a year. The man is resilient.

But resilience funded by unlimited capital is not the same thing as building.

And the AI race is exposing that difference faster than any industry shift I've ever seen.

Because AI capability compounds. It rewards depth over breadth, culture over capital, building over buying.

The labs that are winning aren't the ones with the most money. They're the ones where the best researchers actually want to be.

No amount of money can make people want to be somewhere.

That has to be built.

I wrote a tweet this week about Meta. Stacked the facts. $14 billion, faked benchmarks, sidelined org, researchers leaving, still behind everyone.

The last line was: "You cannot buy what can only be built."

People keep quoting it back to me. I think because it's not really about Meta.

It's about all of us.

The job you're in. The company you're building. The life you're constructing.

Are you buying it or building it?

Are you acquiring the appearance of the thing, or doing the slow, unglamorous work of becoming the thing?

Because the world is shifting fast. And when it shifts — when the airports close, when the benchmarks get exposed, when the money stops working the way it used to — the only thing that holds is what was built from the inside.

Everything else was rented.

And rent comes due.