The Bunker and the Library

Source: “Mark Zuckerberg’s AI ambitions back in the spotlight as Meta execs begin ‘moonshot’ mission for $9.5 trillion valuation and massive payouts” — Amanda Gerut, Fortune, April 28, 2026.

Meta’s board just told us what it believes the future is worth. Five executives — not Mark Zuckerberg, who already holds about $230 billion in stock — were granted seven tranches of options each, with strike prices ranging from $1,116 to $3,727 per share. The stock trades today at $671. To make the highest tranche profitable, Meta would have to reach a market capitalization of $9.46 trillion. No company in human history has ever done that. It would be nearly twice the size of Nvidia, today the most valuable company on earth.

If they get there, the five — Andrew Bosworth, Christopher Cox, Susan Li, Curtis Mahoney, and Dina Powell McCormick — will collect close to a billion dollars each. Meanwhile Meta is spending $115 to $135 billion this year on capital expenditure, almost all of it directed at AI systems whose entire commercial premise is to perform cognitive labor that human beings currently perform.

The capital strand is being priced for a moonshot. The labor strand is being priced as the fuel. That’s the argument.

I want to write this for the person reading the Fortune piece on a Wednesday evening, sensing something is off. You are not imagining it. The accounting that produced this award lets a company book labor as cost and capital substitution as productivity — and it has arrived at its logical endpoint: a payout explicitly tied to making the human substrate of the economy redundant.

The shop floor tells the same story, smaller

I have run plants before this one. At an automotive parts plant in Tuscaloosa, Alabama — molding and assembling components for a German carmaker down the road — we installed automation that lifted throughput sharply. The operators who fed the learning, who spotted the failure modes, who taught the engineers what the line actually does at 3am on a Saturday, did not receive option tranches at $1,116, $1,500, and $3,727. They received the same hourly rate and, if it was a good quarter, a pizza party.

This was not because anyone at that company was greedy. It is because the accounting system we all operate inside treats their contribution as cost and the automation as capital. The productivity gain flows up the ledger to the equity holders. The operators who made the gain possible were recorded as the expense category the plant worked hard to reduce.

Meta is doing the same thing one layer up, at a scale where the social consequences cannot be hidden in a footnote. The engineers writing AI training pipelines, the data labelers in Nairobi and Manila, the content moderators absorbing the worst of the internet — all of them are formed capability that Meta is converting into shareholder value. Five people will collect close to a billion dollars each if the conversion succeeds.

If you sense this is unjust, you are not confused. You are reading the ledger correctly.

What patient formation looks like, and what we saw it become

There is a counter-tradition in industry, and it is worth naming, because it reminds us the current arrangement is a choice.

General Electric built Crotonville in 1956 — the first corporate university in America. Jack Welch poured roughly a billion dollars into it across his tenure and personally taught there about 250 times. The output wasn’t just GE executives. It was a generation of American CEOs — Bob Nardelli at Home Depot, Jim McNerney at Boeing, Jeff Immelt at GE itself. Fortune 500 boards literally bid on Crotonville graduates.

This is not abstract for me. I came to Chrysler in part through Tom LaSorda, who helped build Lansing Grand River into the JD Power Gold plant where I spent some of my best years at GM. At Chrysler we were briefly led by Bob Nardelli, brought in by Cerberus from Home Depot — another Crotonville graduate. I worked one rung removed from the people that system formed, and I saw what happens when patient formation collides with extractive ownership: the formation does not survive the collision, and neither does the asset.

My wife Jayanthi saw the other end of the same arc. She worked at GE’s John F. Welch Technology Centre in Bangalore — the largest R&D campus GE ever built outside the United States, opened in 2000 to extend the leadership factory’s logic to engineering at global scale. She was there toward the end of the GE locomotive business, before it was merged into Wabtec in 2019 and the operating spine Crotonville had been built to serve was sold off the balance sheet. Between us we watched the same institution from both ends — the leadership factory in America, the engineering factory in India — while the financialized parent decided it no longer wanted the operating businesses either was built to serve. The formation did not stop working. The owners stopped wanting what it produced.

IBM is the rhyme. Thomas Watson Sr. opened the Endicott schoolhouse in 1916, and for half a century the Basic Beliefs and the engineering culture were formation. By the 2010s, having stopped forming the capability, IBM tried to buy it — overpaying for Watson, selling Watson Health for parts at a loss, paying $34 billion for Red Hat. Formation is not a thing you can wire-transfer.

Caterpillar is the example most people have never heard framed this way, and the most powerful one because it is still working. Cat’s dealer network — many dealerships now in their third or fourth generation of family ownership — is itself a formation system, developing dealer principals, service technicians, and parts managers across decades through its own learning centers and a college partnership in Peoria.

When Caterpillar moved its headquarters from Peoria to Deerfield and then to Irving, Texas, the symbolic loss for central Illinois w