On May 26, 2026, a 387-foot Feadship superyacht called Launchpad passed through Seattle’s Ballard Locks and slid into Lake Union, where it moored a few blocks from Meta’s offices. It is worth about three hundred million dollars. It belongs to Mark Zuckerberg, who was not aboard. The boat came by itself.
The same day, Meta disclosed it was cutting nearly fourteen hundred jobs in Washington state, roughly a fifth of its workforce there. A crowd gathered along the walkway to look at the boat. Some of them booed and heckled the crew. The lock operator, who had no idea he was writing the lede of every column that would follow, told GeekWire it was the biggest one he had handled in fourteen years. The next day Wingman arrived, the two-hundred-and-sixty-two-foot, hundred-million-dollar support vessel that carries the crane and the helipad and the spa, and which somebody accurately described as a floating garage for billionaires. An artist set up an easel on the shore and painted the yacht on fire.
Divide three hundred million dollars by 1,395 laid-off employees and you get about two hundred and fifteen thousand dollars a head. The boat that came to look at Seattle on layoff day cost more than a year of employment for every person Meta cut in the region. The timing was a coincidence. I believe that. It does not matter even slightly, and the fact that it does not matter is the entire story.
The part that is not funny
This week the Wall Street Journal reported that the AI backlash has tech executives fearing for their lives. The facts in that piece are real and none of them are amusing. On April 15, a man followed an employee through a badge reader into Anthropic’s San Francisco lobby carrying an envelope marked with a senior executive’s name and told a security guard that the executive was going to be killed, that he had come to warn someone. It ended without violence and without an arrest. Five days before that, a Texas man attempted to firebomb Sam Altman’s home. He was charged with attempted murder and attempted arson, and investigators recovered a manifesto calling for the killing of AI chief executives and investors. Altman responded by posting a photograph of his husband and their baby, saying he hoped it might dissuade whoever came next. A threat-monitoring firm logged a sevenfold increase in digital threats against AI leaders and data centers between late February and May.
So let me say the obvious thing once, plainly, so nobody has to wonder where I am standing. Throwing a Molotov cocktail at a house with an infant inside it is the act of a lunatic. The man who did it is a criminal defendant, not a folk hero, and anyone treating him as one has confused a felony for an argument. I want nothing to do with any version of this story where that gets winked at.
Now. Having said that.
You repeatedly told everyone you are working on a future without us
In May of 2025, Dario Amodei sat down with Axios and said AI could eliminate half of all entry-level white-collar jobs within one to five years and drive unemployment somewhere between ten and twenty percent. He said we need to stop sugar-coating this. He said that the producers of the technology have an obligation to be honest about what is coming. He described a world in which cancer is cured, the economy grows at ten percent a year, the budget is balanced, and a fifth of the country has nothing to do. He said all of it, notably, on a day he had spent onstage demonstrating the product that would do it.
Twelve months later, Fortune covered him onstage in Lower Manhattan beside Jamie Dimon, reaching for the Jevons Paradox, the nineteenth-century observation that efficiency gains grow demand rather than shrink it. The headline wrote itself: the man who spent a year warning of a white-collar bloodbath is changing the narrative. And the Journal’s own reporting confirms this is not one guy having second thoughts. Industry leaders who had been issuing dire warnings about what AI would do to the workforce have pivoted to talking up the benefits. They have not, of course, slowed the models down.
Alex Karp, to his credit, said the quiet part into a live microphone at a conference on AI and labor. Tell people their job is going to disappear, he observed, and they go for the pitchfork.
Yes, Alex. That was the campaign. There was a budget for it. Doom was the pitch deck. A technology that might end the world is an extraordinary thing to own, and for two years it was described that way constantly, by the people raising money against it, because a technology that might end the world is worth a trillion dollars and a technology that tidies up your spreadsheet is worth considerably less. The apocalypse was load-bearing. It held up the valuation. And it worked. The public believed every word. That was the one part of the plan that went perfectly.
Where I am standing
I should say this plainly too, because it determines whether anything else I write here is worth your time. I build with this stuff. I ship production agentic systems for people who pay me because the thing has to work on Monday morning. MCP servers, RAG pipelines, vector search, the whole unglamorous plumbing. I am not an AI skeptic and I am not going to pretend the technology is fake, because I would be lying and you would catch me inside a paragraph. It writes real code for me every day.
That is exactly why the executive claims irritate me. I know what it does. I know what it does not do. And the gap between those two things is where a hundred thousand people lost their jobs.
The trade
Through June of this year, Challenger, Gray and Christmas counted AI as the stated reason in 101,743 announced American job cuts, roughly a quarter of everything, nearly double the 54,836 attributed to it across all of 2025. Since the firm began tracking the category in 2023, about fifty-nine percent of the all-time total arrived in the first six months of 2026 alone. In May, AI accounted for 38,579 cuts, forty percent of everything announced that month and the highest monthly figure ever recorded for the reason. It led all reasons for four consecutive months. Technology announced 139,156 cuts in the first half, up eighty-three percent year over year, close to a third of every job lost in the country.
Over the same stretch, Meta, Amazon, Microsoft and Alphabet committed something on the order of seven hundred and twenty-five billion dollars in capital expenditure, up about seventy-five percent from last year. Meta alone projects a hundred and forty-five billion. Andy Challenger’s read on his own data is about as dry as a man can get: whether or not the individual jobs are being replaced by AI, the money for them is.
That is the trade, stated without decoration. Payroll became GPUs.
And it did not work
Here is where this stops being a story about greed and becomes a story about competence, which is worse.
Klarna is the canonical case and everybody knows the first half. Seven hundred customer service jobs replaced by an assistant built with OpenAI, two-thirds of the queue handled by machine, forty million dollars of annual profit, headlines on every outlet that has ever run a headline. The second half got a fraction of the coverage. Quality collapsed on anything requiring judgment, nuance, or the ability to talk down a furious human being. The projected savings never fully materialized. The company started rehiring, and the cost of recruiting, onboarding and training the replacements exceeded what the cuts had saved in the first place. Sebastian Siemiatkowski’s summary ran four words. We went too far.
He is not an outlier. He is the template. IBM ran the same experiment on its HR function and rehired when the system could not handle anything subjective. Commonwealth Bank of Australia reversed forty-five layoffs after concluding the roles had never been redundant to begin with. A Careerminds survey in February found two-thirds of companies that executed AI-driven layoffs had already begun rehiring, more than a third of them bringing back over half of what they had cut, most starting within six months of the announcement. Orgvue found that thirty-nine percent of business leaders had reduced staff because of AI and that fifty-five percent later conceded those decisions were mistakes. Gartner surveyed three hundred and fifty executives at companies doing a billion dollars or more and found no meaningful relationship between cutting workers and improved returns, and now projects that by 2027 half the companies that blamed AI for headcount reductions will have rehired for the same functions under different titles. Forrester calls it the layoff boomerang. Bloomberg went and looked at British job losses attributed to AI and found they were mostly driven by ordinary economic conditions, with the AI story serving as cover for cuts that were happening regardless.
Read that sequence again, because it is the whole indictment and it has nothing to do with anybody’s boat. They told you the machine was coming for your job. They cut the job. The machine could not do the job. They hired somebody back to do the job, quieter, cheaper, sometimes offshore, under a different title, and issued no press release about it. And the executive who ran that play did not give the money back, because the money was never really about the machine. The money was about the quarter.
Then they sent you the bill
Meanwhile, somebody had to pay for the electricity.
Fortune reported this week that the data center buildout has already pushed American electricity costs up by roughly twenty-three billion dollars. PJM’s own independent market monitor attributed sixty-three percent of the 2025/26 capacity price increase, some 9.3 billion dollars in a single year, to data centers. The capacity auction went from $28.92 per megawatt-day to $269.92, then hit its $333 ceiling in December and still failed to secure enough power to rule out blackouts. Residential rates are up forty-two percent nationally across five years, and considerably more in the corridors where the buildings actually went. Ninety-four percent in the District of Columbia. Seventy-four in Maryland. Seventy-three in Maine. Fifty-eight in New York.
John Steinbach has lived in the same house in Manassas, Virginia for nearly forty years. His January bill was two hundred and eighty-one dollars against about a hundred the month before. He is not an activist. He is a man who opened an envelope.
None of this is an accident or an unfortunate side effect. It is the default setting. Under the tariff rules covering most of the country, the cost of new generation and transmission is socialized across every ratepayer on the system unless a specific rule says otherwise, and for most of the country that rule does not exist yet. The retiree in Manassas is a co-investor in the buildout. She did not get a term sheet. In Virginia, roughly eighty percent of the jurisdictions that have been reviewed are bound by non-disclosure agreements, so she cannot read the deal she is funding either.
A million dollars a job
And then, having paid for the power, the community pays again.
Good Jobs First, who do the unglamorous work of actually counting this, report that data center subsidies are structured to reward capital investment rather than employment, with the result that taxpayer cost routinely exceeds one million dollars per permanent job. Not per job per year. Per job. Georgia revised its FY2026 projection upward by 664 percent, to two and a half billion dollars. Ohio’s tax department disclosed a 1.6 billion dollar loss for 2025 and the governor paused the program for new applicants days later, which tells you that nobody in the building had the faintest idea what it cost until somebody sat down and added it up. Virginia is at 1.94 billion for FY2025 counting state and local. Texas is tracking toward as much as nine billion over five years. Indiana admitted to 655 million and separately handed Amazon Data Services a package estimated at 8.2 billion dollars, the largest known state subsidy in the dataset.
Here is the number that should end the conversation. In a Mortenson survey of data center owners, the actual buyers, asked directly, only about three percent ranked tax credits and local incentives as the most important factor in choosing a site. Three percent. The states have been paying billions of dollars for a decision that the people making it say was made on entirely different grounds. They were not buying the industry. They were, at most, buying the county, in a seller’s market, at a price they set themselves, against no competing bid anyone ever verified.
That is not capitalism. That is not even a bad negotiation. That is a man walking into a dealership and offering sticker plus nine percent because the salesman looked tired.
The part the Journal never quite says
So yes, the public is angry, and the public is right to be angry. And here is the fact I think matters more than anything else in this entire story, which I did not find in the coverage of executives and their bodyguards.
The anger is working. And it is not working through arson.
In the first quarter of this year alone, community opposition blocked or delayed roughly a hundred and thirty billion dollars of American data center projects, approximately matching the total for all of 2025. Cancellations went from six in 2024 to twenty-five in 2025 to twenty in a single quarter of 2026. The number of active opposition groups more than doubled in three months, from 396 to 833 across forty-nine states. Seattle’s council received better than ninety-three thousand constituent emails about five proposed campuses. Gallup has seventy-one percent of Americans opposed to a data center near them, forty-eight percent strongly, which is higher than opposition to a nuclear plant and shows no meaningful variation by age, race, education, income, or whether you live in a city or a field. Two days ago New York became the first state in the country to halt new large-scale data center construction outright. JLL now ranks community support as the second most important site-selection criterion in the entire industry, behind only speed to power.
That is what actually has them worried, and it should. Zoning boards are beating them. County commissioners are beating them. Retirees with printed-out utility bills are beating them. The arsonist accomplished precisely nothing except making the industry sympathetic and handing every executive in the Valley a reason to stop answering questions. Every Molotov thrown at this business buys it a week of cover it did not earn. The single most effective thing anyone has done to this industry in eighteen months was to show up to a public meeting on a Tuesday night.
The bodyguard who cannot look like a bodyguard
Which brings me back to the security detail, and to the strangest and most revealing thing in the Journal’s entire piece.
A Silicon Valley security firm told the reporters that tech executives, unlike musicians or politicians, tend to request less conspicuous protection. The musicians want the enormous guy. The executives want the armed professional who does not look like one.
Sit with that for a second. They want the bodyguard. They cannot be seen wanting the bodyguard. Because being visibly protected means being visibly afraid, and being visibly afraid breaks the character, and the character is the asset. You cannot be Prometheus with a man in an earpiece standing behind you. You cannot simultaneously be the person building the most consequential technology in human history and a regular guy who is honestly a little puzzled why anybody would be upset with him. The grandiosity is load-bearing for the valuation and unaffordable for the personal life. They are going to have to pick one, and we all know which one they will pick, because it is the one with the money in it.
Zuckerberg’s 2025 compensation came to 25.1 million dollars, and Meta disclosed that the majority of it went to personal security at his residences and during his travel. His paycheck is a bodyguard. That was true before any of this started. The man has been living inside the answer for years.
They can go sit on the boat if they want. The boat is a terrible hiding place. It is a GPS transponder with a hot tub, it has a crew of twenty who all own phones, and MarineTraffic is free. But that was never the point, and I am not suggesting anybody needs to hide, because I do not want a single thing to happen to any of them and neither should you.
I am pointing out that the boat came to Seattle on layoff day, all by itself, with nobody aboard, and parked where the people who had just been fired could see it from the window.
Nobody had to threaten anyone for that to be true.
About the Author
Gal Ratner is the founder and CTO of Inverted Software and WhiteStar Labs, and Chief Architect at Prana Entertainment. He has spent nearly thirty years shipping production software on the Microsoft and .NET stack for clients including Microsoft, Sony, Rockstar Games, 2K Games, Best Buy, and Allegiant Air. He was employee number six at Break.com and a Los Angeles Business Journal CTO of the Year finalist. His current work is production agentic AI: MCP servers, the Microsoft Agent Framework, RAG pipelines, SQL Server 2025 vector search, and the PLogger observability framework. He writes about the AI industry from inside it, which is a polite way of saying he has to make the promises work after the executives finish making them.
He lives in Las Vegas, trains Brazilian jiu-jitsu under Sergio Penha, rides motorcycles, co-hosts Edge Grip Podcast, and wrote a novel called The Archive of Lost Suns. He is the .NET AI guy that ships.