The Productivity Mirage
The Productivity Mirage
Labor Markets · Artificial Intelligence · Monetary Policy
Sunday, 10 May 2026 · Issue CCXLVII · Private circulation only
The jobs number beat by eighty-five thousand. Read it again. Then read the fine print.
Here is what happened on Friday: the Bureau of Labor Statistics reported that the U.S. economy added 115,000 jobs in April, landing nearly twice Wall Street's consensus estimate of 62,000. Markets held their gains. Yields drifted lower. Commentators reached for words like "resilient" and "solid footing." A perfectly normal Friday, apparently. Nothing to see. Move along.
Except.
Labor force participation slipped again — to 61.8%, down from 61.9% in March and a full seventy basis points below a year ago. The so-called real unemployment rate, which catches those working part-time because they can't find full-time work, jumped by 445,000 to 4.9 million. Wages? Up just 0.2% month-on-month — the same lukewarm number as March, and not the kind of wage growth that makes a dent against a CPI still running at 3.3% year-on-year after gasoline prices surged over twenty percent in a single month. The quits rate is near its floor. People aren't leaving jobs because they have nowhere to go.
The headline was fine. The subtext was a slow puncture.
The jobs number beat by eighty-five thousand. Cloudflare fired a thousand on the same morning. Both of these things are true, and together they tell you far more than either does alone.
Because while the BLS was tallying healthcare and warehouse hires, NET was busy announcing it would cut roughly twenty percent of its entire workforce — over eleven hundred people — in what the company's 8-K, filed without ceremony on May 7th, called a transition to an "agentic AI-first operating model." Revenue was up twenty-five percent in the same quarter. Growing fast, firing hard. Not a company in distress. A company in a hurry.
Matthew Prince didn't dress it up. He said the roles being eliminated "are not the roles we need for the future." In the same breath, he noted that AI usage inside Cloudflare had grown over six hundred percent in three months. You can write your own conclusion about what connects those two sentences. The market wrote its own, too: NET fell roughly 14% on the day, which was interesting — not because the layoffs spooked investors, but because investors don't yet know how to price the trade-off between a leaner cost structure and the signal of a company publicly discarding twenty percent of its people.
Cloudflare wasn't alone. BILL slashed up to thirty percent of headcount. Upwork cut a quarter of its workforce. Of the 83,387 announced layoffs tracked in April, the tech sector alone contributed 33,361 of them — and for the second consecutive month, the leading cited reason was artificial intelligence. Not restructuring. Not macro. AI. This is the first time in the history of that particular dataset that a technology has appeared as a primary layoff driver two months running. Write that down somewhere.
Meanwhile, TTD — Trade Desk — fell 13% after missing on both earnings and guidance, collecting a string of downgrades before the ink was dry. Ad-tech is supposed to be the AI beneficiary, the great aggregator of the algorithmic attention economy. Apparently the memo hasn't fully arrived. The irony of an advertising-technology company getting punished while the companies that replaced their use-case with software thrive is not subtle, but markets are not obligated to be.
The aggregate picture is becoming legible, if you want to look at it directly. The big four cloud providers — Alphabet, Amazon, Meta, Microsoft — are now projected to spend north of $660 billion on infrastructure in 2026. Meta raised its full-year capex guidance to a range of $125–145 billion, a figure so large it would rank as a top-twenty economy by government expenditure if it were a country. Microsoft spent roughly $32 billion in a single quarter. These are not R&D budgets. This is physical infrastructure: chips, power, cooling, fiber. A bet placed in concrete and silicon that the productivity gains are real and durable and coming soon.
The U.S. economy added 115,000 jobs. Nonfarm productivity grew at just 0.8% annualized in Q1, the slowest pace in four quarters. The companies spending the most on the revolution are posting some of the weakest productivity numbers in recent memory. Either the gains are coming later and the investment thesis is patient capital, or a meaningful chunk of that six hundred and sixty billion dollars is what John Kenneth Galbraith might have called the competitive exercise of spending in order not to be left behind.
The Fed, for its part, knows exactly how uncomfortable this all is. Jerome Powell's last FOMC meeting as Chair was in April. His term as Chair expires May 15th — five days from now. Kevin Warsh is expected to step into the role currently held by Stephen Miran on an interim basis. Four FOMC members dissented in April: three wanted to rip out the easing bias entirely, one wanted a cut. The committee is not unified. It is not even particularly coherent.
What Powell leaves behind is a central bank that held rates at 3.50–3.75% through an oil shock driven by a war in the Middle East, a labor market bleeding participation at the margins while producing just enough top-line numbers to forestall political pressure, and an inflation dynamic that has technically cooled from its peak but sits thirty basis points above the two percent target with energy prices providing no comfort on the forward trajectory. Whoever chairs the Fed next will inherit all of that and the question of whether the AI capex supercycle ever actually shows up in TFP data — or whether it was always, primarily, a financial markets story.
Markets have rallied for six consecutive weeks. The S&P closed Thursday at 7,337. The forward P/E for the index is 20.9x, above both the five-year and ten-year averages. That is not irrational in a world where the dominant cohort of publicly listed companies is producing genuine earnings growth. But it does mean that every upside surprise in the jobs data, every earnings beat from Alphabet or Amazon, every positive headline on the Iran framework deal, is already baked in at something close to perfection.
The market is pricing a clean resolution to the Middle East conflict, a soft landing for the labor market, and productivity gains from AI arriving just in time to justify the capex. It may get all three. But the same Friday that gave us a jobs report that beat by eighty-five thousand also gave us Cloudflare announcing it no longer needs eleven hundred of its employees because the technology it helped build made them unnecessary. The headline said resilient. The filing said "agentic AI-first." They were both describing the same economy.
Decide for yourself which one you're holding.
*For informational purposes only · Not investment advice
Upvoted! Thank you for supporting witness @jswit.