Fired to pay for the AI, then rehired to fix it

Klarna's chief executive, Sebastian Siemiatkowski, found a large word for a small discovery.

We just had an epiphany: in a world of AI nothing will be as valuable as humans. — Sebastian Siemiatkowski

It was, in his telling, an epiphany. The admission that follows is the operative part: an efficiency drive had, by his own account, degraded what the company sold.

As cost unfortunately seems to have been a too predominant evaluation factor when organizing this, what you end up having is lower quality. — Sebastian Siemiatkowski

What the Klarna chief executive framed as revelation was something he had already said in plain terms a year earlier, as a warning, and then shelved.

If you want to stay customer-obsessed, you can't rely [entirely] on AI, — Sebastian Siemiatkowski

Klarna has since resumed recruiting human customer-service representatives ^[1]. The loop is not Klarna's alone. Ford rehired or promoted 350 veteran engineers in June after its automated quality systems came up short ^[1]; its chief operating officer, Kumar Galhotra, described the mechanism from the factory floor.

We had been relying more and more on automated quality systems and not getting the desired results. — Kumar Galhotra

Commonwealth Bank of Australia had to unwind its own version last November, after conceding a round of dismissals had been built on an error.

did not adequately consider all relevant business considerations and this error meant the roles were not redundant — Commonwealth Bank

Meta, which had cut roles to fund its AI buildout, is now rehiring middle managers because its people were spending their days correcting the machines' mistakes ^[2]. Set the scope before the turn: a Stanford study this summer found AI's aggregate effect on employment has been small — no white-collar apocalypse — with the sharpest wound a decline in entry-level hiring since 2022 ^[3]. This is a misallocation story, not an extinction event. The first half of the trade was the layoffs themselves. They were not distress but reallocation: more than 142,000 tech jobs cut by June to fund an estimated $700 billion to $725 billion of AI infrastructure ^[4]. Meta eliminated 8,000 roles in a quarter it posted $26.8 billion in net income; Block cut 40 percent of its staff to refocus on AI ^[4]. The spending did produce explosive earnings — on the sellers' ledgers. Nvidia booked $81.6 billion in quarterly revenue, 92 percent of it from data centers, while Micron's revenue grew 346 percent and its stock 680 percent in a year ^[5]. Nvidia's vice president of applied deep learning, Bryan Catanzaro, put the arithmetic plainly.

For my team, the cost of compute is far beyond the costs of the employees. — Bryan Catanzaro

Uber's chief technology officer said his 2026 coding budget was already gone by April.

I’m back to the drawing board because the budget I thought I would need is blown away already. — Praveen Neppalli Naga

That is the arithmetic that made payroll the controllable line item and compute the one that is not. The repricing is the newest part of the story. Goldman Sachs found that half of S&P 500 companies mentioned AI's margin potential on second-quarter calls, yet only 2 percent quantified the earnings impact and 11 percent named a specific use case ^[6].

AI's next wave will likely reward companies that use the technology to improve decisions, lower costs and strengthen margins — not just those building AI infrastructure. — Ellen Zentner

The prices are following. Alphabet's credit-default swaps — the cost of insuring against a default on its debt — jumped as much as 5.3 basis points to 84 after its earnings landed flat ^[7]. Microsoft borrowed $25 billion to fund AI ^[8], though BlackRock's latest estimate — that the giant cloud providers' compute payback has fallen below three years, down from a four-year estimate months earlier — is the standing counter-case that the debt is buying equipment that pays ^[9]. Klarna's own shares are down 68 percent over the past year despite 27 percent revenue growth and a return to profit; the story there is volatile profitability and weakness in its German market, not a collapse in demand ^[10]. Of the 500 firms in the index, just 25 could point to a quantified margin gain from AI, and those are the ones the market is paying for ^[11]. The investor Steve Eisman keeps a colder ledger.

I've gotten much more conservative in my own portfolio. — Steve Eisman

Even this week's capital-spending "misses" read as supply choke rather than retreat: Microsoft and Alphabet both came in below expectations and blamed GPU supplier delays pushing usable capacity into next year ^[12].

Demand is not the issue, but the supply chain is. — Jefferies Group

The buildout is not stopping: nearly 70 percent of CEOs still planned to increase AI spending this year ^[13]. What has changed is what investors will pay to hear: demonstrated margin, not construction. Most of the industry never ran the control group, but one company did. IKEA deployed the same technology the opposite way: its chatbot took nearly half of service volume and saved more than €13 million, while 8,500 employees were retrained into interior-design services that generated more than €1.3 billion in revenue ^[14]. Two numbers, one company: thirteen million euros saved by the machine, one point three billion earned by the people it did not replace.