A MESSAGE FROM BLOOM ENERGY

AI growth is outpacing power infrastructure, making energy strategy a board-level concern. The 2025 Mid-Year Power Report highlights grid delays, rising onsite generation, and the strategic importance of power access.

Read report

“We are seeing exponential growth in demand across our entire customer base. This financing demonstrates investors’ extraordinary confidence in our financial performance and the strength of their collaboration with us to continue fueling our unprecedented growth.” —Krishna Rao, Anthropic’s chief financial officer, discussing the company’s latest Series F funding round, a $13 billion cash injection led by Iconiq, Fidelity Management, and Lightspeed Venture Partners that values the startup at $183 billion.

Anthropic was last valued at $61.5 billion in March. The company’s run-rate revenue—an easily gameable calculation that takes a given month’s revenue and multiplies it by 12 to extrapolate out to annual revenue—has surpassed $5 billion, according to Anthropic. At the beginning of the year, that same number was just $1 billion. And that’s the sound of a bubble expanding.

And now for the main event…

Last month, I wrote about a team of researchers at a think tank called the Economic Innovation Group, who published a report debunking some of the hysteria about the feared A.I. jobpocalypse. Notably, the researchers found a paucity of evidence suggesting that this doomsday was around the corner. After breaking occupations into categories based on their exposure to A.I. obsolescence, the researchers discovered, ironically, that the “unemployment rate for the most A.I.-exposed workers” was actually rising less quickly than for “the least exposed workers.” Somewhat humorously, the report was titled AI and Jobs: The Final Word (Until the Next One)—a reference to the fact that we’re still in the early days of understanding how this technology will truly impact the workforce, despite vocal protestations that we’re headed down a path of no return.

For years, the industry narrative of untold reinvention and destruction has been narrated by the A.I. soothsayers and evangelists most invested in its success. In 2016, Geoffrey Hinton, the so-called godfather of A.I., famously predicted that radiologists would be out of a job within five to 10 years. “I think that if you work as a radiologist, you are like Wile E. Coyote in the cartoon,” he said. “You’re already over the edge of the cliff, but you haven’t yet looked down.”

Of course, this became a rampant talking point. Elon Musk said at a conference in 2024 that “in a benign scenario … probably none of us will have a job.” Dario Amodei, the C.E.O. of Anthropic, said in May that A.I. will probably erase half of all entry-level white collar jobs in a couple of years, leading to an unemployment rate as high as 20 percent. And OpenAI chief Sam Altman said recently that entire job categories will be “just like totally, totally gone” due to A.I.-induced replacement.

Now, a cohort of more sober and circumspect researchers is adding their flavor to the industry scholarship. While A.I. models have improved drastically, there remain extraordinary challenges—reliability, hallucination, bias, etcetera. Rather than a job wipeout, what we’ve seen is certain professions, like Hinton’s radiologists, adopting A.I.-powered tools to enhance what they do. At the end of the day, the work is undertaken by human beings, who are then held accountable to other humans. Businesses writ large, meanwhile, have struggled to adopt and deploy these systems at all.

To wit: Just a few short weeks after the E.I.G. paper, the next word has surfaced in the shape of another report, written by Stanford economists Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen. Canaries in the Coal Mine?: Six Facts About the Recent Employment Effects of Artificial Intelligence borrowed some methods from the folks at E.I.G., but the Stanford economists came to a broadly similar conclusion (with a few major caveats) despite looking at a different dataset. Using data from ADP, the largest payroll provider in the U.S., they identified that employment has declined 20 percent since 2022 for early-career employees (aged 22-25) in exposed fields such as software development and customer service. But Chandar noted that it’s not clear that “these findings are fully driven by A.I. Many other things changed in the U.S. economy at the same time.” It is entirely possible that a substantial portion of this decline was the result of rampant overhiring by Big Tech during the pandemic. Meanwhile, and critically, employment remained stable among older, middle-aged groups.

A MESSAGE FROM BLOOM ENERGY

As AI adoption accelerates, power has become the defining constraint—and opportunity—for data center growth. Our latest 2025 Mid-Year Power Report reveals a dramatic shift in how industry leaders are planning for the future.

Read the report.

The Stanford team concluded that seniority, training, and experience were key factors in job retention. But, at a macro level, their analysis aligned with the E.I.G. finding that the jobpocalypse narrative has been overhyped. At least, we have no evidence of widespread replacement so far. Joshua Gans, a professor at the University of Toronto’s Rotman School of Management and the chief economist of Creative Destruction Lab, wrote that these results supported the hypothesis that A.I. will act as a complement to, rather than a substitute for, human labor.

Nathan Goldschlag, the director of research at E.I.G., said that the report largely matched his own findings on the A.I. doomsday scenario in the labor markets. “It is interesting how quickly the effects fade with age, (with) results so concentrated among very young workers,” he tweeted. “It is also interesting that they find no earnings effects. If demand for young A.I. (exposed) workers is falling, then wages should be falling too. Perhaps this is because the effects are just kicking in, but it raises more questions to be sure.”

In the long arc of history, the idea that a technological innovation will simply replace labor has never actually borne out. Instead, the labor markets have just continued to evolve: In the late 1800s, nearly half of Americans worked on farms; by 1970, the share of farmworkers in the U.S. had fallen to around 3 percent. So while job displacement due to technological adoption undeniably occurs, new jobs have always been created on the other end. Software engineering, for instance, a field with tens of millions of practitioners, didn’t exist until the 1960s. “As long as there are human beings, there will be a market, and there will be a labor market,” Dr. Fabian Braesemann, a research lecturer in A.I. and work at the Oxford Internet Institute, told me. “So I do not at all buy into this frenzy around mass unemployment.”

Braesemann also noted that the conclusion of these recent papers agrees with his own research. Historically, he said, technological progress changes the requirements of participation in the labor market. If that transition moves more slowly than demographic change, the process usually goes unnoticed. If the process occurs more swiftly, as with A.I., “then we might see a phase where people struggle to reskill fast enough—or where the market struggles to offer enough new jobs in other industries or new occupations—so they might become unemployed.”

Braesemann expects there to be some weirdness in the data in the early days of widespread adoption, as lower-level work begins to evolve into more of a directory role, with employees overseeing A.I. agents. Meanwhile, entry-level workers might struggle to move to other parts of the labor market fast enough. (This might explain some of the findings in the Canaries study.) Braesemann is also certain that, as with the rise of the internet, smartphones, and personal computers, we will see the birth of entirely new categories of jobs—a fact that few dispute. It’s just not clear what they’ll be or when they’ll appear.

For now, though, as the MIT economist Simon Johnson told me in July, A.I. is currently reducing labor costs more than it is adding those all-important new tasks. “I think the key thing is there are two processes: One is automation replacing people, and the other is new task creation. You need the new task creation to be strong enough to deal with what automation is doing to the number of people being displaced,” he said. “And the problem is that A.I. is coming very fast. This is the downside of a massive investment push: You get more new stuff sooner—which we usually like—but if that means it’s harder to adjust in the labor market for many millions of people, that’s going to be a problem.”

In my own conversations with companies that have been eager to adopt the technology, I’ve come across much the same scenario—a largely successful push to reduce labor costs by enabling small teams to do more with fewer resources. The reality is that the scope of the impact this technology is having on the labor market will likely be one of those hotly debated open questions for a long time. But as Noah Smith points out, maybe that’s the best-case scenario—as long as it remains a debate, almost by definition, nothing terrible will have happened.

That’s all for today. I’ll see you Thursday.

Ian

Need help? Review our FAQ page or contact us for assistance. For brand partnerships, email ads@puck.news.

You received this email because you signed up to receive emails from Puck, or as part of your Puck account associated with {{customer.email}}. To stop receiving this newsletter and/or manage all your email preferences, click here.

Puck is published by Heat Media LLC. 107 Greenwich St, New York, NY 10006