Businesses are racing to replace people with AI, and they're not waiting to first find out whether AI is up to the job.

Why it matters: CEOs are gambling that Silicon Valley will improve AI fast enough that they can rush cutbacks today without getting caught shorthanded tomorrow.

• While AI tools can often enhance office workers' productivity, in most cases they aren't yet adept, independent or reliable enough to take their places.
• But AI leaders say that's imminent — any year now! — and CEOs are listening.

State of play: If these execs win their bets, they'll have taken the lead in the great AI race they believe they're competing in.

• But if they lose and have to backtrack, as some companies already are doing, they'll have needlessly kicked off a massive voluntary disruption that they will regret almost as much as their discarded employees do.

Driving the news: AI could wipe out half of all entry-level white-collar jobs — and spike unemployment to 10-20% in the next one to five years, Anthropic CEO Dario Amodei told Axios' Jim VandeHei and Mike Allen this week.

• Amodei argues the industry needs to stop "sugarcoating" this white-collar bloodbath — a mass elimination of jobs across technology, finance, law, consulting and other white-collar professions, especially entry-level gigs.

Yes, but: Many economists anticipate a less extreme impact. They point to previous waves of digital change, like the advent of the PC and the internet, that arrived with predictions of job-market devastation that didn't pan out.

• Other critics argue that AI leaders like Amodei have a vested interest in playing up the speed and size of AI's impact to justify raising the enormous sums the technology requires to build.

By the numbers: Unemployment among recent college grads is growing faster than among other groups and presents one early warning sign of AI's toll on the white-collar job market, according to a new study by Oxford Economics.

• Looking at a three-month moving average, the jobless rate for those ages 22 to 27 with a bachelor's degree was close to 6% in April, compared with just above 4% for the overall workforce.

Between the lines: Several companies that made early high-profile announcements that they would replace legions of human workers with AI have already had to change course.

• Klarna, the buy now-pay later company, set out in 2023 to be OpenAI's "favorite guinea pig" for testing how far a firm could go at using AI to replace human workers — but earlier this month it backed off a bit, hiring additional support workers because customers want the option of talking to a real person.
• IBM predicted in 2023 that it would soon be able to replace around 8,000 jobs with AI. Two years later, its CEO told the Wall Street Journal that so far the company has replaced a couple of hundred HR employees with AI — but increased hiring of software developers and salespeople.

Zoom out: Every modern era of technological transformation has disrupted the labor market, from the Industrial Revolution of the early 19th century to the assembly-line automation of the early 20th and the container ship-driven globalization of the millennium.

• The transitions have often been rough, but economies emerged bigger and with more jobs, not less.
• Some AI experts fear the change could be so much faster with AI that there will be no time to adapt. Others view AI as a fundamentally different kind of tech that will force society to invent new approaches to jobs and salaries, like the notion of a universal basic income.

Our thought bubble: Predicting employment levels has always been tough because there are so many complex variables to consider.

• Even if Amodei is right and AI cuts a devastating swath among office workers, there are other demographic forces at work that could make it harder for businesses to find the human workers they still need.
• For instance: The largest generation in history is retiring as boomers age out of the workforce. The Trump administration is working overtime to limit immigration. Other black-swan crises will erupt that could boost or limit unemployment.

What we're watching: The sociopolitical skews of AI's workforce impact are volatile and hold a great potential for splitting coalitions and dividing allies.

• The populist wing of Trump's MAGA movement is likely to resist AI-driven change even as the president's tech-insurgent allies push for more investment and weaker regulation.
• More broadly, Americans overall say that, unlike impatient CEOs and China-fearing office-holders, they want to see AI introduced with more care and less haste, per the 2025 Axios Harris 100 poll.
• If the speedy road gets bumpy, the public might have another "throw the bums out" fit.

The New York Times has inked a multiyear AI licensing deal with Amazon to bring its journalism to a number of Amazon-owned products and experiences, the company said yesterday.

Why it matters: It's the Times' first-ever deal with an AI company for its content.

• The newspaper company is currently engaged in a copyright infringement lawsuit against OpenAI and Microsoft.

Zoom in: The licensing deal will give Amazon access to editorial content from the New York Times, its cooking app, and its sports site, The Athletic, for AI-related uses, the company said.

• The content will be used to fuel real-time answers to user queries via summaries and short excerpts of Times content that appear across Amazon products and services, such as the Alexa voice assistant.
• The deal also gives Amazon access to the Times' content for training its large language models.

Between the lines: When asked why the deal doesn't include content from the Times' consumer recommendation site, Wirecutter, a spokesperson told Axios that "Amazon and Wirecutter have a longstanding relationship."

The big picture: More news companies are finding ways to strike deals with some AI firms while taking legal action against others.

• News Corp. struck a lucrative five-year deal with OpenAI last year but later sued AI search engine Perplexity.
• The Atlantic, Condé Nast and Politico (through its parent Axel Springer) have all announced commercial deals with OpenAI, but are suing Cohere.

Disclosure: Axios and OpenAI have a licensing and technology agreement that allows OpenAI to access part of Axios' story archives while helping fund the launch of Axios into four local cities and providing some AI tools. Axios has editorial independence.

• Scoop: The Trump administration is planning to change the name of the AI Safety Institute to the Center for AI Safety and Leadership. (Axios Pro)
• Exclusive: Microsoft is bolstering its policy and legal leadership — promoting former Trump administration official C.J. Mahoney to general counsel, while also tapping former Biden Administration deputy attorney general Lisa Monaco to lead global policy efforts. (Axios)
• Former Facebooker Palmer Luckey's Anduril is teaming up with Meta to build helmets, glasses and other VR and AR wearables designed for soldiers operating AI-powered weapon systems. (Wall Street Journal)
• Grammarly has raised $1 billion in fresh funding as it aims to continue its expansion into a broader AI platform. (Reuters)
• China's DeepSeek has made an improved model available to its API customers and freely available via an MIT license. (VentureBeat)

It's not as bad as "snakes on a plane," but pigeons on a plane doesn't look like much fun either.