AI slowdown warnings hit data centers, markets and hiring, even as demand persists

Digital Realty CEO Andrew Power said in an interview with CNBC’s Property Play that pledges for a slowdown by Anthropic, OpenAI and xAI do not mean “pencils down” for AI or the real estate that supports it. “There’s tremendous digital transformation happening that is not connected to AI,” Power said. “There is tremendous cloud computing growth.” He said hyperscalers have had to choose between growing commercial cloud businesses or allocating capacity to AI labs, and not all markets will be affected equally. Digital Realty operates in Northern Virginia, Dallas, Chicago, Singapore, Tokyo, Frankfurt and Amsterdam, where demand has outpaced supply for several years. The company has $20 billion under construction, up from $10 billion at the end of 2023, and has evolved its funding model to raise private capital and form joint ventures, Power said. “I’m not suggesting today is an end-of-the-world storm or anything like that,” he said.

Analysts said a slowdown would not directly remove the physical needs of AI. Andrew Batson, global head of data-center research and strategy at JLL, said the real growth in data centers over the next few years is in inference—businesses and citizens adopting the tools into daily workflows. Only one in four Americans uses AI daily, he said, leaving significant runway for adoption and data-center demand even if model releases slow. Batson pointed to institutional money from Blackstone, BlackRock and KKR, which he said “have high conviction in this space.” JLL estimates the real-estate portion of data-center investment could account for $3 trillion in the next five years, while McKinsey has said AI could represent about 70% of global data-center capacity demand by 2030 and total capital outlay could reach nearly $7 trillion.

In CNBC’s Market Memo, Mike Santoli wrote that the “youthful phase of AI”—when promise was unlimited and winners were easy to spot—is over. Semiconductor shares were already 20% below their June highs before a 4% slide Monday, and the S&P 500 tech sector’s forward price-to-earnings multiple has contracted from 29 to 21 over the past year, Santoli wrote. He said the AI narrative has become more fixated on risks than opportunities, and that data centers now have a lower approval rating than Congress. Still, he noted that the purge of high-momentum AI hardware and industrial stocks has reset tactical sentiment, and that big platforms doing most AI capital spending—Microsoft, Alphabet and Meta—could take a beat before accelerating. He also compared macro conditions to 1999, when the Fed began raising rates as 10-year Treasury yields hit 6% and a tech-capex boom raged, after cutting rates three times late the prior year.

MIT Technology Review reported that the AI industry has taken a “doomer turn,” with Dario Amodei, Sam Altman, Elon Musk and Demis Hassabis suddenly agreeing that the latest generation of large language models is not safe and that the industry needs to figure out what to do. The publication noted that cynicism is easy: with trillion-dollar IPOs in sight, OpenAI and Anthropic need to reassure investors while hinting at the power of the models they have created. The shift has not been universal. MIT Technology Review also pointed to reports that Trump has called AI safety fears a “hoax” and rejected more safeguards, saying stronger guardrails could undermine America’s AI advantage, while Anthropic’s co-founder has said AI kill switches may need to be mandatory and Bill Gates has said the world has passed AI’s risk thresholds. In a separate experiment, Google DeepMind found that AI agents asked to solve math problems split into rival factions, and when some cheated, others tried to stop them—a first-of-its-kind whistleblowing behavior with implications for alignment researchers.

CNBC reported on resistance to AI interviews in hiring. Roughly two-thirds of 1,200 U.S. job seekers said they have had an AI interview, according to an April Greenhouse survey, and nearly four in 10 have withdrawn from a hiring process over one. In a separate Greenhouse report last year, 70% of hiring managers said AI helps them make faster and better hiring decisions with fewer recruiter resources, and one in two recruiters said AI improved hiring overall, but only 8% of candidates believe AI makes hiring fairer. Art Hebbeler, a 65-year-old IT program director, said an AI bot interviewed him after he expected a recruiter; he hung up early. Aaron Holmes, 41, said an AI interviewer had no natural cadence, left his questions unanswered and left “a really bad taste.” Tammy Wright, an instructional designer with nystagmus, said AI software repeatedly interrupted her to tell her to look straight ahead, calling it “demoralizing.” The American Civil Liberties Union says predictive hiring tools that analyze facial, audio or physical interactions risk automatically rejecting or scoring lower candidates based on disabilities, race and other protected characteristics. HireVue said in 2021 that it stopped visual analysis of candidates, saying it has far less correlation to job performance than other elements of its algorithmic assessment.