What AI leaders at Davos 2026 said about layoffs and growth

Davos no longer felt like a discussion of whether AI is real or not. Instead, it felt like a discussion of how quickly AI will impact daily work, who benefits first, and who gets caught standing still.

Over five days in the Swiss Alps, the topic people kept coming back to, in private meetings as well as on stage, was artificial intelligence. The tone oscillated between optimism, worry, and a kind of acceptance that the buildout is happening anyway.

The mood of the week on AI was jobs first, disruption second

The line repeated most often was simple: “Jobs, jobs, jobs.” Nvidia CEO Jensen Huang used it to reassure everyone that AI is already generating employment in fields like energy, semiconductors, and infrastructure, even though concern about job loss is growing.

But Davos also offered a blunt counterpoint. In a different session, Anthropic CEO Dario Amodei forecasted that 50% of white-collar jobs might disappear by 2031. This reframed the “jobs” discussion as a near-term labor shock, rather than just a long-term transition.

The pushback came just as strongly from policymakers. IMF Managing Director Kristalina Georgieva issued a warning about an AI “tsunami” for the labor market. She pointed to an IMF study estimating that 60% of jobs in advanced economies and 40% globally could be affected. 

She noted that changes are already underway, with jobs in advanced economies already being “enhanced” by technology. The more ominous threat, however, came when she spoke of the youth: “entry-level jobs” , the typical launching point for a career, are now the easiest to erase.

The subtext of all these views was the same. The pitch for AI is about productivity, but the battle is over who absorbs the pain and who captures the gain.

AGI Timelines Transformed into an On-Stage Argument

If there was one Davos AI storyline built to generate headlines, it was the split over how close existing models are to human-level intelligence.

Google DeepMind CEO Demis Hassabis said current machines are “nowhere near” AGI. He suggested the field still has “missing ingredients,” putting AGI roughly five to ten years out.

Yann LeCun, Former Chief AI Scientist of Meta, took it a step further, arguing that the large language model approach won’t get us there at all. He claimed the industry needs systems that can build stronger “world models” and understand cause and effect.

The key takeaway wasn’t that Davos produced a single timeline. It was that the people with the strongest credentials and the biggest budgets are still debating whether the current path is a straight line to human-level intelligence, or a dead end that needs a new approach.

Chips, Power, and the physical wall behind the buildout

Davos also made another quieter truth harder to ignore: the AI boom is no longer just a model story, it is a capital-and-infrastructure story.

TSMC CEO C.C. Wei said they went beyond normal checks to ensure AI demand was solid enough to warrant the massive spending required to expand capacity. He framed this as a bet that only makes sense if AI adoption translates into real financial returns.

But the larger limit is even simpler than that. AI software can grow rapidly. Physical power cannot. As models become larger and deployments get wider, the bottleneck changes from “who has the best code” to “who can get the chips, the data centers, and the electricity to run them.”

In other words, the limit on AI is no longer imagination; it’s physics, permitting, grids, and capex. This is why energy companies and infrastructure players are increasingly considered gatekeepers of who gets to compete.

AI started inching towards hardware and always-on assistants

One of the most tangible indicators of the week came from the hardware discussion at OpenAI.

At Axios House Davos, OpenAI Chief of Global Affairs, Chris Lehane said the company is “on track” and provided a public window for when its first device could surface: the second half of 2026. However, he declined to confirm what the device actually is.

That timeline marker did two things. First, it reinforced that OpenAI is serious about moving past the app. Second, it highlighted the issue plaguing past AI devices: if the device is slower than a phone, or can’t easily integrate into a daily routine, it becomes a novelty.

OpenAI’s advantage, if they can pull it off, is control, shipping their own assistant on their own hardware, rather than depending on another company's operating system.

What Davos Really Settled

Davos did not resolve the question of timelines or job impact. But it did settle one thing: the discussion is no longer about feature upgrades. It is about an economic force with a supply chain, a power bill, and real labor consequences.

The "wait and see" period is ending because the stack is forming before everyone’s eyes: models on top, chips and infrastructure below, and hardware forms starting to emerge to make AI feel omnipresent.

Whether you believe Huang’s optimism or Amodei’s warning, the consequence is clear: in 2026, the winners won’t just be the ones with the best demos, but the ones who control the path from model to chip to power to product.