Agentic AI Remains a Core Talking Point
“Agentic AI,” a term describing software that completes tasks autonomously for humans, retained its status as one of the forum’s most frequently cited buzzwords. Executives agreed that practical uses already exist, though the sophistication of those agents varies considerably by industry.
Uljan Sharka, chief executive of the start-up Domyn, told participants current agents still require a significant degree of human oversight and are not ready to replace employees outright. By contrast, Fabricio Bloisi, chief executive of Prosus, offered the most optimistic forecast. He revealed that Prosus already operates roughly 30,000 software agents and predicted the emergence of companies run largely by such systems within five years, underscoring his belief that the concept represents more than temporary enthusiasm.
Geopolitical Tensions Shape AI Timelines
Beyond technical hurdles, many leaders highlighted geopolitical unpredictability as a factor that could either accelerate or delay AI adoption. Sharma noted that where political flashpoints arise will matter as much as the technology itself. Export controls, data-sovereignty rules and supply-chain uncertainties all influence how quickly cutting-edge tools spread across borders.
The rivalry between the United States and China surfaced repeatedly. Demis Hassabis, chief executive of Google DeepMind, estimated that Chinese AI models trail their Western counterparts by only a few months. His comment reinforced the view among delegates that any gap in capability is narrowing, heightening competitive pressure for companies and governments on both sides of the Pacific.
Physical AI Emerges as the Next Growth Area
While software agents monopolized attention last year, “physical AI” gained momentum in Davos. The label refers to embodied applications such as industrial robots, autonomous vehicles and other machines that combine advanced perception with decision-making algorithms. Sharma described the segment as the “next wave,” projecting it could be five to six times larger than the market for agentic AI within the same time frame.
Sassine Ghazi, chief executive of chip-design toolmaker Synopsys, said he originally believed physical AI would become mainstream in the late 2020s but now expects widespread adoption sooner. Jensen Huang, chief executive of Nvidia, called AI-powered robotics a “once-in-a-generation” chance for Europe, citing the continent’s deep industrial manufacturing base.
Imagem: Internet
The technology’s growing visibility was evident even in informal settings. During one evening reception, a humanoid robot joined guests at the dinner table, illustrating how quickly experimental prototypes are reaching public venues.
Energy Supply and Infrastructure Questions Persist
Several chief executives also warned that energy infrastructure may struggle to keep pace with AI’s computing demands. Training large models and operating data centers at scale requires substantial electricity, much of it still generated by fossil fuels. Without new sources of low-carbon power, they cautioned, governments could face difficult trade-offs between climate commitments and digital expansion. Recent studies by the International Energy Agency underscore the challenge, estimating data-center electricity use could double this decade if efficiency gains stall.
Investor Focus Turns to Tangible Returns
From an investment standpoint, the overarching message from Davos was that markets will reward proof of value rather than potential alone. Last year’s enthusiasm lifted share prices of AI-linked firms even before revenues materialized. Executives now expect analysts to scrutinize metrics such as cost reduction, margin improvement and new revenue streams when evaluating future performance.
That shift mirrors comments from cloud providers who say corporate customers are aligning AI spending with clear business objectives. Instead of purchasing broad access to large language models, companies increasingly ask vendors to tailor tools for customer service, software development or supply-chain optimization—areas where returns can be tracked.
Outlook for 2026 and Beyond
Davos attendees agreed that AI will remain the central narrative in technology over the next 12 months, but the conversation has matured. The debate is moving from “can it be done” to “does it pay off,” with corporate adoption patterns, geopolitical currents and infrastructure capacity all shaping the pace of progress.
Whether the most ambitious predictions—such as companies managed largely by autonomous agents—materialize within five years will depend on sustained advances in model reliability, regulatory clarity and energy availability. For now, boards and investors appear poised to evaluate AI projects with the same rigor applied to any capital expenditure, marking a new phase in the technology’s evolution.
Crédito da imagem: Getty Images
