WEF Davos 2026 debates AI bubble risks and superintelligence race

Davos 2026 crystallized a tension that’s been building for two years: frontier AI is both the hottest growth story in tech and a potential macro‑risk if the economics don’t hold. Satya Nadella warned that hyperscalers could "quickly lose the social permission" to burn huge amounts of electricity on models that don’t deliver clear health, education or productivity gains.​([computerworld.com](https://www.computerworld.com/article/4121561/ai-needs-a-course-correction-say-world-economic-forum-speakers.html)) Elon Musk painted a near‑term future of superhuman AI and ubiquitous robots, but in the same breath highlighted energy as the binding constraint. AI‑first ministers like India’s Ashwini Vaishnaw pushed back against mega‑models, arguing smaller systems can do “95% of the work” with far less power.

For the race to AGI, this is a subtle but important pivot: the debate is shifting from "can we build it" to "can we afford it and justify it." If capital markets start treating AI data‑center buildout like a bubble, funding for multi‑trillion‑parameter experiments could tighten, pushing labs toward efficiency research, smarter architectures and agentic systems rather than brute‑force scaling. Conversely, if Nvidia and the cloud providers keep showing astronomical profits, the Davos warnings may simply mark the wall‑of‑worry phase of a long AI capex cycle.