AI hardware panel: LLMs are just table stakes; retention and iteration decide survival
At Huxiu's 13th AI100 closed-door session, speakers from Rokid, Xiaodu, Tingzhihui Technology and a Singapore incubator discussed what it takes for AI hardware companies to survive now that models and agents are widely available. Their shared view: wiring in AI only gets hardware to a passing grade — the real gaps are understanding of physical scenarios, accumulating unique data, AIOS-centered hardware-software integration and actually completing tasks. The panel also cited Meta's Muse Charm AI keychain as evidence that form factors are fragmenting into niches, and suggested metrics such as 30-day retention to separate novelty purchases from real demand.
Offers practitioners a frontline framework for judging form-factor choices, data assets and retention metrics in AI hardware.