You're tuned in to Tech Beat, your daily read on the technology shaping how we live and work.
Google is taking an unusual step with its next AI infrastructure push. A chip codenamed Frozen v two would embed parts of the Gemini model's architecture directly into silicon, potentially delivering six to ten times the efficiency of current setups. Investors appear to have already taken notice, which tells you something about how seriously the industry is weighing hardware as the next frontier in the AI race.
That story connects to something else out of Google today. The company released three new Gemini models — three point six Flash, three point five Flash-Lite, and Flash Cyber — but the one developers were watching for, Gemini three point five Pro, was notably absent. No explanation was offered, and that silence is fueling real questions about where Google's flagship model roadmap actually stands right now.
Meanwhile, a more unsettling concern is drawing attention from the security community. Researchers at Zhejiang University have outlined a theoretical attack they're calling Bit two Watt, in which roughly one thousand hijacked GPUs could be used to destabilize a local power grid by exploiting the dramatic swings in energy demand that AI workloads create. The mitigations exist, but the vulnerability itself is a reminder that data centers are now physical infrastructure in ways we're only beginning to reckon with.
Stay curious, stay skeptical. Tech Beat out.
