What Google's TurboQuant can and can't do for AI's spiraling cost

ZDNET reports on Google’s TurboQuant, a real-time quantization approach that compresses the key-value cache used by large language models to reduce memory usage (claims include up to 6x reduction and successful tests on Llama, Gemma, and Mistral models) without degrading accuracy; the piece frames TurboQuant as a potential cost- and memory-saving advance for AI inference and local deployments but not as a security-related incident.