Everyone's watching the model wars. The bigger story is silicon. Reuters confirmed DeepSeek is building its own inference chip — hardware tuned to run models, not train them. The news alone nudged Nvidia's stock down. The 'so what': inference is now the faster-growing half of AI compute. If big labs make their own chips, Nvidia's moat narrows fast. More at www.contentbuffer.com Will every big lab design its own chip? #AI #Nvidia #DeepSeek #AIHardware #Semiconductors