Semiconductor News & Analysis Feed

NVIDIA has achieved a groundbreaking 5x reduction in token cost for DeepSeek v4 AI models just one month after its launch, through full-stack inference software optimizations on its Blackwell GPU platform. This development underscores the importance of 'cost per token' as a key metric for AI total c

NVIDIA has achieved a groundbreaking 5x reduction in token cost for DeepSeek v4 AI models just one month after its launch, thanks to full-stack inference software optimizations on its Blackwell GPU platform. This development underscores the importance of 'cost per token' as a key metric for AI total

As organizations transition from AI pilot projects to full-scale AI factories, infrastructure decisions have shifted from peak chip specifications to cost per token—measuring how many useful tokens can be delivered per dollar, watt, and within required latency targets. NVIDIA’s inference software st

As organizations transition from AI pilots to production AI factories, infrastructure decisions have shifted from peak chip specifications to cost per token—how many useful tokens can be delivered per dollar, watt, and within required latency targets. NVIDIA’s full-stack inference software, co-desig