Latest from OpenAI, Anthropic, Google, Meta & more
1 stories
A startup emerging from stealth with $12 million in backing has identified a critical inefficiency plaguing the trillion-dollar AI infrastructure race: GPUs are burning far more power than necessary, wastefully spiking electricity costs for the companies racing to build the largest language models. Niv-AI's software promises to measure and manage those power surges in real time, potentially reshaping the economics of AI development at a moment when energy consumption has become the limiting factor in scaling up artificial intelligence. If the technology works as promised, it could hand a significant cost advantage to early adopters while intensifying pressure on competitors to rein in their sprawling, power-hungry data centers.