Positron AI has secured 230M USD in Series B funding at a 1B USD valuation to scale its energy-effic...
The AMW Read
The article introduces a new hardware player focused on memory-centric architecture to solve the inference bottleneck, updating the silicon/infrastructure map.
NoveltySignificance
AI Infra · Player MapSilicon Substrate
Positron AI has secured 230M USD in Series B funding at a 1B USD valuation to scale its energy-efficient inference hardware. The company is tackling the memory bottleneck with its Asimov silicon, which packs 2304 GB of RAM per device compared to 384 GB in Nvidia upcoming Rubin GPU. By delivering 5x more tokens per watt, Positron is moving the industry from brute-force compute toward memory-centric architectures. This transition is essential for cost-effectively deploying multi-trillion parameter models at scale. 🚀
