Key facts
- Japanese AI companies are seeking alternatives to Nvidia's expensive GPU servers.
- The focus is on developing less costly server options.
- These alternatives include systems utilizing Neural Processing Units (NPUs) and cache servers.
Many artificial intelligence-related companies in Japan are facing high costs associated with servers equipped with Nvidia's graphics processing units (GPUs). In response, efforts are being made to develop and offer more affordable alternatives. These emerging solutions often incorporate Neural Processing Units (NPUs) and cache servers, which are seen as more cost-effective options compared to solely relying on increased GPU capacity. One such AI data center server is equipped with NPU semiconductors developed by Rebellion.