Intel has officially launched the K-AI Full-Stack Alliance in Seoul, partnering with 23 South Korean companies. The initiative aims to develop a comprehensive AI ecosystem, covering hardware, software, and industry-specific solutions to support sovereign and open AI infrastructure within the local market.
The alliance includes a diverse range of participants, such as server manufacturer KTNF, AI chip startups FuriosaAI and DEEPX, large language model developer Upstage, and cloud service provider MegazoneCloud. These members will collaborate across the entire AI technology stack.
During the event, Intel showcased its next-generation data center inference accelerator, the D270 GPU from the D series. The chip delivers 26,200 TFLOPS of FP4 performance with a power consumption of 350 watts.
Intel claims the D270 outperforms the Nvidia RTX 6000 Pro in several benchmarks. According to company data, the new accelerator reduces latency for the Gemma model by 30 percent, improves AI agent performance by 2.2 times, and boosts financial simulation speeds by up to 5.9 times.
