AI NewsWaveOct. 29 AI Briefing: From FuriosaAI's 820 Billion Won Raise to Gemini 4 Carbon
Key points
- Big news from Korean AI chips and memory
- Google's new model and the Nvidia acquisition probe
- Government physical AI support and corporate AI service launches
FuriosaAI announced on the 29th that it has completed a Series D round of 820 billion won. AI Times, TheElec and others reported a post-investment valuation of about 3.8 trillion won, and Digital Daily reported that the company plans to use the funds for commercialization of its second-generation AI accelerator RNGD and for next-generation product development.
Samsung Electronics said on its third-quarter earnings call on the 29th that HBM4 sales more than tripled from the previous quarter. According to Chosun Biz, HBM4 accounts for about 60% of total HBM sales, and iNews24 reported that the company has sold all of next year's HBM volume at its target price. TechM reported the outlook that the memory shortage will last through 2028.
Google unveiled the AI model Gemini 4 Carbon, with stronger coding and cybersecurity capabilities, on the 28th (local time). Chosun Biz reported that it came two weeks after the introduction of Gemini 4 Argon, and Digital Today, citing The Verge, reported that it ranked first on the Artificial Analysis Intelligence Index v5.0.
The U.S. Department of Justice is investigating Nvidia's acquisition of Hugging Face. Newsis, citing CNBC, reported that the department is reportedly examining effects such as the impact of the acquisition on AI competitors. The deal is worth $12.93 billion.
At a meeting on physical AI policy direction on the 29th, the Ministry of Science and ICT said it will build a training center next year to produce and use training data for physical AI and will carry out a field pilot project with Korea Post, AI Times reported.
SK AX launched G-Fit Studio, a hiring solution that uses AI simulation to analyze applicants' problem-solving, communication and AI-use skills. TechM reported that it can interview 300 people per hour, and AI Newspaper (인공지능신문) reported that it evaluates about 1,800 people in six hours.
Kakao announced on the 29th that it has released two Kanana-2 embedding models as open source on Hugging Face, as foundation technology for personalized AI agents. The models have 3 billion and 1.3 billion parameters.
Miricanvas unveiled an AI connector linked to ChatGPT at OpenAI DevDay Exchange Seoul on the 22nd. According to VentureSquare, it creates draft designs such as presentations, card news and posters based on ChatGPT conversations, and a full launch is targeted within the year.