South Korea boosts its AI chip industry with $642M amid ChatGPT frenzy

The news touches on a couple of notable currents in the world of tech right now. For one, there is the current OpenAI/ChatGPT frenzy: recent advances in generative AI point to a future where AI could create a new wave of services and make existing processes more efficient, and so a lot of consumers and organizations — tech and non-tech — are considering how this could work for them. All that AI data crunching, however, needs immense processing power: cue an opportunity and demand for better chips and architectures to run those services more efficiently, and a subsequent race among companies and countries to be the ones to meet that demand.

Added to that is South Korea’s macroeconomic potential. There is a long-lingering question mark over what role China will play in the years ahead in the world of hardware and advanced technologies — something that has geopolitical as well as economic implications — and that’s spurring R&D and ambitions from a wider set of countries to complement and indeed compete better with China in the tech industry. Indeed, South Korea’s efforts are not alone: they come as Japan, India, the US and others explore how they too can do more in the advanced chip race. South Korea’s advantage is that it’s already a behemoth: it’s not just the home to several iconic mobile brands and services, but it’s also a major supplier of components for an even longer list of players.

Today’s announcement is part of the country’s bigger digital strategy plan. Unveiled in September 2022, the government’s digital strategy budget places bets on six major technologies — AI, AI semiconductors, 5G and 6G communication, quantum, metaverse and cybersecurity. AI is a big part of that vision: Seoul at the time set aside approximately $795.3 million (1.02 trillion won) for AI semiconductors and $235.3 million (301.8 billion won) for next-generation AI from 2022 to 2026.

(Note: investment numbers for the AI semiconductor industry between the September 2022 budget and today’s news vary not just because of the timescales, but because the funds cover various projects, and these have different timelines.)

As its first project, Seoul will hire AI chipmakers and cloud companies as early as April for two data centers, known as neural process unit (NPU) farms, which will only use domestic AI chips. The announcement indicates that the two data centers are projected to open as early as 2024.

Three AI chipmakers in South Korea — Rebellions, Sapeon Korea and FuriosaAI — will join the bid for the data centers by teaming up with one of the cloud firms such as Naver Cloud, KT Cloud, KaKao Cloud and NHN Cloud, sources familiar with the situation told TechCrunch.

Rebellions, Furiosa AI and Sapeon Korea spokespeople confirmed to TechCrunch that they plan to bid for the data centers without providing further details.

We don’t know which cloud companies will partner with which the AI chipmakers yet, but Rebellions has been backed by Korean telco KT and Temaske’s Pavilion Capital; Furiosa AI has received funding from Naver. Sapeon, a Korean telco giant SK Telecom subsidiary, which has raised funding from SK Hynix, SK Telecom and SK Square, already offers its first AI chip X220, manufactured by TSMC and launched in 2020, to NHN Cloud.

Rebellions and Furiosa AI have not yet moved into mass production. Both aim to commercialize their products next year, according to the industry sources.

Rebellions strategically announced earlier this week that it had launched an artificial intelligence (AI) chip called ATOM, manufactured by Samsung and used in data centers, chatbot AI and computer vision.

Furiosa AI is also set to unveil its second AI chip for Chatbot AI applications and plans to commercialize the second AI chip in the first quarter of next year.

Apart from the data center bid, the AI chip industry’s most notable strategic deal has been a partnership between the nation’s two largest tech firms, Samsung and Naver, which will work together to develop AI chips.

South Korea aims to develop up to 50 types of AI chips, focused on system semiconductors. It believes it has the potential to secure 20 percent of the global AI chip market by 2030. The global AI chip market is projected to reach $263.6 billion by 2031, 37.1 percent up from $11.2 billion in 2021, according to a report by Allied Research.