Korea’s Trillion-Dollar AI Semiconductor Play: Outmaneuvering Nvidia
The artificial intelligence (AI) semiconductor market is experiencing exponential growth, solidifying its position as a critical battleground for global technological supremacy. Projections indicate the data center AI chip market alone will expand from $124 billion in 2024 to a staggering $860 billion (approximately ₩1200 trillion) by 2030, representing an impressive 38% average annual growth rate. To seize this immense opportunity, South Korea has committed to a massive investment of nearly $1.2 trillion (around ₩1350 trillion) over several years, establishing new semiconductor hubs and AI data centers.
Nvidia remains the undisputed leader in the AI accelerator chip market. Despite its market share in AI accelerators, which peaked at 87% in 2024, being projected to decline slightly to 75% by 2026, Nvidia continues to hold a dominant position in this rapidly expanding sector. Its Graphics Processing Units (GPUs) are indispensable for AI model training, and the company’s robust ecosystem underpins its market control. However, as the AI industry’s focus shifts from training to inference, Neural Processing Units (NPUs) are gaining prominence due to their superior power efficiency and cost-effectiveness.
NPU Development: South Korea’s Strategic Edge
South Korea is actively pursuing a ‘K-NPU industrialization strategy’ with the ambitious goal of becoming a top-three AI powerhouse. Domestic companies are strategically concentrating on developing low-power, high-efficiency NPUs optimized for inference and edge AI, rather than directly confronting Nvidia’s GPU-centric ecosystem. Rebellions and FuriosaAI stand out as key players driving Korea’s NPU ecosystem. Rebellions, widely regarded as South Korea’s most promising challenger to Nvidia in AI inference, plans to mass-produce its Atom NPU with Samsung Electronics. The company’s Rebel100 NPU, an HBM3e-based design, caters to larger models and is already commercially deployed in KT Cloud, SK Telecom’s AI platform, and LG Electronics. Rebellions secured $400 million in a Pre-IPO round in March 2026, reaching a valuation of $2.34 billion and bringing its total cumulative funding to $850 million.
FuriosaAI is also emerging as a formidable competitor to Nvidia. Their second-generation NPU chip, Renegade (RNGD), is fabricated on TSMC’s 5-nanometer process, integrating 41 billion transistors and 48GB of High-Bandwidth Memory (HBM). The Renegade chip demonstrated more than twice the power efficiency of Nvidia’s hardware when running Meta’s Llama model. Collaborating with Samsung SDS, FuriosaAI launched Korea’s first NPU-as-a-Service (NPUaaS), broadening access to high-performance, energy-efficient AI inference infrastructure for enterprise customers. Meanwhile, Mobilint is making significant strides in physical AI with its NPUs designed for edge devices.
Government support is a crucial catalyst for these domestic NPU companies. Under the ‘K-NVIDIA Fostering Project,’ the South Korean government invested ₩250 billion ($166 million) in Rebellions, actively nurturing domestic AI semiconductor startups. Furthermore, in August 2026, the National Information Society Agency (NIA) signed agreements with FuriosaAI and Rebellions to procure NPU servers totaling ₩10 billion ($7.2 million). These servers will power government-wide AI services, including an AI chatbot for public officials. The government also plans to convert over 50,000 public CCTV cameras to domestic NPU-based AI monitoring systems over the next five years. This proactive public-sector adoption is expected to vitalize the domestic NPU market and provide essential references for companies targeting international markets.
Notably, Rebellions recently engaged with Nvidia executives to discuss potential technological cooperation. This interaction suggests a possibility for complementary integration of domestic NPU technology with Nvidia’s GPU ecosystem, opening new avenues for global market expansion. The government is also demonstrating heterogeneous computing systems that link AMD GPUs with domestically developed NPUs.
Outlook and Implications
South Korea’s aggressive investments in AI semiconductors and dedicated NPU development efforts represent a significant push to disrupt Nvidia’s market dominance. Should the government’s strategic demand creation combine effectively with the technological advancements of domestic companies, South Korea is well-positioned to emerge as a new force in the AI semiconductor landscape. Investors should closely monitor the commercial performance of Korean NPU firms, particularly their real-world application cases in both public and private sectors, and their ability to forge global partnerships. The success of the ‘K-NPU industrialization strategy’ will serve as a critical indicator for the future trajectory of the Korean economy. Power-efficient NPUs for AI inference are becoming indispensable as AI extends beyond data centers into physical AI domains, including automotive, robotics, factories, and smart devices.
References & Sources
