Can Eridu curb Nvidia's dominance? The rapidly emerging artificial intelligence (AI) technology signals significant changes not only in the competition for computing power but also in the network infrastructure that connects it. Recently, Eridu, a U.S. startup that developed a network switch optimized for the AI era, has garnered attention by successfully raising approximately $230 million (about 310 billion won) in a large-scale investment round. What has particularly drawn buzz is its challenge to the AI hardware market, currently dominated by Nvidia, with a new approach centered on network innovation. This investment is highly significant as it comes at a time when managing network traffic within data centers is becoming increasingly critical and complex due to the explosive growth of AI workloads. Until now, AI technology development has primarily focused on hardware. Among these, Nvidia has played an unrivaled leading role in the market with its high-end GPU (Graphics Processing Unit) technology, including AI accelerators. Nvidia has dominated the AI chip market, building a powerful ecosystem, and many companies rely on Nvidia's hardware and software platforms. However, with the rapid increase in AI workloads and the growing demand for large-scale parallel processing, the limitations of existing approaches have begun to surface. Specifically, it has been pointed out that conventional network infrastructure, primarily designed for general-purpose computing, struggles to efficiently handle the massive parallel processing and ultra-high-speed data transfer requirements essential for AI model training and inference. To overcome these limitations, Eridu offers high-performance, low-latency network switch solutions specifically tailored for AI applications. The company focuses on resolving network issues that cause bottlenecks within data centers, aiming to eliminate data movement bottlenecks in AI computing environments and maximize communication efficiency between AI chips to enhance the overall performance of AI systems. This innovation goes beyond merely reducing bottlenecks, presenting the potential to revolutionize the performance of entire AI systems. In the AI era, data movement and communication optimization are emerging as core values, just as important as individual computing power. During the training and inference of large-scale AI models, thousands or tens of thousands of AI chips must operate simultaneously, exchanging vast amounts of data. If the network becomes a bottleneck in such scenarios, the performance of the entire system will inevitably degrade, regardless of how powerful the AI chips are. Eridu's network switch is designed to address this very issue by improving communication efficiency between AI chips and facilitating smoother parallel processing during training. This approach is noteworthy for being network-centric rather than solely hardware-focused, and it demonstrates the potential to create a new competitive landscape in the AI infrastructure market. Eridu's emergence clearly indicates the growing importance of not just the chips themselves but also the network infrastructure connecting them within the AI hardware ecosystem. Until now, AI technology development has primarily concentrated on faster chips, more memory, and more efficient algorithm development. However, as AI models grow exponentially in size and the volume of data required for training explodes, the role of networks in efficiently connecting and coordinating these elements is becoming increasingly crucial. Eridu has found its differentiation point precisely here. While Nvidia dominates the AI chip market, Eridu is entering the market by attempting to differentiate itself with 'data movement' solutions essential for the AI era. This approach is not merely a technological innovation; it has the potential to transform the very structure of the AI infrastructure market. Until now, AI infrastructure investments have primarily focused on computing hardware like GPUs. However, Eridu's successful fundraising demonstrates that investors are beginning to recognize the new value of network infrastructure. The substantial $230 million investment reflects the expectation that network technology will become a core infrastructure in the AI era. This clearly shows that the advancement of AI technology requires not only hardware design but also significant changes in physical infrastructure and network technology. AI Era: Highlighting the Importance of 'Data Movement' Another significant change that Eridu's technological innovation will bring is the diversification of the market's competitive landscape. Currently, the AI hardware market is under Nvidia's strong dominance, with many companies relying on its ecosystem. In this situation, the emergence of innovative startups like Eridu offers new options to the market. While Nvidia also provides networking solutions, Eridu is attempting to differentiate itself with a specialized a