Fumiki Negishi, HPE HPC·AI APJ (Asia-Pacific, Japan) General Manager and Executive Vice President, diagnosed that the focus of competition in AI infrastructure is shifting from securing GPUs to integrated solutions. HPE views Korea—a market with rapidly growing AI investments—as a core APJ region and plans to strengthen cooperation there.
In an exclusive interview with MoneyToday on the 9th, Vice President Negishi stated, "Customers want not only GPUs but also solutions that combine software, services, networks, and storage." He added, "The rules of AI competition have changed. Simply securing many GPUs is no longer enough to win in this new game."
HPE’s global cloud and AI segment operating profit margin for the third quarter of fiscal year 2026 (May–July) rose from 7% a year ago to 17%, while data center networking revenue surged by 112% compared to the same period last year. Although Vice President Negishi could not disclose APJ-specific results, he explained that demand for AI infrastructure remains strong.
HPE collaborates with NVIDIA across all aspects of AI infrastructure. Vice President Negishi likened their relationship to "a supplier of ingredients and a chef." He emphasized, "If NVIDIA provides high-quality ingredients and recipes, HPE combines them to create the final dish that customers can actually use. We also have our own recipes and ingredients."
Regarding Korea, he said, "In APJ, Korea holds a comprehensive vision for AI across all areas and is actively putting it into action." He noted, "Japan has launched government-led projects in recent years, but Korea’s pace of implementation is faster." He also cited Korea’s diverse industrial base and the presence of global memory companies as key strengths.
The 'Han River' supercomputer, the sixth-generation system being built by HPE at the Korea Institute of Science and Technology Information (KISTI), represents a flagship Korean project. It combines the 'Cray Supercomputing EX4000' with GPUs, HPE’s proprietary 'Slingshot' network, and storage to handle both scientific computing and AI tasks simultaneously.
Vice President Negishi stressed that integrating GPUs with other infrastructure is critical. While acknowledging, "Without sufficient GPU access, competitors will inevitably fall behind," he added, "Securing many GPUs is merely the starting point, not the ultimate goal."
He identified service capabilities as a key challenge for future AI infrastructure expansion. Although power and infrastructure can be scaled with investment and time, services require skilled personnel and training, leading to his conclusion: "Ultimately, people become the bottleneck."
He also anticipated the spread of 'AI grids.' An AI grid connects distributed data centers and edge computing resources via networks when large-scale data centers face power or cooling constraints, enabling AI workloads to be processed across multiple locations. Vice President Negishi observed, "When scale-up reaches its limit, shifting to scale-out is a recurring trend in the computing industry."
HPE plans to expand its role as Korean companies connect AI investments to tangible outcomes. Vice President Negishi stated, "We are actively involved in many projects in Korea. Companies must consider where and how to use AI most effectively. We can help achieve that."
in TECH
