Q&A with QCT President Mike Yang: The Key Role Generative AI Models Play and Its Evolution
The server-related market continues to expand, as the US semiconductor giant NVIDIA became the world's largest market capitalization on the New York Stock Exchange at the end of May and into the month of June. Taiwanese company Quanta Cloud Technology (QCT), a global data center solution provider and a subsidiary of Quanta Computer Inc., is also riding this trend and strengthening its business development in anticipation of the further spread of artificial intelligence (AI). Looking back on COMPUTEX 2024, Dempa Shimbunsha, a major Japanese newspaper focused on the electronics industry, spoke to QCT President, Mike Yang, about their future strategy, including that for the Japanese market. Q. What impact has generative AI, the main theme of this exhibition, had on the market?A. With the advent of generative AI, the market is expanding explosively. Until now, AI was used for single-purpose tasks, but it is now widely available for general use, such as text-to-image generation. On the other hand, training generative AI requires huge amounts of data, and the demand for computing power is also increasing. Public clouds meet these needs, and many companies are beginning to see the value of the investment. Q. Generative AI, such as Microsoft's new AI function for PCs "Copilot" and OpenAI's GPT-4, is evolving every day. Furthermore, the accuracy of generation continues to improve by using confidential data within companies through RAG (Retrieved Augmented Generation). What is the outlook for the server and cloud markets moving forward?A. With NVIDIA at the center, we have integrated CPUs (central processing units) and GPUs (graphics processing units) into servers, and we have systems that are best matched for the purpose of generative AI. This market, which has hardware and software and a business model, cannot go bad. This trend will continue for at least the next 10 years. Q. What are QCT's focus areas?A. As a company that handles both hardware and software, we are constantly studying the latest AI technologies. However, the ultimate focus is still on hardware. We are particularly confident in the power solution for the cabinet of our new AI computing platform, the NVIDIA GB200 NVL72, which is equipped with NVIDIA's cutting-edge Grace Blackwell Superchip. Other companies are also developing similar systems, but ours is based on a liquid cooling system design that we've been working on for the past three years. This technology is essential for reducing power consumption to support large-scale use in data centers. In the future, we will ship 50% air-cooled and 50% liquid-cooled servers. This is likely to become the trend.Q. What are your expectations on future business developments?A. The market is sustained by the demand for powerful GPUs driven by the evolution of AI models. We need to consider whether we are on the side of operating public cloud services where AI runs or on the side of using them. NVIDIA's development pace is extremely fast, with technological updates occurring annually. If we can't keep up with industry progress, we'll quickly fall behind. This was reaffirmed at this exhibition, which was attended by many top executives. The progress of generative AI is also within our expectations. We are confident that we can provide solutions as the best partner. Q. What do you think of the Japanese market?A. Compared to the United States, Japan is still just starting their cloud service deployment. One reason may be that Japanese companies tend to focus more on the domestic market. I believe the government should take a national approach, such as investing in and developing large-scale data centers. There is a need to increase investment in data centers and find profitable business models. QCT would like to support this effort to build up our presence in Japan.
[Source] 【始動 AI時代-COMPUTEX2024から】〈7〉(dempa-digital.com) . |