Google and Apple Threaten Nvidia's Dominance; Six Key Technologies from HBM to Quantum Computing in Focus

KAIST Professor Kim Jeong-ho predicted that Nvidia's AI data center market share will fall from the current 90% to 60-70% within 10 years, and that Google and Apple will be the main companies threatening Nvidia's dominance. He explained that Google and Apple will emerge as competitors to Nvidia based on their financial power and ecosystems.

Currently, Nvidia holds an overwhelming share in the AI data center market, but analysis suggests that the competitive landscape could change through Google's self-developed AI chip, TPU, and Apple's vast financial resources and existing ecosystem.

Professor Kim Jeong-ho presented six key technologies to watch in the future AI era. The first is the memory revolution, represented by HBM, HBF, and HBS, which he emphasized as essential for improving AI performance.

The second is 3D packaging technology, a method of stacking multiple chips vertically to maximize performance. The third is chiplet technology, which efficiently connects multiple processors, and the fourth is glass substrates, gaining attention as a next-generation substrate. The fifth is silicon photonics, and the last, sixth, includes next-generation computing technologies such as quantum computing, neuromorphic, and optical computing. These technologies are expected to be key drivers of AI development in the future.

Google possesses its self-developed AI chip, TPU (Tensor Processing Unit), which is evaluated as a 'full-stack' capability with all necessary elements for AI development, from hardware to software. Professor Kim Jeong-ho analyzed that Google's full-stack strength will serve as a powerful competitive advantage in the AI era.

Furthermore, by utilizing profits from various revenue models like YouTube for AI data center investment, financial risks can be reduced. Professor Kim Jeong-ho noted that since Google does not 자체적으로 보유하고 있지 않다 in the memory sector, cooperation with memory companies like Samsung could be an important partnership.

Apple has strong potential to enter the AI market based on its massive cash reserves. It has already built a robust 'circular economy model' through its own devices like iPhones and Macs, and its software ecosystem, making it relatively easy to integrate AI capabilities. Professor Kim Jeong-ho also predicted that Apple might acquire major AI companies like OpenAI if they face financial difficulties in the future.

Professor Kim Jeong-ho emphasized that Korean conglomerates must undergo a major transformation, moving away from their traditional role of supplying parts of the supply chain to a 'full-stack' strategy that leads the ecosystem itself. This is identified as an essential condition for survival in the AI era.

Specifically, he suggested that Samsung should strengthen its full-stack capabilities by developing AI models and securing data centers, while SK Hynix should expand its presence within the AI ecosystem through data center collaborations. For Hyundai Motor, he explained that it should explore new growth engines by pioneering physical AI and agent commerce based on automobiles.

Professor Kim analyzed that it is time for Korean conglomerates to evolve beyond simply supplying components to providing integrated core technologies and services necessary for the AI era.

Professor Kim Jeong-ho suggested that Naver should focus its AI technology on its existing strengths in shopping, search, and news to enter the 'AI agent commerce' sector. This is a strategy for Naver to maximize profits generated in the e-commerce sector through AI and pursue continuous growth based on it.

Professor Kim predicted that Naver would enhance the synergy of its existing services and create new revenue streams through AI agent commerce, similar to how it links search and news services to maintain its shopping platform. Through this, Naver will be able to establish new business models for the AI era, he emphasized.

Professor Kim Jeong-ho emphasized that for South Korea's system semiconductor companies to survive, it is essential to protect and nurture the domestic AI semiconductor startup ecosystem. He expressed concern that if domestic AI semiconductor companies like Rebellions and Divine lose hope, the dream of encompassing South Korea's entire system semiconductor industry will disappear, leaving only memory semiconductors.

He also explained that it is crucial for AI semiconductor startups to secure 'references' by testing and validating their products within domestic data centers. This should provide opportunities to overcome trial and error and advance technological capabilities, and entrepreneurship will play a key role in this process, he added.

Professor Kim Jeong-ho emphasized that to foster innovation in South Korea, the nation must inherit the 'first (第一)' spirit of Samsung founder Lee Byung-chul and the 'have you tried it' spirit of Hyundai Group founder Chung Ju-young. This implies pioneering the future with bold challenges and execution.

He also urged companies to boldly reinvest profits in production facilities and talent development. He suggested that the government should use a portion of the taxes paid by corporations for basic science and technology development, talent nurturing, and building essential infrastructure for the AI era. This underscores the urgent need for national-level investment and support to secure future competitiveness.

Professor Kim Jeong-ho proposed a groundbreaking national project to secure AI talent in South Korea. He suggested introducing a 'military academy-style education system' by establishing AI departments in major universities across the country and providing full tuition and living expenses for students, including through doctoral programs.

Professor Kim explained that the annual budget for such a large-scale support initiative would be in the hundreds of billions of won, which is well within the nation's overall budget. He asserted that such bold investment is necessary to cultivate key talent who will lead the future AI era.

Professor Kim Jeong-ho proposed establishing an innovative national strategic agency, similar to the US DARPA (Defense Advanced Research Projects Agency) model, to achieve South Korea's leap to G3 (top three global powers). This aims to set national future agendas, modeled after Korea's G7 project in the 1990s, and pursue bold challenge research.

DARPA has given birth to countless innovative technologies such as the internet and GPS with the goal of developing technologies that did not exist before. Professor Kim emphasized the need to create an original R&D funding structure that does not overlap with existing overseas research, focusing on high-impact research despite high uncertainty.

Through this, he predicted that Korea could transition from merely being a 'fast follower' of advanced technology to a 'creative leader' that leads the world.

Professor Kim Jeong-ho emphasized that 'power' is a key competitive factor in the AI era and argued that energy policy should be approached practically, not ideologically. He explained that each power generation method has its pros and cons, so a balanced energy portfolio that does not lean towards one side is important.

A stable and efficient power supply system must be built by considering the characteristics of each energy source, including nuclear power, renewable energy (solar, wind), and thermal power (co-generation, LNG). Professor Kim stressed that energy policy needs to be decided quickly and implemented consistently over the long term to secure power competitiveness in the AI era.

Professor Kim Jeong-ho presented three core technologies that South Korea should focus on to secure national competitiveness in the AI era. The first is NPU (Neural Processing Unit) class semiconductors, a sector already supported by the National Growth Fund.

The second is power semiconductors, and the third is silicon photonics technology. These three technologies are essential for improving AI system performance and efficiency, and they are areas where Korea has strengths or significant future growth potential. Professor Kim emphasized that intensive investment and nurturing in these technologies are necessary to enhance national competitiveness.

Professor Kim Jeong-ho explained that future semiconductor systems will evolve into a form of assembling verified components like LEGO blocks, emphasizing that 'packaging,' this assembly technology, will be a core competitive edge. This is because it is becoming almost impossible to design and create an entire system from scratch.

Professor Kim analyzed that packaging technology is essential to secure full-stack competitiveness, and the reason global foundry company TSMC was able to secure large volumes was its advanced packaging solutions. This means that beyond simply producing chips, the technology to integrate them efficiently and precisely will determine success or failure in the future semiconductor industry.

Professor Kim Jeong-ho acknowledged that companies are indeed feeling financial pressure due to massive AI investments. However, he analyzed that the current 'AI bubble theory' is less likely to lead to an actual bubble explosion and is more aimed at adjusting stock prices up and down through the media.

Professor Kim predicted that the AI bubble would not burst all at once, leading to collapse, but rather would be a temporary 'deflation' due to dampened investor sentiment, allowing for 'reinflation' through reinvestment or technological development. In other words, AI investment will continue, and while short-term adjustments may occur, the long-term growth momentum will be maintained, he stated.

Professor Kim Jeong-ho presented a prioritized technology roadmap to watch over the next 10 years. The top priority is the 'memory revolution,' represented by HBM, HBF, and HBS, which he stated is ongoing and will play the most crucial role in improving AI performance.

The second priority is '3D packaging' technology, and the third is 'chiplet' technology, which connects multiple processors. Following these are 'glass substrates' as the fourth priority and 'silicon photonics' as the fifth. Finally, the sixth priority includes next-generation computing technologies such as 'quantum computing,' 'neuromorphic computing,' and 'optical computing.' These technologies are expected to be key drivers that will reshape the future AI industry.

Professor Kim Jeong-ho asserted that the United States will be the ultimate victor in the AI era within the next 10 years. He analyzed that the US's success factor is not just capital, but the historical technological prowess accumulated since the start of Silicon Valley in the 1960s, including semiconductor design, equipment, EDA (Electronic Design Automation), and memory technologies, which will play a decisive role.

In contrast, countries like Russia lack an innovation ecosystem like an AI Valley or Silicon Valley, which will limit their ability to secure AI competitiveness, he pointed out. He predicted that the US will firmly establish its AI leadership based on its long-standing technological infrastructure, talent, and innovation ecosystem.

Professor Kim Jeong-ho emphasized that South Korea must abandon its 'Fast Follower' culture and spirit, which pursued being second, and transition to a vision of becoming a 'Creative Leader,' a first-place nation. To achieve this, he asserted that the country must develop a culture and political system befitting a leading nation.

Specifically, he suggested that national efforts should focus on building many data centers and factories, increasing investment in basic science and technology, and extensively nurturing talent to lead the AI era. He also added that when investing in research, a culture that embraces uncertainty and does not fear failure is needed, and these efforts will contribute to building a 'technology highway'.