AI향 반도체에 필수적인 차세대 메모리 HBM3로 주가를 올리고 있는 SK하이닉스가 창립 40주년을 맞이했다. 이에 창립 기념 특별대담에서 곽노정 대표이사가 미래 반도체 기술 전망과 SK하이닉스의 비전을 밝혔다.

▲ SK Hynix CEO Kwak No-jeong
speaking at a special interview celebrating the 40th anniversary of the company’s founding (Photo: SK Hynix)
The essence of the semiconductor market changes in the AI era… “Customer demands diversify”
No boundaries between memory-CPU-system semiconductors, prospects for technological convergence
SK Hynix, which is raising its stock price with the next-generation memory HBM3 essential for AI-oriented semiconductors, has celebrated its 40th anniversary. In a special interview commemorating the anniversary, CEO Kwak No-jeong revealed the future semiconductor technology outlook and SK Hynix's vision.
SK Hynix CEO Kwak No-jeong held a special conversation with company members on the 10th to commemorate the company's 40th anniversary.
SK Hynix CEO Kwak No-jeong, who presents the future direction of the company, said, “We will innovate memory semiconductors, which have been recognized as general-purpose products, into customized specialty products,” and emphasized, “Only companies that break away from the old ways and satisfy customers will survive.”
As the full-fledged AI era arrives, the specifications of memory semiconductors required by big tech companies are diversifying. SK Hynix announced that it has achieved the world's first mass production of HBM3 and is also accelerating the development of the next-generation product, HBM3E.
.png)
▲Special interview commemorating SK Hynix’s 40th anniversary (Photo: SK Hynix)
The recent boom in generative AI stemming from ChatGPT is accelerating the development of artificial intelligence technology while also spurring development to differentiate AI services and dominate the market by big tech companies.
Accordingly, memory players have traditionally been required to mass-produce small varieties of memory semiconductors such as DRAM and NAND flash to customers, but SK Hynix sees the market outlook as one in which the methods of conducting AI learning are each different, and the memory specifications required by each company are diverse, so memory manufacturers will inevitably have to diversify.
In line with this trend, CEO Kwak predicted that SK Hynix will need to work closely with AI customers at the basic stage from HBM3E, which is scheduled for mass production in 2024, and through this, specifications will be determined, and there will be major changes across the industry, including design and production methods and marketing.
He predicted that the boundaries between memory, CPU, and system semiconductors will disappear and technological convergence will take place in relation to future semiconductor technologies. This means that the scope of memory utilization will expand further.
CEO Kwak said, “It seems that PIM products that embed computational functions into the memory itself can be advanced and applied to quantum computing in the future,” and added that whether SK Hynix can maturely lead this will be a factor that determines the future.