Ice and fire coexist! AI prosperity has not yet broken the sluggish state of storage chips.

The AI boom has fueled the production of AI chips by NVIDIA. Although the market generally expects the AI frenzy to drive significant growth in storage chips, it is clear that this growth has not yet materialized.

On Friday, July 7th, Samsung Electronics, the world's largest storage chip manufacturer, released performance guidance stating that it expects second-quarter operating profit to decline by 96% compared to the same period last year. This will be its worst performance in over fourteen years, indicating that even the Korean semiconductor giant has not fully seized the AI opportunity.

In terms of stock prices, NVIDIA has surged nearly 200% this year, while Samsung Electronics has risen by nearly 26% and SK Hynix has seen an increase of almost 50%. Although the stock prices of these two storage chip giants have also risen, Samsung Electronics is even lagging behind the performance of the NASDAQ, and SK Hynix is only slightly better than the overall market.

Storage chips play a crucial role in enabling more advanced AI-based online tools, such as OpenAI's ChatGPT. Working in conjunction with graphics processing units (GPUs) manufactured by companies like NVIDIA, storage chips help accelerate computational speed, making them essential for building faster and more complex AI applications.

Although some analysts believe that the generative AI boom has revived the US tech industry - both tech giants and startups have regained hope due to the prosperity of AI, and even housing prices in the Bay Area have rebounded as a result - generative AI cannot save the broader decline in the tech industry, including the sales of storage chips.

Samsung Electronics expects second-quarter operating profit to be 6 trillion Korean won, equivalent to $458 million, compared to 14.1 trillion Korean won in the same period last year. In addition, Samsung Electronics' second-quarter revenue is expected to decline by 22% to 600 trillion Korean won. Another Korean company, SK Hynix, which is also a storage chip giant, expects an operating loss of 2.8 trillion Korean won in the second quarter of this year, with revenue down 53% YoY to 6 trillion Korean won. Samsung Electronics and SK Hynix will announce their full financial reports later this month. During last week's earnings conference call, the management team of Micron Tech identified new opportunities brought by AI. The company's CEO, Sanjay Mehrotra, stated during the call:

As we have mentioned before, the demand for DRAM in AI servers is six to eight times that of regular servers, and the demand for NAND is three times that of regular servers. In fact, some customers have already started deploying AI servers with higher memory capacity.

According to data from the World Semiconductor Trade Statistics (WSTS), the global revenue for all types of chips is expected to decline by around 10% this year, reaching $515 billion. WSTS predicts that storage chips will experience the most severe decline among major semiconductor types, with revenue expected to decrease by 35% to $84 billion.

Nevertheless, the future prospects of storage chips remain promising in the AI wave.

To adapt to the AI era, Samsung Electronics, SK Hynix, and Micron Tech have all introduced the next generation of DRAM chips called HBM, specifically designed for artificial intelligence systems. HBM, also known as "High Bandwidth Memory," stacks multiple layers of DRAM together and can be packaged with GPUs produced by NVIDIA as a single unit. This significantly improves computational efficiency.

SemiAnalysis, a chip industry consulting firm, indicates that the price of HBM is approximately five times that of standard DRAM chips, potentially bringing greater overall profits to manufacturers. Currently, HBM accounts for less than 5% of global storage chip revenue, but it may exceed 20% by 2026.