SK Hynix, a major player in the global memory chip industry, declared on Thursday that its profit for the second quarter reached its highest point in 6 years, maintaining its dominance in cutting-edge memory chips crucial for AI computation.

These are the second-quarter figures from SK Hynix in comparison with LSEG SmartEstimate, which leans towards estimates from consistently accurate analysts:

• Revenue: 16.42 trillion Korean won (around $11.86 billion), versus 16.4 trillion Korean won
• Operating profit: 5.47 trillion Korean won, versus 5.4 trillion Korean won

The operating profit in the June quarter reached its highest level since Q2 of 2018, bouncing back from a 2.88 trillion won loss in the same period the previous year.

Revenue from April to June surged by 124.7% from 7.3 trillion won reported a year ago. This marked the highest quarterly revenue in the company’s history, data available since 2009 reveals.

SK Hynix mentioned on Thursday that a steady increase in the overall prices of its memory products, fueled by robust demand for AI memory like high-bandwidth memory, resulted in a 32% revenue increase compared to the previous quarter.

The South Korean corporation supplies high-bandwidth memory chips catering to AI chipsets for firms such as Nvidia.

Shares of SK Hynix dropped by up to 7.81% on Thursday morning.

The decline occurred as the South Korean Kospi index tumbled by as much as 1.91% following the drop in U.S. tech stocks overnight, triggered by disappointing earnings from Alphabet and Tesla. These reports offer investors the first insight into the performance of megacap companies in Q2.

Leaders in memory solutions like SK Hynix have been aggressively boosting their high-bandwidth memory capacity to meet the surging demand for AI processors.

High-bandwidth memory necessitates more wafer capacity than standard dynamic random access memory products – a type of computer memory used to store data – a situation exacerbated by the tight supply, as SK Hynix highlighted.

SK Hynix stated, “The need for investments is rising to cater to the demand for conventional DRAM as well as HBM, which demands more wafer capacity than regular DRAM. Hence, the capital expenditure level for this year is anticipated to surpass initial estimates.”

While overcapacity is predicted to rise next year due to increased industrial investments, a substantial portion is anticipated to be utilized for ramping up HBM production. Therefore, the scarcity in supply for conventional DRAM is likely to persist.

In a note dated June 12, SK Kim from Daiwa Capital Markets mentioned their expectation of a “tight HBM and memory supply until 2025 due to a bottleneck in HBM production.”

“Hence, we foresee a favorable pricing environment and anticipate SK Hynix to achieve robust earnings in 2024-25, benefiting from its prowess in HBM for AI graphics processing units and high-density enterprise SSDs for AI-servers, leading to a reevaluation of the stock,” Kim stated.

The supply of high-bandwidth memory chips has been stretched due to the explosive adoption of AI driven by large language models like ChatGPT.

The AI surge is expected to continue to strain the supply of high-end memory chips this year, caution industry analysts. In May, SK Hynix and Micron disclosed they had exhausted their high-bandwidth memory chip supply for 2024, with the 2025 stock nearing depletion.

Extensive language models demand high-performance memory chips, allowing them to recall details from past interactions and user preferences to generate human-like responses.

SK Hynix has primarily led the high-bandwidth memory chip market, being the exclusive supplier of HBM3 chips to Nvidia until reports emerged that Samsung had validated its HBM3 chips for use in Nvidia processors in the Chinese market.

The company announced its intention to ship the next-generation 12-layer HBM4 starting from the latter half of 2025.