AI memory boom drives SK hynix to record quarterly profit

Source Cryptopolitan

SK hynix posted record second-quarter results on Wednesday as surging demand for high-bandwidth memory (HBM) chips used in AI servers continued to reshape the semiconductor industry.

The South Korean chipmaker reported revenue of 79.3187 trillion won and operating profit of 60.5426 trillion won, up 257% and 557%, respectively, from a year earlier. First-half revenue also crossed 100 trillion won for the first time in the company’s history.

Even with record earnings, SK hynix shares dropped post-announcement as investors expected even stronger results as AI-driven expectations had been growing for months, demonstrating just how high expectations in the AI chip industry have grown.

Why memory sits at the center of the AI trade

High Bandwidth Memory (HBM) has emerged as one of the essential parts of AI infrastructure. By vertically stacking memory chips, HBM provides much more bandwidth and better power efficiency than regular DRAM, enabling Nvidia’s and all other chip manufacturers’ AI accelerators to carry out huge amounts of work more efficiently.

SK hynix, the leading HBM supplier for Nvidia, can be considered one of the strongest indicators of the demand for AI infrastructure.

According to its earnings announcement, SK hynix reported growing demand for its products, as AI continues towards more advanced “agentic” systems capable of performing higher-level functions.

In order to be able to meet future demand, the company has signed long-term contracts with approximately 10 major clients, granting them the right to reserve production capacity years in advance.

The trend manifested in the global memory market is similar. TrendForce anticipates that prices of contracts for DRAM will rise in the range of 58% to 63% for the third quarter, whereas prices of NAND flash could rise about 70% to 75%, with HBM and server memory being prioritized over consumer goods by the manufacturers. The increase in prices is attributed to limited availability of memory in the market, investments in hyperscale cloud, and long-term contracts.

Samsung and Micron are riding the same wave

SK hynix is not the only memory maker benefiting from AI. Recently, Samsung Electronics said that it expects that its operating profit for the second quarter will be 89.4 trillion won and its revenue will be about 171 trillion won, as it gained from strong demand for AI memory although there is a question mark on whether this sudden growth in the industry will continue.

Samsung has also taken an aggressive stance on HBM4. As reported by TrendForce, it is the first company to make HBM4 commercially available in large quantities after commencing mass production in February. It is also reported that Samsung has already generated more than $1 billion in income from HBM4 sales within only four months of its product release.

The company states that its commercial HBM4 offers data transfer performance of 11.7 Gbps per pin, which is well above the JEDEC minimum of 8 Gbps, and boasts better power and thermal efficiency.

Micron Technology has also published record results for the quarter. CEO Sanjay Mehrotra called memory “a strategic asset” in the age of artificial intelligence, emphasizing that hyperscale customers are signing more multi-year supply contracts to ensure supplies into the future.

The squeeze is landing on phones and PCs

The AI boom is also creating shortages elsewhere.

According to IDC, there may be an “unprecedented memory chip shortage” that could extend to as late as 2027 due to the increasing amount of production devoted to HBM for AI accelerators, rather than PCs and smartphones.

According to the research company, the global supply of DRAM will increase by roughly 16% and the supply of NAND will grow by about 17%, both of which are lower than historical averages.

As supply starts to tighten, smartphone manufacturers will have the option of either absorbing the higher costs of components, lowering the specifications of their devices, or passing the price increases onto consumers. In the downside scenario from IDC, worldwide smartphone shipments might drop by as much as 5.2%.

HBM4 competition is heating up

Competition in next-generation AI memory is also intensifying. SK hynix began mass shipments of HBM4 during the quarter and plans to expand production in the second half.

The company said its latest chips meet customers’ performance requirements while improving power efficiency, and it has already completed sample shipments of HBM4E for future AI accelerators.

Nevertheless, Samsung began producing HBM4 commercially before anyone else and therefore has gained a significant competitive edge. TrendForce predicts that this year the company will earn more than $1.2 billion from the sales of HBM4, becoming the first supplier to exceed $1 billion in HBM4 sales.

On the flip side, SK hynix’s strategy has been to supply the HBM3E market first before expanding production of HBM4, the company expects to ramp up production from Q3 onwards. Looking ahead, SK Hynix intends to expand its advanced chip packaging while improving the production of NAND chips with strict budget planning.

The advent of AI has caused a major change in the business of memory chip producers whose focus has shifted away from mere production to bringing the new generation of faster and better-performing AI memory chips to the market.

 

Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Disclaimer: For information purposes only. Past performance is not indicative of future results.
placeholder
Gold Price Analysis (XAU/USD): Gold Falls to 6-Month Low as Inflation Fuels Rate Hike Bets, A Buying Opportunity or a Falling Knife? Gold hit a 6-month low on Fed rate hike bets. However, strong central bank buying and technical indicators suggest potential tactical bounces and long-term accumulation windows.
Author  Mitrade Team
6 Month 12 Day Fri
Gold hit a 6-month low on Fed rate hike bets. However, strong central bank buying and technical indicators suggest potential tactical bounces and long-term accumulation windows.
placeholder
XRP Price Prediction for July 2026: Can Buyers Finally Break the Downtrend?XRP (XRP) price trades near $1.05, caught between a year-long downtrend and a sudden burst of buying.July has historically rewarded XRP holders. This year the month arrives with on-chain accumulation
Author  Beincrypto
6 Month 30 Day Tue
XRP (XRP) price trades near $1.05, caught between a year-long downtrend and a sudden burst of buying.July has historically rewarded XRP holders. This year the month arrives with on-chain accumulation
placeholder
What to Expect From Ethereum (ETH) in July 2026Ethereum (ETH) enters July 2026 trading near $1,570, close to multi-month lows, after recording its first run of three consecutive red quarterly candles in its history.On-chain data and price charts n
Author  Beincrypto
7 Month 01 Day Wed
Ethereum (ETH) enters July 2026 trading near $1,570, close to multi-month lows, after recording its first run of three consecutive red quarterly candles in its history.On-chain data and price charts n
placeholder
NVIDIA Price Forecast: Michael Burry Shorts NVDA, but Analysts See $299On July 1, NVIDIA (NASDAQ: NVDA) sits at $198.34, failing to break above the former support level that is now serving as resistance between $198 and $205 on the 2H chart's downward blue c
Author  TradingKey
7 Month 02 Day Thu
On July 1, NVIDIA (NASDAQ: NVDA) sits at $198.34, failing to break above the former support level that is now serving as resistance between $198 and $205 on the 2H chart's downward blue c
placeholder
Meta Compute Launch Sends AI Compute Stocks Tumbling GloballyMeta’s plan to sell surplus computing power hit chip stocks hard on Wall Street. Meta’s own shares climbed nearly 9% on the news.The announcement flipped years of assumed AI compute scarcity into a su
Author  Beincrypto
7 Month 02 Day Thu
Meta’s plan to sell surplus computing power hit chip stocks hard on Wall Street. Meta’s own shares climbed nearly 9% on the news.The announcement flipped years of assumed AI compute scarcity into a su
goTop
quote