Nvidia provider SK Hynix posts highest revenue in 6 years on AI chip growth

Norman Ray

International Courant

A person walks previous the emblem of SK Hynix within the foyer of the corporate’s workplace in Bundang, Seongnam on January 29, 2021.

Jung Yeon-Je | AFP | Getty Pictures

SK Hynix, one of many world’s largest reminiscence chip makers, reported on Thursday that its second-quarter revenue hit a six-year excessive, sustaining its management in superior reminiscence chips which might be key to synthetic intelligence (AI).

Under is a comparability of SK Hynix’s second-quarter outcomes with LSEG SmartEstimate, which relies on analyst forecasts which might be persistently extra correct:

Income: 16.42 trillion Korean gained (about $11.86 billion), versus 16.4 trillion Korean gained Working revenue: 5.47 trillion Korean gained, versus 5.4 trillion Korean gained

Working revenue within the second quarter reached its highest degree because the second quarter of 2018, recovering from a lack of 2.88 trillion gained in the identical interval final 12 months.

Gross sales for April by way of June rose 124.7 % from 7.3 trillion gained recorded a 12 months in the past, the best quarterly gross sales within the firm’s historical past, in response to LSEG knowledge obtainable since 2009.

SK Hynix mentioned on Thursday {that a} continued improve within the total costs of its reminiscence merchandise – pushed by robust demand for AI reminiscence, together with high-bandwidth reminiscence – led to a 32% improve in income in comparison with the earlier quarter.

The South Korean large provides high-bandwidth reminiscence chips appropriate for AI chipsets for corporations like Nvidia.

Shares of SK Hynix fell as a lot as 7.81% on Thursday morning.

The declines got here as South Korea’s Cospi index misplaced as a lot as 1.91% after US expertise shares bought off in a single day, following disappointing Alphabet And Tesla earnings. These experiences mark buyers’ first take a look at how mega-cap corporations fared within the second quarter.

“Within the second half of this 12 months, robust demand for AI servers is predicted to proceed, in addition to a gradual restoration in typical markets with the launch of AI-enabled PCs and cellular units,” the corporate mentioned in its earnings report on Thursday.

Benefiting from robust demand for AI, SK Hynix plans to “proceed its management within the HBM market by mass-producing 12-layer HBM3E merchandise.”

The corporate is predicted to start mass manufacturing of the 12-layer HBM3E this quarter after offering samples to main prospects. The merchandise are anticipated to ship to prospects within the fourth quarter.

Tight provide

Main reminiscence expertise corporations reminiscent of SK Hynix have been quickly increasing their HBM capability to satisfy the rising demand for AI processors.

HBM requires extra wafer capability than common dynamic random entry reminiscence merchandise – a sort of pc reminiscence used to retailer knowledge – which SK Hynix says additionally it is fighting tight provide.

“Capital expenditure wants are additionally rising to satisfy demand for typical DRAM and HBM, which require extra wafer capability than common DRAM. Due to this fact, this 12 months’s capex degree is predicted to be increased than what we anticipated initially of the 12 months,” SK Hynix mentioned.

“Though extra capability is predicted to extend subsequent 12 months as a consequence of elevated industrial funding, a good portion of it will likely be used to ramp up HBM manufacturing. Thus, the tight provide scenario for typical DRAM is more likely to persist.”

SK Kim of Daiwa Capital Markets mentioned in a June 12 notice that they anticipate “tightness within the HBM and reminiscence market to persist till 2025, making a bottleneck in HBM manufacturing.”

“Due to this fact, we anticipate the favorable value setting to proceed and SK Hynix to attain strong earnings in 2024-2025. The corporate advantages from its aggressive place in HBM for AI graphics processing models and high-density enterprise SSD (eSSD) for AI servers, which is able to result in a revaluation of the inventory,” Kim mentioned.

The availability of high-bandwidth reminiscence chips has change into scarce because of the explosive adoption of AI, fueled by giant language fashions like ChatGPT.

The AI ​​growth is predicted to proceed Provide of high-end reminiscence chips might be tight this 12 months, analysts have warned. SK Hynix and Micron mentioned in Could they’d run out of high-bandwidth reminiscence chips for 2024, whereas shares for 2025 are additionally practically bought out.

Giant language fashions require many high-performance reminiscence chips. These chips allow the fashions to recollect particulars from earlier conversations and person preferences, with a view to generate human-like responses.

SK Hynix has largely led the high-bandwidth reminiscence chip market, because it was the only real provider of HBM3 chips to Nvidia earlier than changing into its rival. Samsung is claimed to have handed the exams for using its HBM3 chips in Nvidia processors for the Chinese language market.

The corporate expects to ship the subsequent era 12-layer HBM4 from the second half of 2025.

Nvidia provider SK Hynix posts highest revenue in 6 years on AI chip growth

World Information,Subsequent Massive Factor in Public Knowledg

Share This Article
Exit mobile version