Overview
SK hynix is moving quickly to solidify its position in the next-generation high-bandwidth memory (HBM) market, with recent corporate filings and announcements revealing both expanded production capacity and progress on its latest chip architecture. According to Yonhap News Agency, the South Korean memory maker has placed an order with Hanmi Semiconductor for HBM production equipment valued at 44.2 billion won (approximately US$28.7 million), and separately, it has begun shipping samples of its newest HBM product, the 12-layer HBM4E, to major global customers.
Coverage comparison
Two separate reports from Yonhap News, both published in June, captured distinct but complementary developments. The first, dated June 8, detailed a regulatory filing by Hanmi Semiconductor announcing a deal to supply thermo-compression (TC) bonders to SK hynix. The second, dated June 18, reported SK hynix's own announcement of HBM4E sample shipments. Both stories were reported exclusively by Yonhap, and no conflicting accounts from other outlets were available at the time of writing.
The equipment order story focused on production expansion for HBM4, the memory type expected to power Nvidia's upcoming Vera Rubin AI platform. The sample shipment story highlighted SK hynix's technological progress and competitive positioning against rival Samsung Electronics, which began shipping HBM4E samples roughly a month earlier, according to the same Yonhap report.
Key claims
Equipment order for Nvidia-bound HBM4:
According to Yonhap's June 8 report, Hanmi Semiconductor disclosed in a regulatory filing that it had won a contract to supply TC bonders to SK hynix. The deal is worth 44.2 billion won (US$28.7 million), with deliveries scheduled through early September. Industry watchers, as cited by Yonhap, estimated that the agreement likely covers about 15 units, given that a single TC bonder typically costs around 3 billion won. The equipment includes Hanmi's TC Bonder 4.5 Griffin, an upgraded version of its TC Bonder 4.0, which was launched in May of the previous year.
Yonhap reported that the equipment is intended to expand production of HBM4 products for Nvidia, and that the bonders are presumed to be destined for SK hynix's new M15X fab in Cheongju, about 110 kilometers south of Seoul. This presumption, attributed to industry watchers, was not confirmed by either company in the filing.
HBM4E sample shipments:
On June 18, SK hynix announced it had begun shipping samples of its 12-layer HBM4E chip to major global customers. The company said in a press release, as reported by Yonhap, that it "was able to deliver samples of the 12-stack HBM4E on schedule thanks to its advanced HBM development and production expertise for HBM," and that it would "work closely with partners for mass production in a timely manner."
The company claims the new chip delivers more than 20 percent greater power efficiency than the previous memory model and offers improvements in both performance and power efficiency. These enhancements are expected to improve data processing capabilities for AI training and inference workloads, according to the report.
Customer base and competitive context:
SK hynix's customers include major AI companies such as Advanced Micro Devices, Nvidia, and Google, per the June 18 Yonhap report. The sample shipment comes approximately one month after Samsung Electronics began shipping HBM4E samples, which Samsung described as the world's first shipment of the next-generation AI memory product, according to the same report.
Perspectives
SK hynix's perspective: The company frames its HBM4E sample delivery as evidence of its technological leadership and manufacturing expertise. Ahn Hyun, president and chief development officer, said in a statement: "SK hynix has laid the foundation to strengthen its AI leadership with HBM4E based on its market-leading technological capabilities and manufacturing expertise." He added that through close collaboration with partners, the company would "deliver the value needed in the market while reinforcing technology leadership as a full-stack AI memory creator."
SK Group chairman's remarks: On the day of the equipment order announcement, SK Group Chairman Chey Tae-won spoke at a press briefing in Seoul alongside Nvidia CEO Jensen Huang. According to Yonhap, Chey said his company is making all-out efforts to ensure a stable supply of chips to Nvidia. "Most of our cooperation so far has centered on memory, but from now on, we will elevate our partnership to a higher level," he said. This suggests that the equipment order is part of a broader strategic alliance between SK hynix and Nvidia, extending beyond simple supply arrangements.
Industry watchers' take: Industry observers cited by Yonhap interpret the equipment order as a signal of SK hynix's aggressive expansion of HBM4 capacity, particularly for Nvidia's Vera Rubin platform. The estimated number of bonders (15) and the targeted fab location (Cheongju) point to significant scaling efforts, though these details remain speculative.
Competitive context: The timing of SK hynix's HBM4E sample shipment, coming about a month after Samsung's, highlights the intense competition in the HBM market. Both companies are racing to secure orders from AI chipmakers, and the pace of sample deliveries is seen as a key indicator of who will lead the next wave of AI memory adoption.
Outlook
The recent actions by SK hynix — both the equipment order and the sample shipments — underscore the company's push to maintain its leading position in the HBM market, which is critical to powering the next generation of AI hardware. While the equipment order's direct link to Nvidia's Vera Rubin platform is inferred rather than confirmed, the combination of the filing, Chey Tae-won's remarks, and the industry context suggests that SK hynix is gearing up for substantial HBM4 production. As the competition with Samsung intensifies, the coming months will reveal whether SK hynix can convert its sample shipments into mass production orders and maintain its edge in AI memory.