SK hynix is discussing equipment orders with semiconductor equipment makers to increase production of its next-generation high-bandwidth memory, HBM4. Korea's etnews reported on November 15. The orders mainly cover back-end equipment for bonding, cutting and inspection, with installation targeted for the first half of 2027. The added capacity is estimated at more than 20,000 wafers per month, equivalent to about 13% of estimated HBM production capacity at the end of 2025. SK hynix itself forecasts that demand for memory will exceed supply until 2030, so attention is turning to how much this expansion will ease the shortage.
Talks on back-end equipment orders for HBM4 expansion
On November 15, etnews reported, citing multiple people at semiconductor equipment makers, that SK hynix is discussing purchases of the equipment needed to manufacture HBM4. The company is reportedly coordinating timing and quantities with firms that supply back-end equipment for bonding, cutting and inspection. The sources are anonymous, and it has not been disclosed whether formal orders have been placed or what the contract values are. An SK hynix representative told the newspaper that the company has continued to order equipment in line with its plans.
According to people at partner companies, SK hynix aims to bring in and install the equipment in the first half of 2027, and has asked for delivery of some equipment by the first quarter of that year. The leading candidate for the expansion is the P&T6 plant in Cheongju, South Korea. The facility was formerly known as M8. SK hynix decided in 2025 to convert it into an HBM production site and has been preparing its manufacturing lines since early 2026.
What matters in this investment is that it targets the back-end process, which turns finished chips into HBM products, rather than the front-end process that makes the DRAM chips. HBM is completed by stacking and bonding multiple DRAM chips and then inspecting them. Even if enough chips can be made in the front-end process, shipments of the final product cannot increase if stacking and inspection capacity falls short. P&T6 is being built out as a site for this back-end work.
According to etnews's report of February 12, 2026, P&T6 was then at the stage of initial orders to verify that its manufacturing lines worked. The newspaper said SK hynix's HBM back-end operations had little free space for new equipment, so no large-scale orders had been seen until then. P&T6, converted from the former M8 plant, became a site where new equipment could be installed.
The current order talks suggest that equipment evaluation at P&T6 is progressing and that the company is moving toward building a full-scale mass-production line. However, SK hynix has made no official announcement about specific order quantities, suppliers or when the equipment will start operating.
An added 20,000 wafers a month is about 13% of end-2025 HBM capacity
In its November 15 report, etnews cited industry sources as saying that P&T6 would add HBM production capacity of more than 20,000 wafers per month, measured in wafers, once it is fully operational. The newspaper also reported in February an industry forecast that P&T6's added capacity would be around 20,000 wafers per month. It further estimated SK hynix's total HBM production capacity at the end of 2025 at around 150,000 wafers per month.
Comparing these figures, P&T6's added capacity is equivalent to about 13.3% of HBM production capacity at the end of 2025.
\[ \frac{20,000}{150,000}\times100\approx13.3\% \]
However, the 20,000-wafer figure is a forecast by industry sources, and the 150,000-wafer figure is etnews's own estimate. Neither is a production capacity published by SK hynix, and the basis for the estimates is not given. The comparison is also across different points in time: capacity at the end of 2025 versus the added capacity expected when P&T6 is fully operational in the future. Because the November report puts the added capacity at "more than 20,000 wafers per month," the actual scale of the expansion could exceed 13.3%.
Meanwhile, demand for memory, including HBM, remains strong. In its third-quarter 2026 results announced on October 28, SK hynix reported revenue of 104.6155 trillion won and operating profit of 83.3338 trillion won, an operating margin of 80%. The company forecasts that, as AI demand grows, the technical difficulty of leading-edge processes and the time needed to build new fabs will constrain supply expansion, and that demand will continue to exceed supply in the memory market until 2030.
The company has signed long-term supply agreements with 14 customers, including major cloud service providers, and the contracted volumes account for more than 50% of its total production capacity in 2027 and 2028. However, that ratio applies to the company's overall capacity, not just HBM, so it is not comparable to the estimated 13.3% increase at P&T6. SK hynix has also not disclosed a specific HBM supply shortfall, so it is not possible to quantify how much this expansion will narrow the gap between supply and demand.
Expanded HBM4 supply also bears on AI infrastructure development in Japan. In its announcement of July 16, 2026, NVIDIA said it plans to deploy 27,500 Rubin GPUs in a Japanese AI platform being built jointly with Noetra. According to NVIDIA's technical documentation, each Rubin GPU carries up to 288GB of HBM4. If all 27,500 GPUs were equipped with 288GB, the total HBM4 capacity required would reach 7.92PB (7,920TB). The HBM4 supply situation could therefore also affect plans for large-scale AI infrastructure like this.
Expansion is under way beyond P&T6. How far will supply capacity grow?
In assessing the scale of this expansion, it is important to note that P&T6 is not SK hynix's only production expansion site. In its third-quarter 2026 earnings announcement, the company said it would raise M15X production capacity to the maximum by the end of 2026 and expand HBM4 production. Accordingly, it expects HBM4 shipment volume and revenue to both exceed those of HBM3E in the fourth quarter of 2026, making it the company's main product.
Even if P&T6's added capacity is about 13% of estimated end-2025 capacity, SK hynix's overall HBM supply capacity would grow further if production increases at other sites such as M15X. etnews's February report also said that, once equipment evaluation at P&T6 is complete, large-scale additional orders were expected as the company moves to build a full-scale mass-production line.
However, SK hynix has not disclosed the scale of M15X's production increase in specific wafer figures. For now, the only comparison possible is between the reported estimated capacity at P&T6 and the estimated HBM production capacity at the end of 2025. On that comparison, P&T6's added capacity alone is a little over 10% of existing capacity, which is hard to see as large enough to resolve the long-term excess demand the company forecasts.
One benchmark that would change this view is whether P&T6's added capacity exceeds 20% of HBM production capacity at the end of 2025. If the 150,000-wafer estimate is correct, that would mean an increase of more than 30,000 wafers per month.
If SK hynix discloses specific figures in its earnings materials or briefings, either for P&T6's added capacity or for overall HBM production capacity, the scale of this expansion can be assessed more accurately. The current figure of about 13% is based on forecasts by anonymous industry sources and estimates by a news outlet, and should be compared again once official production capacity figures are available.