South Korean Samsung Electronics reported that its entire planned 2027 shipment volume of high-bandwidth memory HBM has already been allocated among customers. The company has completed negotiations on volumes and prices, yet the largest customers continue to demand additional batches of chips. This points to a continuing memory shortage amid the massive buildout of artificial intelligence infrastructure.
Kim Jae-jun, vice president of Samsung's memory division, said this on October 29 during the presentation of quarterly results, reports The Korea Herald.
According to the company representative, negotiations on HBM supply in 2027 concluded at prices that meet Samsung's commercial expectations. At the same time, key customers continue to request volume increases.
This concerns the entire volume of HBM Samsung plans to produce next year, not all memory on the global market. Chips of this type are used in high-performance processors and accelerators for AI systems. Their high bandwidth is needed to process large data sets in data centers.
Demand for such components is growing rapidly due to massive investments by technology companies in data centers and computing infrastructure. Samsung forecasts in its financial report that in 2027 demand for AI memory will exceed manufacturers' capabilities even more than in 2026.
Samsung also warns that the memory chip shortage may persist until 2028. As Herald Business explains, even a significant increase in capital investment will not allow a rapid ramp-up of output. Building new semiconductor fabs takes a long time, and lead times for production equipment continue to lengthen.
Additional constraints come from the transition to next-generation memory. HBM4E and HBM5 chips require larger die sizes and more complex manufacturing processes. This makes it harder to simultaneously increase output of HBM and conventional DRAM.
For makers of servers, computers, and smartphones, this situation means a risk of further component cost increases. Part of production capacity is being shifted to more profitable AI products, limiting memory availability for consumer electronics.
Against this backdrop, Samsung is actively expanding long-term contracts. The company has already signed multi-year memory supply agreements with ten large customers and is in talks with another ten. By 2030, the manufacturer aims to cover 60-70% of its production capacity with such agreements.
At the same time, sales of the most advanced chips are growing. In the third quarter of 2026, Samsung's HBM4 revenue more than tripled quarter-on-quarter. This generation already accounts for roughly 60% of the company's HBM sales revenue.
To expand production, Samsung plans to spend about 84.4 trillion won on capital investments in 2026, of which 76.8 trillion won will go to the semiconductor business. The company is also considering expanding production capacity in South Korea and the United States.
As previously reported by "Kurs Ukrainy", back on October 8 Samsung forecast record quarterly earnings thanks to high demand for AI chips. The report released on October 29 confirmed those financial results. At the same time, the new information about fully contracted HBM supply indicates that strong demand secures orders for the manufacturer already for next year.
Based on materials from: The Korea Herald, Samsung Global Newsroom, Herald Business
