Abstract:
Global artificial intelligence infrastructure construction continues to drive up memory demand, and Samsung Electronics is locking in sales in the next few years through long-term supply agreements. According to Korean media citing securities analysis reports, Samsung has signed large-scale memory supply agreements with large global technology companies. About 70% of general DRAM production capacity is guaranteed through long-term contracts, and all high-bandwidth memory (HBM) planned to be produced in 2027 has received orders. As memory supply continues to be tight, related product prices are expected to rise further, which is expected to drive profit growth in Samsung's semiconductor business, but may also increase the manufacturing costs of smartphones, tablets and other consumer electronics products.
According to a report released by GSMArena on October 20, 2026, South Korea's Yonhap News Agency previously disclosed that Samsung has signed a five-year memory supply contract with a large global technology company. The relevant agreement covers most of Samsung’s general DRAM supply and all HBM products planned to be produced in 2027. For about 30% of general-purpose DRAM production capacity that has not been locked in long-term contracts, market demand is still very strong, and multiple customers are competing for the remaining supply.
This means that Samsung's future memory business is gradually shifting from a model that relied more on spot market prices and short-term orders in the past to a supply arrangement with more long-term visibility. For memory manufacturers, long-term contracts can help companies plan production capacity, equipment investment and product mix in advance, and can also reduce operational risks caused by sudden changes in market demand. However, long-term contracts do not necessarily mean that all products are sold at fixed prices, and the specific price adjustment mechanism still depends on the terms of the contract.
The current tense situation in the memory market is closely related to the rapid expansion of artificial intelligence data centers. Training and inference of large AI models require large amounts of GPUs, dedicated accelerators, and high-speed memory. HBM, in particular, has become an important component of high-performance AI accelerators due to its ability to provide extremely high memory bandwidth in a limited space.
Compared with traditional DRAM, HBM usually achieves high-bandwidth data transmission by vertically stacking multiple layers of memory chips and utilizing advanced packaging technology. This design can meet the data throughput requirements of AI processors in large-scale parallel computing, but the manufacturing process is more complex, and capacity expansion is also restricted by factors such as yield, packaging capabilities, and advanced process resources.
With the rapid growth in demand for AI accelerators, HBM has gradually become a key market for memory manufacturers to compete for. Companies such as Samsung, SK Hynix and Micron are all expanding investment in R&D and production of related products, hoping to obtain long-term orders from global cloud computing service providers, chip design companies and data center operators.
Market forecasts cited by GSMArena say HBM prices may rise significantly next year. However, different reports have different estimates of the increase: the forecast range mentioned by the website is 150% to 200%, while relevant analysis by Korean securities agency KB Securities predicts that HBM's average selling price will increase by 100% to 150% year-on-year in 2027. These figures are agency forecasts, not confirmed market transaction prices. The actual increase will still be affected by factors such as product generation, supply scale, and customer contract conditions.
The new generation of HBM4 products is expected to be an important factor driving price increases. Compared with the previous generation of HBM, HBM4 has further developed in terms of interface bandwidth, data transmission capabilities and system integration, and can meet the more stringent performance requirements of the next generation of AI accelerators. As HBM4's share of Samsung's memory business increases, product structure changes may also push up the overall average selling price.
At the same time, the general DRAM market is also facing supply pressure. The demand for high-capacity DDR5 memory in servers continues to increase, and AI inference also requires a large amount of memory to store model parameters, intermediate calculation results, and KV cache data. Data center operators not only need to purchase GPUs and HBM, but also need to configure a large amount of server memory, which makes the impact of AI infrastructure construction on the traditional DRAM market continue to expand.
KB Securities predicts that the average selling price of DRAM in the first quarter of 2027 may see a double-digit percentage increase from the previous quarter. If this forecast comes true, memory manufacturers will have the opportunity to benefit from both shipment scale and price increases, rather than relying solely on sales growth for revenue.
Samsung Electronics' semiconductor business is handled by the Equipment Solutions Division, which includes memory, wafer foundry, system LSI and other businesses. As demand for high-bandwidth memory and server DRAM continues to grow, the segment's profit performance is expected to improve further. South Korea's KB Securities predicts that Samsung Electronics' operating profit may reach 367 trillion won in 2026 and 555 trillion won in 2027, a year-on-year increase of approximately 51%.
However, these figures are forecasts by securities institutions and are not the performance guidance that Samsung has announced. The final outcome still depends on memory prices, capacity utilization, product yields, capital expenditures, and the actual progress of global AI infrastructure investments.
The rise in memory prices is not entirely beneficial to Samsung itself. Samsung Electronics operates both the semiconductor business and the consumer-oriented mobile experience business, which is responsible for the Galaxy series of smartphones, tablets, laptops and other consumer electronics products.
When DRAM supply is tight and prices continue to rise, Samsung's mobile device business also needs to compete with data center customers for some memory supply. Even if Samsung can coordinate the procurement and production of different business units internally, it cannot completely eliminate the cost pressure caused by the overall price increase in the memory market.
This means that future Galaxy series products may face higher material costs. For flagship mobile phones, the increasing memory capacity, improved processor performance and the addition of AI functions will make memory configuration more important. If DRAM procurement costs rise significantly, Samsung may respond by adjusting product pricing, configuration mix, procurement strategies or other cost items.
GSMArena therefore pointed out that there is price increase pressure on the Galaxy S27 series of mobile phones. However, there is currently no official news from Samsung that can confirm the final price of the Galaxy S27 series, and it cannot be concluded that all models will inevitably increase in price just based on the increase in memory prices. The final pricing still depends on product positioning, market competition, exchange rates, other component costs and Samsung's profit strategy.
From an industry level, Samsung's current order situation reflects that the structure of the semiconductor industry is changing. In the past, the DRAM market has often been affected by cyclical fluctuations in demand for PCs, smartphones and servers. Memory manufacturers may expand production capacity when demand exceeds supply, and suffer price drops when supply exceeds supply. Today, the long-term construction needs of AI data centers are bringing new growth momentum to the high-end memory market.
But this does not mean that the memory industry will no longer experience cyclical fluctuations. Data center investment may be adjusted as financing conditions, commercial returns on AI applications, and technology roadmaps change. As memory manufacturers continue to expand production, new supply pressure may also form in the future. Long-term contracts can increase order visibility, but they cannot completely eliminate market risk.
For Samsung, locking in a large number of DRAM and HBM orders in advance will not only help stabilize future revenue, but also enable the company to arrange advanced memory production capacity in a more targeted manner. For customers, signing long-term agreements in advance can help ensure the supply of key chips and reduce the risk of future memory shortages affecting server deployment.
However, the AI industry’s growing demand for memory may also transmit cost pressure to other electronic products. Smartphones, tablets, personal computers and gaming devices all require DRAM or other storage devices. If the prices of related products remain high for a long time, consumer electronics manufacturers will have to make trade-offs between product selling prices, configuration levels and profit margins.
The large-scale long-term orders obtained by Samsung this time show that AI infrastructure is becoming an important source of demand in the global memory market. In the next few years, the supply capacity and price changes of HBM and server DRAM, as well as the mass production progress of new generation products, will directly affect the profitability of Samsung's semiconductor business, and may also indirectly affect the cost and pricing of consumer electronics products such as Galaxy.
Comments