Samsung warns: memory chip shortage will worsen and prices will rise until 2029

What has happened?

During its second-quarter 2026 earnings conference, Samsung, the world's largest memory chip maker, announced that the supply shortage of DRAM and NAND memories will worsen in 2027 and continue through 2028. Jaejune Kim, executive vice president of the memory business, was blunt: demand for chips for artificial intelligence and data centers far exceeds production capacity, and long-term contracts with the world's five largest data center companies commit between 60% and 70% of its total capacity for at least five years. This announcement comes after Samsung's semiconductor division multiplied its profits by more than 250 compared to the previous year, reflecting the manufacturer's enormous pricing power in a market with limited supply.

But how did we get here? To understand it, we have to go back to the 2020 pandemic, when consumer electronics demand soared and caused a global semiconductor crisis. Although that initial crisis was gradually resolved, the explosion of generative artificial intelligence from 2023 created a new massive demand for high-bandwidth memory (HBM) chips and high-capacity server memory. Samsung, SK Hynix, and Micron redirected much of their production to these markets, neglecting memory for general consumption. Now, with long-term contracts with hyperscalers (Google, Amazon, Microsoft, Meta, and Apple), the capacity available for the rest of the market has drastically decreased.

Why is it important?

The memory chip shortage is not a new problem, but the magnitude and duration anticipated by Samsung change market expectations. Until now, many analysts expected a gradual recovery starting in 2026. However, the confirmation that data center contracts extend until 2029 implies that prices for end consumers—whether users wanting to upgrade their PC RAM or buy an SSD—will remain high for at least three more years. According to Reuters data, Samsung expects consumer DRAM prices to rise another 15-20% in 2027, while NAND could rise an additional 10-15%.

Moreover, Samsung's dominance in the memory market (it controls about 40% of global DRAM production and a similar percentage of NAND) means its decisions have a direct impact on the entire industry. Other manufacturers like SK Hynix and Micron will likely follow a similar strategy, prolonging the supply crisis globally. In fact, SK Hynix has already announced that its data center contracts represent 65% of its capacity, and Micron expects to reach 60% by 2027. This means there is no alternative production buffer in the short term.

The impact is not limited to price. Memory availability is also being affected: delivery times for DDR5 DRAM orders have doubled to 16-20 weeks, and high-capacity NVMe SSDs are subject to rationing. For companies that rely on servers, this translates into delays in expanding their infrastructure and higher operational costs. A Gartner report estimates that memory spending for data centers will grow 30% in 2027, directly impacting tech companies' margins.

Consequences for businesses and users

For computer manufacturers, the rise in memory component prices will translate into an increase in the cost of final products. Consumers will have to pay more for laptops, desktops, and smartphones, or settle for more modest memory configurations. For example, a laptop with 16 GB of DDR5 RAM that cost 800 euros in 2025 could cost 950 euros in 2027. Companies that depend on servers and data centers will also see their operational costs increase: a server with 256 GB of RAM could cost 20% more than two years ago.

On the other hand, the situation benefits memory chip manufacturers, who are seeing their margins soar. Samsung, SK Hynix, and Micron are achieving record profits, allowing them to invest in new factories, but building new chip manufacturing plants (fabs) takes years, so supply will not increase significantly in the short term. Samsung is building a new factory in Taylor, Texas, which will not be operational until 2029, and SK Hynix has expansion plans in South Korea that will not bear fruit until 2028. Meanwhile, these companies' profit margins are around 40-50%, levels not seen since the dot-com bubble.

The consumer market is also feeling the impact on innovation. PC manufacturers like Dell, HP, and Lenovo are delaying the adoption of DDR5 in low-end models to keep prices competitive, slowing the renewal of the installed base. Additionally, users wanting to upgrade their equipment face prohibitive prices: a 32 GB DDR5 kit currently costs around 120 euros, compared to 70 euros two years ago. For businesses, the cost of maintaining servers with DDR4 memory has decreased, but at the expense of lower performance.

What should readers know?

If you are thinking about buying a new PC or upgrading the memory of your current equipment, prices will most likely remain high for the next few years. No significant drop is expected until at least 2029, when long-term contracts with data centers begin to expire and Samsung can allocate more capacity to the consumer market. However, there are factors that could accelerate normalization: the entry of new players like China's YMTC, which is increasing its NAND production, although still with quality and capacity limitations. Also, the possible slowdown in AI demand if investment returns do not meet expectations, but this is speculative.

Meanwhile, it is advisable to watch for possible alternatives, such as using DDR4 memory instead of DDR5 (although the latter offers better performance) or buying lower-capacity SSDs. It is also recommended to follow the evolution of Chinese memory manufacturers like YMTC, which could increase competition and pressure prices down, although for now their impact is limited. Another option is to consider buying refurbished or previous-generation equipment, which still uses DDR4 and is more affordable.

In summary, the memory chip shortage will worsen before it improves. High prices are the new normal until the end of the decade, and consumers should adjust their expectations and budgets accordingly. Companies, for their part, should plan their infrastructure investments with longer time horizons and consider strategies for efficient memory usage.