The prices of Random Access Memory (RAM) have seen a significant jump in the past 18 months, in an unusual crisis for an industry that has long been accustomed to gradually declining memory prices. According to the report, pressure on the memory market is expected to continue for the remainder of 2026, amid a massive increase in demand from artificial intelligence companies and data centers, leading to higher costs for RAM in computers, phones, and tablets.

What is RAM and why is it important?

RAM is the component that stores the data and instructions that the processor needs quickly while running programs. When the processor requires specific data, RAM retrieves it from storage and temporarily holds it so the processor can access it more rapidly. The more RAM a device has, the more data it can keep ready for use. Therefore, adding more RAM is often a more effective way to improve computer performance than relying solely on a faster processor.

3 companies control most of the RAM market

The vast majority of global RAM chips are produced by three major companies: Samsung, SK Hynix, and Micron. These companies own massive factories that require enormous investments and highly sophisticated manufacturing technologies. There are different types of memory, but the most commonly used in consumer devices is DRAM, including DDR memory used in computers and LPDDR widely used in phones and mobile devices. However, there's another type that has become more important with the rise of artificial intelligence: HBM or High Bandwidth Memory, which boasts higher speeds and bandwidth and is frequently utilized in data centers and AI model training.

Artificial Intelligence pulls memory from the market

Although DRAM and HBM are different types of memory, they rely on the same manufacturing resources, including silicon wafers. HBM requires significantly more manufacturing resources; in 2024, Micron CEO Sanjay Mehrotra said that producing one unit of HBM consumes about three times the supply of silicon chips needed to produce DDR5 memory aimed at consumers. This means that increasing HBM production for artificial intelligence data centers comes at the expense of part of the productive capacity that could have been directed towards traditional DRAM memory used in user devices.

DDR5 Prices Have Skyrocketed by Over 400%

Data from PCPartPicker reveals the magnitude of the rise in memory prices over the past 18 months. The price of a package containing two DDR5-4800 modules, each with a capacity of 16GB, has increased from less than $100 at the beginning of 2023 to more than $400 today. Similarly, the price of a 32GB DDR5-6000 module has risen from under $250 to over $1,200—a price that is close to the cost of an entire MacBook Air.

Why Have Prices Reached This Level?

The short answer, according to the report, is artificial intelligence. The broader reason is that AI companies, backed by massive investments, have started booking huge amounts of production capacity for data center components, including memory. For many years, the RAM market was characterized by a high degree of predictability in demand. However, the AI boom has quickly changed this equation as technology companies race to build large data centers and expand computing capabilities.

AI Data Centers Consume Massive Amounts of Memory

The report provides an example of the scale of consumption through xAI’s Colossus data center in Memphis, USA. According to the data provided, the center houses around 555,000 NVIDIA graphics processing units worth approximately $18 billion. This system includes around 520,000 Grace Blackwell 200 units with memory up to 192 gigabytes per unit and 30,000 Grace Blackwell 300 units with memory up to 288 gigabytes, in addition to 30,000 H100/H200 units. The report estimates that this center alone may rely on as much as 112 million gigabytes of memory—an amount equivalent to the memory found in more than 14 million iPhone 17s or over 9.3 million Galaxy S26 phones. It emphasizes that the memory used in data centers is not the same as the memory inside phones; rather, it relies on shared manufacturing resources.

The Crisis Could Last for Years

The problem does not appear to be short-term. According to the report, leaders at memory manufacturers expect continued pressure on supply in the coming years. The CEO of SK Hynix said that conditions could become more difficult by 2027, while the CEO of Micron predicted that conditions of supply shortages or tight supplies would continue beyond 2027. The report also cites Deloitte’s projections that the market may not see a real breakthrough until 2030 at the earliest.

Why don't Samsung, Micron, and SK Hynix make more RAM?

The solution might seem simple: increase production. However, building a new memory factory is not a quick process. Creating a new facility requires billions of dollars and years of development. In addition to highly precise manufacturing equipment and ultra-clean production environments, the report indicates that Micron’s Boise, Idaho plant includes a Class 1 cleanroom where only an extremely small amount of dust is permitted. SK Hynix plans to build two new memory factories in South Korea at a total cost of $38.1 billion, with production scheduled to begin, according to current plans, no earlier than June 2029.

Why do companies prefer HBM production?

An economic factor exacerbates the crisis: HBM memory is more profitable than traditional DRAM. As artificial intelligence companies and major data centers prepare to pay high prices for available supplies, manufacturers are prioritizing memory dedicated to AI workloads. At the end of 2025, Micron decided to exit its Crucial consumer business, citing a desire to improve supply and support for its strategic customers in the fastest-growing sectors.

Memory companies reap huge profits

The financial results show how manufacturers are benefiting from the surge in demand. Micron recorded quarterly revenues of around $28.86 billion and a gross profit margin of 84.9% in June, while SK Hynix announced record revenues in July with an increase in operating profits by 557% year-on-year. Samsung's memory sector also recorded record levels of revenue and profits.

When will RAM prices drop?

There is no definite date yet. The report suggests that prices could fall sharply if the artificial intelligence race slows down or if the investment bubble in this sector bursts before the expected time. However, if current demand continues, the market may remain tight for years. Current forecasts indicate that a return to normal price levels may not occur until 2029 or 2030, and pressures could continue after that depending on the speed of growth in demand for artificial intelligence data centers.