NAND memory prices to rise 200 per cent – Samsung and AI are changing the market

While the industry's attention was focused on graphics processors, the real bottleneck of the AI revolution emerged in data storage, forcing Samsung to double NAND memory prices again this quarter. This 200 percent year-on-year cost increase means that the giants building infrastructure for language models are beginning to displace traditional electronics manufacturers from the component market.

3 Min Read
memory market

Samsung is sending a clear signal: the era of cheap data space is coming to an end. According to recent reports, the chip giant plans to double the price of NAND dice in the second quarter of 2026. This is another such drastic increase this year, meaning that on a scale of just twelve months, these components will become 200 per cent more expensive. For the tech sector, this is a jolt that will change the mathematics of manufacturing everything from smartphones to servers.

The reason for this phenomenon lies in the architecture of modern artificial intelligence. Until now, investor attention has been focused on HBM memory, the ultra-fast RAM required for GPUs. However, powerful AI systems, such as the Llama models from Meta, need not only to ‘think’ fast, but also to retrieve data instantly. Classic hard drives have proved too slow to keep up with the speed of modern computing clusters. As a result, technology giants have begun buying up NAND resources en masse to build SSD arrays capable of powering AI infrastructure.

The memory market is a de facto oligopoly, with Samsung and SK Hynix setting the tone. When one player raises prices, the rest usually follow suit, consuming the margins of the smaller players. Laptop and phone manufacturers will face a difficult choice: either pass the costs on to end customers or apply digital ‘belt-tightening’ by offering devices with smaller disk capacities as standard.

However, the biggest victim of this change will not be Silicon Valley corporations, which can afford any price in the AI arms race. The SME sector will be hit hardest. While Meta or Google are securing supplies for their data centres, local companies building their own workstations or server rooms will pay twice as much for the same hardware as they did a year ago. In 2026, NAND memory has gone from being a cheap bulk commodity to a strategic resource whose price now reflects the ambitions of the world’s richest players.

Share This Article