The price of memory has turned back the clock
RAM prices are now as high as they were in 2007 due to AI demand. A software performance expert named Daniel Lemire studied historical RAM pricing. He found that current memory prices are roughly the same as they were in 2007. The sharp rise in cost is not due to manufacturing limits. It is driven by a huge surge in demand for a specific type of memory used in AI systems.
Lemire described the situation as a major anomaly in tech history. In a social media post, he said he could not think of another time when hardware prices returned to levels that were last seen in the 2000s. "It is impossible to predict what will happen," but he believes we need to either find ways to use less memory in AI systems or develop faster and more efficient memory solutions.
AI is driving the memory crisis
Data collected for the Stanford DAM Project shows that DDR5 memory, a high-performance type of RAM, now costs between $13.28 and $11.41 per GB. When adjusted for inflation, the real price ranges from $12.74 to $10.94, which matches what DDR3 memory cost in 2011. The long-term trend of falling memory costs has come to a halt — or even reversed — as the AI industry's demand for high-bandwidth memory (HBM) has exploded.
Elon Musk, during a recent earnings call for SpaceX, pointed out that demand for HBM is rising by 200% each year. That is 10 times faster than the industry’s ability to increase memory output. Industry output is growing by about 20% per year. The chairman of the SK Group, a major holding company that owns SK hynix, a top memory chip manufacturer, also confirmed the issue. He called RAM prices "abnormally high" and urged the industry to ramp up production to bring costs down.
Broader consequences of the memory shortage
The spike in memory prices is not just an issue for big data centers. It is affecting a wide range of consumer devices. For example, smartphones, gaming consoles, and even modern cars are being impacted by the shortage. Even memory modules built with chips from Chinese manufacturer CXMT, which many users choose as a cheaper option, have seen their prices go up. Module prices are rising alongside those using chips from the major companies like Micron, Samsung, and SK hynix.
Experts are warning that the current memory shortage situation is not sustainable. No one is sure exactly when or how the market will eventually balance out, but it is clear that something will have to change before the problem worsens. The challenge is that no one can predict what that change might look like or when it will happen.
To understand the significance of this problem, consider how much technology has advanced in just a few decades. Back in 1946, the U.S. government paid $400,000 for a machine called ENIAC, which could perform about 5,000 additions per second. By today's standards, that is incredibly slow. In contrast, one of the cheapest smartphones available today — the Moto G Play — costs just $99.99 and is powered by a Snapdragon 680 processor with an AI performance of 3.3 trillion operations per second. This stark comparison shows how far technology has come and how much it has become more affordable.
However, the recent increase in memory prices does not reflect a decline in technology or a drop in manufacturing efficiency. Instead, it highlights how rapidly the AI industry is developing and how much memory it requires. The race to build better AI systems is pushing the demand for memory to levels that manufacturers can barely keep up with. This is creating a mismatch between demand and supply.
The long-term effects of this shortage are still unknown. For now, it is clear that the cost of memory is at a critical point, and the situation will likely have lasting impacts across the tech industry. Whether the industry finds ways to use memory more efficiently or develops new technologies to meet demand, one thing is certain: the way we use and build computers will change.