RAM Unit Prices Reverse to 2007 Levels: AI Demand Erases 20 Years of Progress

AFFILIATE_PRODUCTS:

20 Years of Price Declines Erased in a Few Months

Historical computer memory price comparison data published by Daniel Lemire—a software performance researcher, scientist, and GitHub developer—is sending shockwaves through the technology industry. Memory prices, which had been falling exponentially for decades, have had 20 years of progress wiped out in just a few months due to the surge in AI demand.

“The unit price of RAM is at almost the same level as 2007,” Lemire said. “I don’t recall any similar case before. This is a historical anomaly” (from a report by Jowi Morales of Tom’s Hardware).

Latest DDR5 Prices and Historical Comparison

According to third-party data confirmed by Tom’s Hardware, the current price per 1GB of DDR5 memory ranges from $13.28 to $11.41. The last time memory was traded in this price range was 2008, when DDR2 was approximately $15 to $11 per GB.

In inflation-adjusted 2024 dollars, current DDR5 prices range from $12.74 to $10.94, comparable to 2011 DDR3 prices ($11.85 per GB). This is supported by price data published by the Stanford DAM Project (data compiled by David Shim).

First Cost Reversal in Technology History

Over the roughly 80 years since the advent of digital computers, technology has advanced exponentially and hardware costs have consistently declined. The ENIAC in 1946 was developed with a government investment of $400,000, equivalent to approximately $6.85 million in today’s inflation-adjusted terms. Its computing power at the time was 5,000 addition operations per second.

Today, smartphones like the Moto G Play, available for under $100, are equipped with a Snapdragon 680 processor capable of 3.3 trillion operations per second (3.3 TOPS). While directly comparing the ENIAC with modern processors is difficult, it is sufficient to illustrate the scale of technological progress and cost reduction.

HBM Demand Driving Prices

The direct cause of this price surge is not a technological regression or a supply bottleneck. It is the enormous demand for HBM accompanying the AI race. Memory manufacturers are expanding supply, but the capacity consumed by AI infrastructure is outpacing it.

Lemire pointed to this breakdown in the supply-demand balance as follows:

“Memory production is increasing by about 20% per year. Normally, that is an astonishing growth rate for a large, mature industry. But consider that demand is also growing at 20% per year.”

Long-Term Outlook and Technical Challenges

Lemire suggests two possible paths to break through this situation. One is building AI systems that do not require large amounts of memory. The other is technological innovation that produces larger and faster memory devices.

At present, neither has been implemented as a concrete solution. It is certain that memory prices will remain elevated and continue to impact both AI development and consumer computing.

Editorial Opinion

In the short term, the surge in HBM demand will continue to spill over into consumer DRAM. Rising memory costs for smartphones and PCs will accelerate pass-through to product prices, and as symbolized by the confirmed price increase of the Google Pixel 11, device makers will continue to be forced to fundamentally change their pricing strategies. The price divergence between AI-oriented and general-purpose memory could widen further over the next 3 to 6 months.

In the long term, structurally high memory prices will raise the barrier to entry for AI development. As training costs for large-scale models swell, cloud concentration of computing resources will advance further, and there is a risk that a structure in which AI development is consolidated among a few hyperscalers becomes entrenched. The fact that semiconductor manufacturing innovation nullified 20 years of price declines in just a few months suggests that memory efficiency will become one of the most important design parameters in future chip design.

The fundamental question raised by this event is whether supply-side technological innovation alone can suppress prices, given the premise that AI demand continues to grow at more than 20% annually.

References

• ” Scientist says RAM pricing has reverted to normalized 2007 levels — memory prices have been falling exponentially for decades, but the AI shortage undid 20 years of progress in a matter of months ”, by Jowi Morales — Tom’s Hardware, 2026-08-07T16:58:22.000Z (ARR)
• Source URL: https://www.tomshardware.com/pc-components/ram/scientist-says-ram-pricing-has-reverted-to-normalized-2007-levels-memory-prices-have-been-falling-exponentially-for-decades-but-the-ai-shortage-undid-20-years-of-progress-in-a-matter-of-months