Your RAM just got cheaper by about $40 on a kit that jumped from $87 to $370. Celebrations are not really in order. The global DRAM shortage that turned a routine PC upgrade into a financial decision has been strangling hardware budgets since late 2025. AI data centers gobbled up the world’s memory supply, prices tripled, and consumer PC buyers could do nothing but watch helplessly. Now, a handful of things have happened in quick succession that are nudging prices and chip manufacturer stocks back down, even if it’s ever so slightly. Here’s what’s actually going on, what it means for RAM and SSD costs, and whether any of this should change your upgrade plans.
Also read: Google’s TurboQuant explained: The JPEG approach to AI compression
On March 25, Google published TurboQuant, a compression algorithm that reduces how much memory large language models consume during inference, specifically targeting what’s called the KV cache – the part of a model’s memory that stores previous calculations so it doesn’t have to rerun them.
Wall Street didn’t wait around to think it through. Between March 19 and March 31, Micron’s stock went into freefall, bottoming out around the $318 range after peaking at a 52-week high of $471.34 which is nearly a 30% correction in under two weeks. The irony is not lost on me. Just before the crash, Micron had posted record revenue of $23.86 billion, nearly tripling its year on year performance. SK Hynix and Samsung also fell 6% and nearly 5% respectively on the Korea Exchange. Japanese flash memory company Kioxia dropped nearly 6%.
Analysts were quick to cool the panic. Ben Barringer, head of technology research at Quilter Cheviot, called the Google breakthrough “evolutionary, not revolutionary,” adding that it does not alter the industry’s long-term demand picture. TurboQuant only touches inference, not training. It doesn’t reduce overall DRAM demand.
Also read: Why memory prices are rising in 2026 and why it affects every device you buy
Partly, and it fits a pattern that is worth understanding. OpenAI was expected to purchase roughly 40% of the world’s DRAM wafer supply from Samsung and SK Hynix, but those deals were letters of intent rather than binding contracts and ended up not materialising. When that anticipated demand disappeared from order books, spot prices reacted.
This is the dirty secret of the AI infrastructure boom, a significant chunk of the demand pressure driving consumer RAM prices to the moon was built on projections and commitments rather than locked-in contracts. When those commitments shift, and they often do, memory markets feel it immediately.
Oracle and OpenAI scrapped plans to expand a flagship AI data center in Abilene, Texas, after negotiations stalled over financing and OpenAI’s changing infrastructure needs. OpenAI walked away because it wants clusters equipped with next-generation Nvidia chips at new sites rather than expanding existing capacity running older processors.
The broader Oracle-OpenAI deal covering 4.5 gigawatts of data center capacity across multiple locations remains on track. But the Abilene is more significant than a single cancelled expansion. OpenAI is becoming more selective about where it deploys infrastructure, and that selectivity, even in isolated decisions, is enough to shift global memory procurement expectations.
The Stargate pullback, the Samsung deal falling through, TurboQuant – none of these individually break the DRAM market. Together, they’ve added the first real question mark in AI’s relentless appetite for memory since the shortage began.
DDR5 spot prices dropped 22% over four days following TurboQuant’s release, with some Corsair 32GB DDT5-6400 kits falling from $490 to $380 in a single day. Sounds dramatic? That same Corsair kit had a historical low of $87. We went from crisis pricing to slightly-less-crisis pricing on a product that should cost a fraction of what it does.
The structural problem is unchanged. Micron shut down its consumer brand Crucial in December 2025 after 29 years, fully exiting the consumer memory market to focus on enterprise and AI customers. SK Hynix’s HBM capacity is fully sold out through the end of 2026. Samsung, SK Hynix, and Micron control over 90% of global RAM production, and building High Bandwidth Memory for Nvidia’s data center GPUs is roughly ten times more profitable per wafer than making DDR5 for our gaming rig. They are not coming back to the consumer market as an act of goodwill.
As for SSDs, I don’t expect much relief there either. HDDs and SSDs are actually getting more expensive even as RAM shows early signs of softening, with NAND flash caught in the same AI-driven supply squeeze.
On the question of what happens next, prices could drift further down if AI capex keeps cooling and more infrastructure commitments fall apart. They could plateau here if procurement stabilises. Or they could climb again the moment the next round of data center contracts lands. Realistically, I don’t see a return to the 2024 pricing anytime soon. Analysts including Gartner, IDC, and TrendForce put the earliest window for meaningful relief at late 2027, with some estimates stretching it all the way to 2028. If you can wait, wait. If you can’t, buy only what you need and revisit when the market actually recovers. I don’t think this is a recovery, it’s more of a wobble.
Also read: UWMS explained: Artemis II astronauts will use a $23 million toilet in space