Valve launched the Steam Machine this past June with a frustrating surprise: a price point of $1,049, or roughly $300 above their original target. When discussing the price increase, the company blamed it on one factor. Rising memory prices.
The memory shortage was so bad that Valve rationed launch units through a randomized reservation queue rather than open sales.
The memory chip shortage has become the defining consumer tech story of 2026 so far. We all know why it’s happening, but let’s dive into the details to separate the facts from the overblown headlines.
Where the memory went
Three companies: Samsung, SK Hynix, and Micron, control more than 95 per cent of global DRAM production. Within the last year, all three have shifted their priorities to specialized stacked DRAM for AI accelerators. These chips, called high-bandwidth memory (HBM), earn three to five times the revenue per wafer compared with standard memory wafers for consumer devices. So, as long as the dependence persists, the companies have strong incentive to focus on HBM almost exclusively.
This has pushed up prices very quickly. TrendForce reported on June 1 that conventional DRAM contract prices climbed 93 to 98 per cent in the first quarter of 2026. In particular, we are seeing an estimated $725 billion in 2026 capital spending by the biggest four cloud service providers, according to Deloitte. The majority of this is for AI infrastructure.
When the bill hits
Unfortunately, these trends are set to get worse before they get better. Micron made its new priorities on AI chips very explicit in December, where they announced they would retire Crucial, its 29-year-old consumer memory and SSD brand.
We are now starting to see the repercussions in consumer electronics. Sony’s PS5 Digital Edition now sells for $599.99 in the US, $200 above its initial launch price. Microsoft moved the Xbox Series X to $650, and we’ve already seen how Valve’s Steam Machine prices have moved, and their Steam Deck already jumped more than half. IDC expects the global PC market to shrink 4.9 per cent this year and smartphone shipments to fall about 2 per cent as these trends continue.
When is relief coming?
Not soon. A new fab takes 18 to 24 months to build, plus more time to reach usable yields. Micron’s newest plant won’t produce meaningful volume until roughly the third quarter of 2028, and Deloitte’s July analysis suggests the crunch may not ease until 2029.
On the demand side, most of the big AI companies are building out more datacenters and buying up the newest chips for them, with consistently growing demand all of 2026. TrendForce has revised its 2026 AI server forecast upward twice this year, most recently to 31 per cent growth.
Unfortunately this leaves the best advice to consumers as acting sooner rather than later. Prices will likely continue to rise for the next three years as shortages continue or get worse. It is possible that with an AI bubble pop prices could come down within six months to a year, but no trendline is showing that currently.
