The RAM you can buy today doesn't reflect what manufacturers are actually paying. Valve engineers working on the Steam Deck told Bloomberg this week that memory shortages are "getting worse," and retail pricing lags the manufacturing reality by three to six months. Translation: the price increases haven't hit consumers yet, but they're already baked in.
Key Takeaways
- Valve's hardware team says RAM shortages are worsening and retail pricing lags manufacturing costs by 3-6 months
- AI infrastructure demand is contributing to the shortage
- Future Steam Deck production runs will face higher component costs as retail pricing catches up
What Valve's Engineers Said
Valve hardware engineers Yazan Aldehayyat and Pierre-Loup Griffais confirmed in the Bloomberg interview that memory shortages are accelerating. Current retail stock, they said, is "lagging" by approximately three to six months — meaning the prices you see on Newegg or Amazon today don't reflect what it costs to actually source components right now.
"We knew there was going to be an issue with sourcing," Aldehayyat said, as reported by Kotaku. The team expects memory prices to climb further as retail inventories deplete and new stock enters at higher procurement costs.
The warning comes as Valve prepares for future production runs of the Steam Deck. The company acknowledged that units manufactured in the coming months will cost more to produce due to rising memory component pricing, though it hasn't said whether that increase will be passed to buyers.
Why the Lag Exists — and What It Hides
Here's the supply chain detail most coverage skips: manufacturers lock in component orders months before products reach retail. When memory prices rise at the procurement level, that cost doesn't show up in consumer pricing until existing inventory clears out and replacement stock arrives at the new higher cost. Valve's engineers are saying that clearing is about to happen.
The article cites the "AI-pocalypse-fueled RAM shortage" as part of the cause. That's shorthand for a real allocation shift: data center operators and cloud providers procuring memory for machine learning workloads are competing with consumer hardware manufacturers for the same supply. High-margin data center contracts get priority. Gaming devices and consumer PCs get what's left.
This isn't speculation. It's the same dynamic that appeared in Meta's infrastructure buildout, which requires massive memory procurement to support AI training workloads. When hyperscalers order at scale, smaller manufacturers negotiate from a weaker position.
What Valve Didn't Disclose
The company has not said how much memory costs have increased for its manufacturing partners, nor has it specified which memory types — DDR4, DDR5, LPDDR5 — are experiencing the most acute shortages. Available reports do not quantify the total demand increase driven by AI infrastructure or specify how long the supply-demand imbalance is expected to persist.
Valve's engineers offered a timeline for when retail pricing will reflect current manufacturing costs, but they did not provide a forecast for when supply constraints might ease.
What to Watch
If Valve's timeline holds, retail RAM pricing should increase within the next three to six months as current inventory depletes and new stock enters at higher cost. Component distributors and system builders will likely adjust pricing during that window.
The next earnings calls from DRAM and NAND manufacturers — Micron, SK Hynix, Samsung — will clarify capacity allocation priorities and production ramp timelines. Changes in allocation toward consumer markets or announcements of new fab capacity would signal potential relief. Until then, anyone planning a build or procurement should assume prices are moving up, not stabilizing.
Why It Matters
Memory pricing directly affects infrastructure costs for AI development, cloud computing, and data center expansion. Valve's warning confirms that current retail prices do not reflect the supply reality manufacturers already face. Organizations planning hardware procurement should anticipate higher costs in the near term, particularly for memory-intensive workloads. This shortage reflects a structural shift in memory demand driven by AI infrastructure competition.