Rahul SomvanshiIf you’re planning a PC build or upgrade in the next 12 months, the CEO of OpenAI has an opinion about your RAM — and it isn’t flattering. Sam Altman said on a recent podcast that putting memory in personal computers for local AI is a “quite terrible” use of a resource the world doesn’t have enough of. The statement landed as consumer DDR5 prices are already climbing steeply and industry forecasts point higher.
Speaking on the Cheeky Pint podcast in October 2026, Altman argued that running AI models on personal hardware wastes memory that belongs in cloud data centers. “A lot of people want local inference. And somehow, they assume, that the problem is solved now… You want maximum utilization in the cloud if you’re trying to deal with this very supply constrained environment,” he said. Altman noted that local AI features are currently being used for only about 5% of on-device inference tasks — a figure he cited to argue that local hardware investment is disproportionate to actual usage.
The backdrop is a genuine global DRAM shortage. Major memory manufacturers — Samsung, Micron, and SK Hynix — have shifted fabrication output away from consumer PC parts toward high-margin enterprise server DRAM and High Bandwidth Memory for AI accelerators. As hyperscalers scale up data center capacity, they are consuming the largest share of available wafer production, pushing consumer memory prices sharply higher.
If you plan to build or upgrade a PC in the next 12 months, expect to pay steep premiums on memory. As hyperscalers lock up global DRAM wafer capacity for cloud data centers, memory makers are shrinking consumer product lines. With prices rising sharply and industry forecasts pointing higher, buying hardware for local AI or gaming is becoming substantially more expensive.
What the DRAM crunch means for PC builders and local AI users
Altman’s argument is not about energy efficiency — it is about silicon supply. He frames cloud centralization as the correct response to a physically scarce resource: there is only so much memory fabrication capacity on Earth, and he believes cloud utilization squeezes more performance per wafer than distributing chips across millions of individual PCs.
The argument has direct consequences for privacy-conscious developers running open-source AI models locally and for PC builders who relied on affordable RAM. Micron has reduced its consumer DRAM lineup in recent product cycles to focus on enterprise AI memory contracts, and budget DDR5 options have become harder to find at lower price points.
The shortage is structural, not cyclical. Fabrication capacity for HBM and enterprise server DRAM requires retooling the same fabs that produce consumer DDR5. Until new fab capacity comes online, consumer supply stays constrained by hyperscaler demand.
Why are DDR5 RAM prices rising so fast in 2026?
DDR5 RAM prices are rising in 2026 because AI data centers are consuming the majority of global memory fabrication capacity. Major manufacturers have prioritized high-margin enterprise server DRAM and High Bandwidth Memory over consumer PC parts, creating a supply crunch that has pushed retail prices sharply higher. Industry analysts warn further price increases are likely before year-end.
The DRAM shortage is tied to AI infrastructure build-outs expected to continue through 2027. How much consumer prices ease depends on whether memory manufacturers expand fab capacity faster than hyperscaler demand grows. Follow Karmactive’s tech coverage for updates on memory pricing and AI hardware developments.
