The statement
Sam Altman, CEO of OpenAI, sparked controversy by stating that it is "quite terrible" from a global resource allocation perspective to have DRAM memory distributed across personal computers when it could be concentrated in data centers. The comments were made during his appearance on the _Cheeky Pint_ podcast, hosted by Stripe co-founder John Collison, and published on October 29, 2026.
This issue is not new — the memory shortage affecting both AI infrastructure and consumers has been the subject of multiple reports since early 2026. What changed was the directness of Altman's statement: instead of using corporate language, he explicitly called the current situation "quite terrible" and argued that as much RAM as possible should be redirected to the cloud.
What he said
Altman's specific comments were clear. He stated that many users want to run local AI inference, assuming that open-source models on their computers would be sufficient to reduce centralized infrastructure demand. "But from a global allocation perspective, actually, it's quite terrible to have everybody have a bunch of the world's limited supply of DRAM sitting in their own computer and being used 5 percent of the time," he said.
Altman estimated that personal computers use their memory only 5 percent of the time — a figure he provided without supporting data, but which illustrated his point: idle memory in home machines would be wasteful when there is insatiable data center demand. "You want maximum utilization in the cloud if you're trying to use resources efficiently," he argued.
When asked about solutions to the memory crisis, his response was even blunter: "Bad." He added that the market is not solving the problem adequately, citing the failure of the "invisible hand" to allocate scarce memory optimally.
The global shortage context
Altman's comments did not come in a vacuum. The DRAM shortage is a documented fact affecting the entire technology chain. According to J.P. Morgan Global Research, DRAM prices have risen more than 400 percent between early 2024 and late 2026, driven by exponential demand from AI data centers.
Micron, one of the world's largest memory suppliers, declared that over 75 percent of its 2027 production is already committed to long-term contracts with infrastructure companies. The company also announced it would stop shipping its Crucial brand of consumer DRAM by the end of 2026, redirecting all production to "larger strategic customers in faster-growing segments" — that is, AI companies.
Intel, meanwhile, saw its shares plummet nearly 30 percent in a single day — the worst collapse in 52 years — partially attributed to the PC market weakening due to the unaffordability of components. Intel CEO Lip-Bu Tan stated in February that there would be no relief for the memory shortage until 2028.
The shelf effects are already visible: used PlayStation 5 Pro consoles reached prices of $1,399 following official price increases in March; Nvidia's RTX 5080 disappeared from the market, and the RTX 5090 faced near-discontinuation. Apple has already raised prices on several Macs and iPads due to memory costs.
The ecosystem contradiction
Altman's argument runs into an obvious ecosystem contradiction: companies like Nvidia and Microsoft, partners of OpenAI, have been actively promoting local AI capabilities on next-generation PC hardware, launching laptops like the RTX Spark machines with high memory capacity. At the same time, Altman is saying that memory in these computers is wasteful.
This tension reveals the strategic fragmentation of the AI industry. On one side is the centralized infrastructure vision — closed models running in massive data centers, with scaled consumption of memory, energy, and capital. On the other is the open-source model and local inference movement, which has gained momentum with Chinese models like those from the Llama, Qwen, and DeepSeek families.
The emergence of open-weight models has been a catalyst for the local AI trend, challenging exclusive reliance on expensive cloud services. This is not merely a technical question — it is a dispute over who controls artificial intelligence infrastructure. If users can run models locally, the subscription-based cloud model of OpenAI loses part of its perceived value.
What remains
Altman's statement exposed something the market has been feeling for months: the memory shortage is not ending anytime soon. With circuit board manufacturers fully booked through the end of 2030 and AI demand growing exponentially, the scenario of elevated prices for consumers and businesses appears entrenched for years to come. The question is no longer whether memory will become affordable again, but how far the high cost might push users away from the traditional hardware ecosystem.
Sources: The Verge, Tom's Hardware, The Outpost AI
✓ Independent sources cross-checked and verified before publishing
