The boom in artificial intelligence has profoundly disrupted the balance of the RAM market, leading to a shortage of inventory and soaring costs that are now weighing on consumer hardware (and beyond). Amid this tense situation, Sam Altman, the CEO of OpenAI, recently expressed a strong opinion on resource allocation. During an interview aired on the YouTube channel Cheeky Pint, he argued that it would be more efficient to direct the global supply of memory toward AI-dedicated data centers rather than storing it in local devices.
At around the 29-minute mark of his remarks, Altman described the situation as “terrible” when DRAM is locked inside consumer hardware. According to his analysis, these devices only utilize memory for about 5% of their operating time. In contrast, data center infrastructure operates continuously, thereby optimizing the use of the overall resource and driving demand toward greater utilization.
This remark specifically targeted laptops with local AI capabilities, such as Microsoft’s Surface Laptop Ultra, which can accommodate up to 128 GB of unified memory. Altman explicitly mentioned these machines designed to run open AI models such as DeepSeek, which he considers direct competition for OpenAI. His argument suggests that having this memory on the user’s device is a waste compared to using it in the cloud.
However, this statement has sparked heated controversy among users and industry analysts. Many observers interpret these remarks as an attack on hardware ownership, fearing that AI giants may seek to turn owners into subscribers to cloud services. This prospect is part of a challenging economic reality in which the massive construction of data centers has caused the cost of DRAM to increase fivefold since last year.
The consequences are already being felt in the consumer market. Hardware prices are reaching levels that are unaffordable for many consumers (€500 for a 32 GB DDR5 kit). Video game consoles are now approaching the $1,000 mark, which is holding back sales. The IT sector is no exception, with a significant decline in computer shipments this year, contributing to the sharpest historical drop in Intel’s stock price.
The shortage is also affecting high-end components. The Nvidia RTX 5080 and RTX 5090 graphics cards have disappeared from NVIDIA’s shelves in the United States after reaching astronomical prices. Even the photography industry has not been spared, with Canon warning that memory shortages could make cameras more expensive. Faced with this situation, memory manufacturers differ on how long the crisis will last.
SK Hynix believes current conditions will persist beyond 2030, while Micron argues that demand could remain strong due to humanoid robots, even if the AI bubble bursts. For his part, Acer’s CEO accuses memory manufacturers of trying to artificially prolong this situation to maximize their profits. He anticipates a drop in prices as early as next year, offering hope of relief for buyers of computer hardware.