What if the RAM/GPU shortage is actually deliberate? — 테오 is onlineSkip to main contentFirst of all, a necessary disclaimer: I’m not a conspiracy theorist. Period.<br>Now that we have that out of the way, let’s talk about a theory that struck me this morning.<br>What if the shortage is deliberate?<br>What if the RAM and GPU shortage is actually deliberate?<br>Deliberate by whom, you may ask? By the companies that depend on us subscribing to their services: the big AI companies.<br>But first, let’s look at some events that took place in 2025.<br>The rise of Chinese AI models<br>In January 2025, Alibaba released Qwen2.5, which shocked the tech industry with its performance. During the same month, DeepSeek-R1 was also released. Meanwhile, Moonshot AI was working on its own model, Kimi, which showed promising results.<br>All of the above(but especially DeepSeek-R1)caused havoc in the AI industry. Nearly $1 trillion was wiped from the tech market’s capitalization.<br>However, it wasn’t only the release of these models that caused concern. Chinese AI labs also released them under permissive licences, which meant that anyone could download and use them for free.<br>That alone was enough to cause panic in the AI industry, since its business model was suddenly under threat. Why pay for a subscription when you can run a free model with similar performance?<br>Then came the hardware shortage<br>By early 2026, RAM and GPU shortages were starting to be felt, although not nearly as much as by mid-2026.<br>For example, on 3 June 2026, I bought two 16 GB DDR4 RAM modules for €26.45 each. Currently, the same RAM is selling for €235.90 from the same online shop.<br>The shortages and price increases were reportedly caused by AI companies buying up as much RAM and as many GPUs as they could get their hands on.<br>Proper hoarding.<br>One could argue that this buying spree was necessary to keep up with the growing demand for AI services. But what if it also caused a convenient, accidental(or perhaps not entirely accidental)side effect?<br>If you, as a consumer or hobbyist, want to run a model such as Qwen2.5 or DeepSeek-R1 locally, you need a powerful computer with plenty of RAM and one or more capable GPUs.<br>If you can achieve a good number of tokens per second and get decent results locally, why would you continue paying for a subscription?<br>A convenient side effect?<br>If you were running one of those AI companies, how would you prevent that?<br>You would make sure that the RAM and GPU shortage continued, making it increasingly difficult and expensive for hobbyists and ordinary consumers to run these models themselves.<br>Simple as that.<br>Of course, the truth may be somewhere in between.<br>We have a saying in my country: kill two birds with one stone .<br>With one massive hardware-buying campaign, AI companies may have managed to accomplish two things at once: build enormous data centres for their services while also keeping RAM and GPUs scarce and expensive for everyone else.<br>Protection from Chinese competition<br>Recent reports also suggest that the US administration is considering restricting the use of Chinese AI models, including Kimi.<br>You can read more in this Axios article: US weighs restrictions on Chinese open-source AI models.<br>This is clearly a move intended to protect American AI companies, but it also suggests that they are genuinely worried about the competition.<br>In the end, we the users, may get the short end of the stick.<br>Hardware that was once widely available and relatively affordable has become scarce and extremely expensive. For now, there are few signs that prices or availability will return to normal anytime soon.<br>Let&rsquo;s hope for the best.