{"slug": "the-memory-shortage-that-no-one-saw-coming", "title": "The “memory shortage” that no one saw coming", "summary": "A global memory supply shortage driven by surging AI demand caught nearly all experts off guard, according to a post on Grey Enlightenment. The shortage, which began in 2025 and continues into 2026, has driven memory stocks like Micron, SanDisk, and Seagate sharply higher while Nvidia stock stalled, yet only a handful of non-AI analysts predicted the bottleneck. The author notes that predicting second-order effects of technology is inherently difficult, comparing the surprise to pandemic-era supply chain disruptions.", "body_md": "As a follow-up to my last post, “[The AI Experts Who Couldn’t Predict AI](https://greyenlightenment.com/2026/07/26/the-ai-experts-who-couldnt-predict-ai/),” I want to expound on the “memory shortage” of 2025-2026 and how no one saw it coming, even experts in AI.\n\nA seemingly sudden demand for memory due to AI has led to shortages for personal computers and other applications. This has garnered widespread media attention, including even a Wikipedia article, “[2025–present global memory supply shortage.](https://en.wikipedia.org/wiki/2025%E2%80%93present_global_memory_supply_shortage)” Same for considerable discussion on Reddit, with new threads being posted daily, such as from annoyed gamers who now find memory in short supply:\n\nSearch interest also swelled in 2025, according to Google Trends for “memory shortage:”\n\nMemory stocks such as SanDisk, Micron, and Seagate, which had been lagging Nvidia stock from 2022-2025, suddenly rocketed, all while Nvidia stock did nothing:\n\nMy question is, how did everyone miss this? From 2022-2024 there was all this demand for GPUs to train the models. The bottleneck was computing power, not storage, and it was assumed it would stay that way. Then in mid-to-late 2025 there was suddenly a “memory shortage” as a sort of follow-on effect or a second-order consequence of the “AI boom.”\n\nNo one saw the shortage, including many AI experts. Outside of a handful of analysts and investors, almost no one anticipated that AI demand would produce a memory shortage. No one made a connection between “surging GPU demand in 2023” and “memory shortages two years later.” No one was saying in 2024, “Buy Micron stock to profit from the impending bottleneck in memory.” It just sorta happened all at once–suddenly there was a shortage.\n\nIt had become almost like a “meme” in its suddenness and unpredictability, like how Covid would lead to supply chain disruptions and shortages 2 years later, which no one predicted in 2020-2021. Again, the leading tech/AI bloggers, podcasters, and pundits on Twitter failed to see the memory shortage, so don’t feel bad if you didn’t. You are in good company with people who have built careers from speculating about technology, like Drawkesh Patel.\n\nBut a few people did. For example, according to a search of Twitter posts from 2024 to early 2025 for keywords pertaining to memory and AI, I only found three (yes, three) individuals who made the connection between GPU demand, AI, and memory shortages. None of them are involved in AI. They aren’t names anyone would be familiar with. They don’t have large brands or popular podcasts.\n\nApril 2024:\n\nThe demand for memory is inextricably linked to demand for AI and AI chips.\n\nSK Hynix, Micron, Samsung all set up very well here.\n\nMemory names and stocks will heat up. Demand gains will lead to pricing power,\n\nmargin expansion and eps growth.It has been a rough few years for…\n\n[https://t.co/xcBIzaxHPp]— Daniel Newman (@danielnewmanUV)\n\n[April 25, 2024]\n\nMarch 2024:\n\n[$MU][$NVDA]In the wake of Nvidia’s launch of powerful new artificial intelligence chips, Goldman Sachs is predicting major growth for memory chips used in AI systems. They see three major memory makers as prime beneficiaries of the booming HBM market: SK Hynix, Samsung Electronics…— The Inner Circle Trading Group DP David Prince (@epictrades1)\n\n[March 29, 2024]\n\nFebruary 2025:\n\n[$MU]Can AI Demand Send This to $150?Technical Setup: Micron has been trading in a range between 85-112 since the end of July. If we can get thru 112 we can see a move to 150!\n\nEarnings: Reported record breaking Q1’25 revenue of $8.71 billion. Same period last year revenue was…\n\n[pic.twitter.com/dUodooevDY]— EliteOptionsTrader (@EliteOptions2)\n\n[February 20, 2025]\n\nSo why is this? Predicting follow-on effects from a technology is really hard. In some sense, it’s easier to predict the emergence of a technology than its downstream consequences.\n\nDeep learning and machine learning were a big deal from 2015-2020. Then from 2020-2021 there was the Metaverse, as a sort of transition between the rise of OpenAI and the denouement of the “machine learning era”. VR was something that would finally be taken seriously after so many earlier attempts had failed, although Meta would continue in that tradition, with [$80 billion](https://finance.yahoo.com/sectors/technology/articles/mark-zuckerbergs-80-billion-metaverse-000000105.html) of losses as of 2026, until shutting most of it down.\n\nThen–boom–LLMs and chatbots in 2022-2023. Machine learning never went away, but, AFIK, no one made the conceptual leap from applications such as self-driving cars in 2021, which relied on computer vision and other machine learning applications, to chatbots that could disprove long-standing math conjectures by 2026. Even in 2023 when LLMs were mainstream, the main objection was the technology devaluing artists, not that it would threaten to make mathematicians or coders obsolete. Again, almost no one made that leap.\n\nThis was typified by the viral May 2022 essay, “[AI-art isn’t art](https://www.theintrinsicperspective.com/p/ai-art-isnt-art)” by blogger Erik Hoel, just as applications such as DALL-E were becoming popular. On the opposite side of the debate, from * The Verge* optimistically proclaimed in Jun 2022, “How DALL-E could power a creative revolution.” At this time, math and coding were considered “safe” from the technology, as the focus was on writing and the arts.\n\nEven when given ample opportunity, no one made the leap or connection from “art and writing” capabilities to “math and coding”. Like the memory shortage, it all sorta happened all at once. Now, as of 2026, AI writing ability has stalled out. It’s still clunky and (mostly) obvious, as it was in 2023. Same for AI art: artists have adapted or moved on. The world didn’t end, although it did spawn endless “think pieces” about creativity and AI. Now with AI supplanting even mathematicians, the debate has taken a more serious tone.\n\nThe iPhone for example was revolutionary, not because of its design as often assumed, but as a follow-on effect, the App Store, which launched a year later on July 10, 2008. That was the actual revolution and is why the iPhone became such a huge success and cultural phenomenon. This had far-reaching secondary effects, such as possibly leading to [lower](https://davidoks.blog/p/why-the-atm-didnt-kill-bank-teller) employment of bank tellers, because everyone was using their phones for banking.", "url": "https://wpnews.pro/news/the-memory-shortage-that-no-one-saw-coming", "canonical_source": "https://greyenlightenment.com/2026/07/28/the-memory-shortage-no-one-saw-coming/", "published_at": "2026-07-28 17:51:10+00:00", "updated_at": "2026-07-28 18:09:12.452155+00:00", "lang": "en", "topics": ["ai-infrastructure", "ai-chips", "ai-policy"], "entities": ["Micron", "SanDisk", "Seagate", "Nvidia", "SK Hynix", "Samsung", "Goldman Sachs", "Daniel Newman"], "alternates": {"html": "https://wpnews.pro/news/the-memory-shortage-that-no-one-saw-coming", "markdown": "https://wpnews.pro/news/the-memory-shortage-that-no-one-saw-coming.md", "text": "https://wpnews.pro/news/the-memory-shortage-that-no-one-saw-coming.txt", "jsonld": "https://wpnews.pro/news/the-memory-shortage-that-no-one-saw-coming.jsonld"}}