Chey Tae Won’s latest remarks land with the kind of blunt force that only a man staring down a decade of scarcity can deliver. He told Bloomberg that the global memory crunch will drag on until around 2030 because wafer supply is already more than twenty percent behind demand and shows no sign of catching up. That alone would be enough to make the PC industry sweat, but the real story is how thoroughly AI has hijacked the entire semiconductor economy.
Chey’s remarks echo what Gizmodo has been reporting for months. The big three memory makers, Micron, Samsung, and SK Hynix, have all shifted enormous portions of their production toward high-bandwidth memory for AI accelerators. These chips are so profitable and so urgently needed for the likes of Nvidia’s Vera Rubin and Grace Blackwell systems that consumer DRAM has become the neglected middle child of the semiconductor family. A single Vera Rubin chip can swallow up to 288 GB of HBM, which is nine times what a typical gaming PC uses. Multiply that by hundreds of chips per data center rack, and you begin to understand why your next laptop is suddenly flirting with luxury pricing.
Chey confirmed that this imbalance is not going away soon. Even with SK Hynix building a new 13 billion dollar facility, it will not be online until late 2027. Samsung, meanwhile, is reportedly hesitant to scale too aggressively because it fears a future glut if the AI bubble cools. In other words, the companies that could fix the shortage are also the ones most incentivized to keep the spigot only half open.
The shift is not subtle. SK Hynix, Samsung, and Micron have all redirected production toward high-bandwidth memory for AI accelerators, the chips that power Nvidia’s most in‑demand systems. This pivot has created a shortfall in conventional DRAM, the kind that keeps laptops, phones, and game consoles running. Prices are rising across the board, corporate plans are being derailed, and even car manufacturers are feeling the squeeze. It is the kind of industry‑wide distortion you get when one sector becomes so profitable that everything else becomes a rounding error.
I felt that distortion firsthand when I tried to upgrade an older ultrabook. A 32 GB kit that would have cost me a casual impulse buy a few years ago now felt like I was bidding on a vintage synthesizer. A friend who runs a small creative studio told me he has delayed refreshing his team’s desktops because the memory alone would eat his quarterly budget. These are not dramatic stories. They are the new baseline.
Chey’s comments make it clear that this is not a temporary imbalance. Even with SK Hynix building new facilities, the timeline for meaningful relief stretches years into the future. He said it takes four to five years just to secure additional wafers, which means the industry is locked into this trajectory, no matter how loudly consumers complain. Samsung, for its part, is reportedly hesitant to expand too aggressively because it fears a future glut if AI demand cools. That creates a strange stalemate where the companies that could fix the shortage are also the ones most incentivized to let it ride.
The deeper problem is that AI has become a gravitational force pulling in capital, electricity, engineering talent, and now memory at a pace no other sector can match. The industry likes to talk about AI as a rising tide that lifts all boats, but right now it looks more like a riptide dragging everything else out to sea. The profits from supplying memory to AI accelerators dwarf what companies can earn from consumer DRAM, so the market has effectively decided that your next laptop can wait.
This is how you end up with a world where data centers get fat on terabytes of HBM while consumers are told to be grateful for eight gigabytes in a midrange phone. It is how you end up with PC makers quietly raising prices, smartphone manufacturers trimming RAM configurations, and hobbyist boards creeping into luxury territory. And it is how you end up with Chey Tae Won calmly predicting that the crunch will persist until 2030 while the rest of the industry pretends this is all just a temporary inconvenience.
The uncomfortable truth is that AI is not just consuming attention and investment. It is consuming the physical components that make the rest of the tech world function. The RAM crisis is not a supply chain hiccup. It is a resource reallocation event, and unless something dramatic changes, the next four years are going to feel like a long wait for a train that keeps getting delayed because someone keeps rerouting the tracks toward a shinier destination.
