RAM crisis could last until 2030 — and then get ready for the robots

A RAM supply shortage means memory manufacturers cannot produce enough chips to match what customers want to buy. Demand exceeds supply, which can make memory harder to obtain and contribute to higher component prices. SK Hynix says this imbalance could last through 2030.

The company points to two main causes. Leading-edge manufacturing processes are technically difficult, and new production facilities require a long lead time. That means chipmakers cannot quickly add enough capacity when demand rises. SK Hynix’s Q3 2026 statement described this as a continuing structure in which demand exceeds supply.

The shortage is being supported by strong enterprise and AI demand. SK Hynix also expects longer-term demand from physical AI, including autonomous driving and robotics. Its forecast is therefore not only about current data centers. It includes several expanding technology markets that could keep memory demand ahead of production.

The article highlights a potentially huge memory requirement for future physical AI. Micron estimates that humanoid robots could need 200GB or more of RAM, plus multiple terabytes of storage in each unit. That is the clearest scale mentioned in the article.

The comparison with today’s personal computers is not quantified. The source does not state how much RAM or storage a typical current PC has, so an exact multiple cannot be calculated from its figures. It does, however, present the robot requirement as a major future source of memory demand.

This matters because robots may add many new memory-consuming devices beyond computers and data centers. Their eventual impact will depend on how widely humanoid robots are adopted and how much memory each one needs. SK Hynix similarly identifies robotics and autonomous driving as longer-term physical-AI markets.

When RAM supply stays below demand, consumers can face higher prices and fewer readily available products. Computer makers, laptop brands, and people upgrading PCs may all compete for limited memory. The article connects the shortage with a “super-inflated component climate,” making purchases more painful for ordinary users.

The pressure is not limited to one type of computer. Enterprise systems and AI data centers are demanding large amounts of memory, while future autonomous vehicles and robots could create additional demand. Those buyers may compete with PC manufacturers and individual consumers for production capacity.

The exact price and availability effects are not quantified in the article. However, it clearly presents the situation as concerning for consumers, even while chipmakers enjoy strong profits. If demand continues to outpace new capacity, PC owners and gamers could face persistently expensive or harder-to-find RAM through the decade.

AI data centers need substantial memory infrastructure, and enterprise technology is increasingly participating in that expansion. SK Hynix says AI-powered productivity gains are encouraging more investment in data centers. It expects this cycle to broaden beyond specialist AI companies to general enterprises and nation-level initiatives.

That investment creates strong demand for memory chips. SK Hynix identified enterprise and AI demand as important short-term growth drivers. Its Q3 2026 revenue exceeded 100 trillion won, a first for the company in a single quarter. The result illustrates how strongly these markets are supporting memory manufacturers.

SK Hynix expects the global AI data-center infrastructure market to keep growing steadily. That forecast supports continued demand for memory, even if some software-AI activity slows later in the decade. The company also sees physical AI, such as autonomous driving and robotics, as a longer-term source of growth.

Building memory-chip capacity is slow because advanced production depends on difficult leading-edge processes. Manufacturers must also construct and prepare new production facilities before those factories can add meaningful output. SK Hynix describes the required timeline as a long lead time.

This creates a mismatch between sudden demand and gradual supply growth. If AI data centers or enterprises increase orders quickly, existing factories may not be able to respond immediately. New facilities cannot simply be switched on, so available memory can remain below market demand while construction and production preparations continue.

SK Hynix used these constraints to explain its forecast that shortages will persist through 2030. The article does not provide a specific construction timetable or factory count. It does make clear that technical difficulty and slow capacity expansion are central reasons the company expects the supply-demand imbalance to last.

Physical AI refers here to systems that operate in the real world, including autonomous vehicles and robots. These machines may need memory to support their computing and onboard functions. SK Hynix identifies physical AI as a mid-to-long-term demand source.

Micron offers a concrete example: it estimates that humanoid robots could require 200GB or more of RAM and multiple terabytes of storage each. If such robots become widely adopted, every unit could add significant demand for memory products. The article does not state how many robots will be built.

This could extend the memory squeeze beyond today’s AI data-center boom. SK Hynix says demand may shift or broaden toward physical AI, while the article notes that the outcome depends on adoption and each robot’s memory needs. Autonomous driving could provide another major application alongside robotics.

Supply is the amount of memory manufacturers can produce, while demand is how much customers want. If factories provide more chips than buyers need, availability is usually comfortable. If customers want more memory than factories can make, supply becomes tight and prices can rise. The article describes that second situation continuing through 2030.

Manufacturing capacity cannot expand instantly. SK Hynix cites difficult leading-edge processes and the long lead time for new facilities. As AI data centers, enterprises, and eventually physical-AI systems increase orders, existing capacity may remain under pressure until additional production becomes available.

Semiconductor production therefore moves in cycles of investment and expansion rather than immediate response. Strong demand encourages chipmakers to invest, but new factories take time to contribute output. In the article, this creates a “virtuous cycle” for manufacturers and shareholders, while consumers experience expensive or constrained memory.