Over the past year, the most painful question for buyers of computers and phones has been "is it expensive?" AI has consumed vast amounts of global memory and chip production capacity, leaving limited resources for consumer products, which naturally drives up prices. Now, the issue has shifted to a more basic question: "Is it in stock?"
A couple of days ago, it was reported that ordering a MacBook Air from the official Apple website now shows delivery times stretching into late August or even September, far exceeding the usual one-to-two-week lead time.
Retail channels outside of Apple's direct stores have been more direct: the supply tightness for the MacBook Air this year is the worst they can recall. Apple has even taken the unusual step of noting in its back-to-school materials that "MacBook Air availability is subject to supply," subtly steering consumers toward the entry-level MacBook Pro.
While the shortage of the MacBook Air alone might not be surprising, when combined with significantly extended delivery times, retailer feedback, and a shift in product recommendations, the message becomes clear. If a trillion-dollar company like Apple has to ration its flagship MacBook Air, the memory crisis has entered a new phase. It's no longer about affordability; it's about availability. Price increases were the classic way to balance supply and demand鈥攔aise costs, then raise prices. But now, even if giants like Apple, Google, OpenAI, and Amazon are willing to pay several times the price, the global storage supply chain cannot increase production capacity in the short term. In simple terms, the global storage procurement market has shifted from "how much are you willing to pay?" to "you can't get it at any price."
According to multiple media reports and research firms, nearline HDD (hard disk drive) capacity in the global commercial storage market is already locked up through 2028-29, and enterprise SSD (solid-state drive) capacity is booked through 2026-27. For hot products like HBM used in data centers, and LPDDR and NAND used in phones, long-term orders are impossible, with supply contracts only available on a quarterly or even monthly basis.
The competition among major buyers has even shifted from "planned capacity" to future "new capacity." The term "price without supply" doesn't mean not a single chip can be bought; it means manufacturers cannot obtain enough chips that meet their requirements at a reasonable cost and within a feasible delivery cycle. This is a significant problem for service providers like OpenAI, but it's a crisis for consumer electronics companies like Apple. In the past, memory and storage chips for phones went to the highest bidder; now, the highest bidders just get to the front of the line.
This shift is already evident in the smartphone industry. Counterpoint's statistics show that global smartphone shipments in the second quarter of 2026 fell 11% year-over-year, marking the lowest Q2 level since 2013.
The report specifically noted that DRAM and NAND shortages forced manufacturers to repeatedly raise prices and cut allocations for low-margin models. By the third quarter, some phone makers had already implemented second or even third rounds of price increases. Meanwhile, the PC industry, hit even harder, has adopted a more creative response: if new components are too expensive, just bring back the old ones.
The most typical example is NVIDIA restarting production of the RTX 3060, a graphics card launched five years ago. Intel and AMD have also introduced new platforms compatible with DDR4 memory. Intel has explicitly stated it will "continue to offer products compatible with older memory technologies." Motherboard and memory manufacturers are retreating from the expensive and scarce DDR5 platform and increasing their DDR4 product lines, creating an abstract scenario of "moving backward."
This is not a victory for budget PC builders. The old platforms offer decent performance, but the key is they avoid the sky-high prices of new-generation memory. What manufacturers are essentially selling is a form of insurance against the crisis in the supply chain. Even Microsoft, known for its bloatware, has stopped embedding Copilot everywhere in Windows 11 and is now researching ways to reduce system memory usage and improve the experience on 8GB devices.
The fact that even an operating system is seriously reconsidering how to survive on 8GB of RAM is telling. Under Andy and Bill's law, software used to push users to upgrade hardware. Now, software is being forced to ration resources for hardware. Separately, according to a Reuters report, on July 27, Changxin Memory Technologies (CXMT) debuted on the STAR Market. Its IPO price was 8.66 yuan, and it closed at 49 yuan, a surge of about 466%, with a total market capitalization estimated at 3.28 trillion yuan. Since only about 7% of CXMT's enlarged total share capital was available for trading at launch, enormous market demand squeezed into a tiny float, making CXMT's listing an instant phenomenon as the top memory stock on the A-share market.
In the following days, stocks of Samsung Electronics, SK Hynix, and even the overall Korean market saw a significant decline. Reuters attributed the volatility to a combination of doubts about AI investment returns, a global tech sell-off, and the rise of Chinese competitors, with CXMT's listing serving as one catalyst in the highly unstable memory market.
However, the valuation frenzy surrounding CXMT's listing reveals the contradictory psychology of the capital market. Everyone hopes new players will increase supply and lower prices, but they can't help but price new entrants based on a scenario of long-term shortages. Help is indeed on the way, but fabs cannot produce chips based on stock prices. The "wait-and-see" crowd can no longer afford to wait. From computers and phones to set-top boxes, game consoles, and even car infotainment systems, almost all modern electronic devices depend on DRAM or NAND. The capacity that AI data centers are siphoning off upstream, directly or indirectly, ultimately travels down the supply chain, manifesting as higher prices, lower specifications, and longer delivery times for every individual consumer.
This crisis has evolved from a cost problem into an availability problem. This scenario was already previewed two decades ago. Today's AI and storage chips are almost a direct parallel to the solar photovoltaic industry boom and the subsequent polysilicon shortage.
Although the storage industry remains cyclical, with some analysts predicting prices could begin to correct in 2027 and sales volumes showing a clearer recovery by 2028, the problem is that the industry and Apple can wait, but your needs for school, the holidays, or an emergency cannot.
While the current AI bubble will eventually subside in the long term, not all purchasing needs can be deferred indefinitely. Now that the memory crisis has shifted from "can't afford" to "can't find," the strategy of those who wait must also change.
For the next six months to a year, whether you're buying a computer for school, building a PC for work, or upgrading your phone, in this chaotic environment, the safest approach is to know your needs and buy immediately. Because in the coming period, the longer you wait, the more price increases you will encounter.
Comments