AI data centers are crowding out memory capacity, and PC and smartphone shipments both fell sharply in the third quarter of 2026

Driven by the explosive growth of generative artificial intelligence, the global semiconductor market is experiencing an unprecedented period of activity. NVIDIA GPU sales are booming, while high-bandwidth memory (HBM), which consists of stacked DRAM chips, remains in a state of prolonged supply shortage. Prices for general-purpose DRAM and NAND flash have also soared, and memory manufacturers' capacity expansion has reached unprecedented levels.

However, when we look at personal computers and smartphones—the representative end products that use semiconductors in large quantities—the situation is entirely different. In the third quarter of 2026 (July to September), global PC shipments fell 20.1% year over year, and smartphone shipments fell 22.0% year over year. According to statistics from the U.S. market research firm IDC, both figures hit record lows.

Although the AI boom should logically trigger an explosive increase in semiconductor demand, why have PC and smartphone shipments plummeted? Investigating the reasons behind this reveals a paradox: the AI boom itself is actually putting pressure on PCs and smartphones.

This article reviews the long-term trends in PC and smartphone shipments, analyzes what may happen in 2026, and examines the special nature of the current semiconductor market.

The PC market has long been in the doldrums

First, let us look at the long-term trend in PC shipments. Figure 1 shows PC shipments and their year-over-year growth rates from the first quarter of 2009 to the third quarter of 2026.

Figure 1: Quarterly PC shipments and year-over-year growth rate (through Q3 2026). Data source: Author, based on IDC data.

Before around 2011, quarterly PC market shipments were about 90 million units, peaking at 95 million, but afterward the market entered a long period of stagnation. Since 2012, negative growth continued for many consecutive quarters, and by around 2018 quarterly shipments had shrunk to about 60 million to 70 million units.

This period coincided with the rapid spread of smartphones. Many tasks that previously required a computer—such as sending and receiving email, browsing the web, using social media, and watching videos—could now be done on a smartphone. As a result, ideas such as "one computer per household is enough" and "there is no need to replace computers frequently" began to spread widely, and the PC replacement cycle lengthened accordingly.

Then, in 2020, the COVID-19 pandemic struck. Remote work and online classes spread rapidly, and demand for PCs surged. In the third quarter of 2020, PC shipments grew 20.9% year over year, and in the first quarter of 2021 they posted an astonishing increase of 57.9%. However, this was merely the "COVID-19 boom." In 2022, the market rebounded, leading to falling demand and inventory adjustments, and the PC market declined sharply. In the fourth quarter of 2022, the growth rate fell to about 28%, hitting bottom and rebounding.

In 2025, the PC market recovered again. As shown in Figure 1, PC shipments in each quarter of 2025 grew by about 10%–15% compared with the same period the previous year. The main reason was Microsoft's ending of support for Windows 10. Microsoft ended support for Windows 10 on October 14, 2025. As a result, replacement demand surged for old PCs that could not run Windows 11, mainly from corporate users. In other words, the growth of the PC market in 2025 was not mainly due to a "generative AI-driven AI PC boom," but rather to large-scale upgrade demand caused by the end of Windows 10 support.

PCs will suddenly stall in 2026

However, the situation changed abruptly in 2026. PC shipments grew 3.0% year over year in the first quarter, but fell 3.8% in the second quarter and plunged 20.1% in the third quarter to 62.7 million units. According to IDC, the 20.1% decline was the largest third-quarter drop in its history. In addition, although third-quarter shipments usually exceed second-quarter shipments due to demand for the new school year, third-quarter 2026 shipments fell 9.1% from the previous quarter, breaking this seasonal pattern.

IDC pointed to limited supply of memory and other components, rising component prices, and the backlash from accelerated shipments in the first half of the year in response to price increases as reasons for the sharp decline in third-quarter sales. Because PC manufacturers and distributors stockpiled inventory before prices rose, some PCs originally scheduled to ship in the third quarter had actually already shipped in the first half. According to data from the British research firm Omdia, the share of DRAM and SSDs in PC component costs has risen from about 15% before the memory price surge to about 40%.

PC prices remain high, and demand is weak. The key question is why memory supply is so tight. The answer lies in AI data centers, but before that, let us look at the smartphone market.

The smartphone market is maturing and will decline sharply in 2026

Figure 2 shows global smartphone shipments and year-over-year growth rates from the fourth quarter of 2009 to the third quarter of 2026.

Figure 2. Quarterly smartphone shipments and year-over-year growth rate (through Q3 2026). Data source: Author, based on IDC data.

In the first half of the 2010s, the smartphone market grew at a staggering pace, with annual growth rates as high as 40% to 70% (over 80% in 2010), and quarterly shipments rising from about 50 million units at the end of 2009 to more than 300 million units at the end of 2014. However, around 2016, the growth rate fell to nearly zero, and smartphone growth peaked. This is because global smartphone penetration had stabilized.

Subsequently, improvements in device performance led to longer replacement cycles. Demand to replace devices every two to three years declined, and the global smartphone market had effectively matured. In 2020, the market briefly declined due to the COVID-19 pandemic, then grew rapidly in the first half of 2021, but that exceptional demand ended in 2022, causing the market to post negative growth again. Although there was some recovery in 2024-2025, the growth rate remained only a few percentage points.

Then, in 2026, the situation changed dramatically. The first quarter fell 2.9% year over year, the second quarter fell even more, by 7.4%, and the third quarter plunged to 255.3 million units, a drop of 22.0%. According to IDC, this 22.0% decline was the largest quarterly decline in its history. In terms of shipments, as shown in Figure 2, it has fallen back to around the level of 2013, when smartphones were rapidly becoming widespread.

IDC calls this sharp decline a "memory crisis." According to IDC, smartphone memory costs have risen by about 350% over the past year and are still rising. Especially for low-priced devices, memory now accounts for more than 65% of the bill of materials (BOM). As a result, the profitability of low-priced Android devices has declined, forcing manufacturers to raise prices, change specifications, or even discontinue low-priced models entirely.

In fact, while sales for Samsung and Apple fell only 4%-6% year over year, sales for Xiaomi, OPPO, and vivo—three brands that perform strongly in the low-price market—fell significantly, by about 30%-35%. This is not simply a matter of falling demand, but a vicious cycle: rising memory costs and device prices are forcing consumers to delay replacing their devices.

AI data centers are like giant "black holes"

So where has all the memory that should have gone into PCs and smartphones gone? Figure 3 illustrates the structure. AI data centers currently being built around the world are devouring GPUs, HBM, DRAM, SSDs, and NAND flash like giant "black holes."

Figure 3: Why are PC and smartphone shipments falling?

Especially in the field of generative AI, simply increasing the number of GPUs is not enough. Large amounts of HBM memory are installed around the GPUs, and AI servers are also equipped with high-capacity server-grade DRAM and enterprise-grade SSDs. Memory manufacturers are prioritizing highly profitable AI server applications, which is putting pressure on memory supply for PCs and smartphones. It should be noted that GPUs themselves are not components that directly compete with consumer products. What directly pressures PCs and smartphones is memory, specifically DRAM and NAND flash.

HBM production consumes a large share of DRAM front-end capacity. It is said that producing the same number of bits of HBM requires about three times the wafer capacity of general-purpose DRAM. Therefore, increasing HBM output for AI servers and server DRAM will reduce the capacity available for PC and smartphone DRAM.

The situation is similar for NAND flash. Demand for high-capacity enterprise SSDs from AI data centers has surged, and NAND flash manufacturers are prioritizing profitable enterprise applications. As a result, supply to the consumer market is constrained.

In other words, a chain reaction is now occurring: "surging demand for AI servers → capacity prioritized for HBM, server DRAM, and enterprise SSDs → DRAM and NAND for PCs and smartphones become scarce → memory prices soar → manufacturing costs for PCs and smartphones rise → shipments fall due to higher prices and fewer models."

As shown in Figure 3, this chain reaction affects not only PCs and smartphones but all consumer electronics that contain memory, such as game consoles. From the perspective of consumer electronics manufacturers, the current situation is one of "severe shortages of DRAM and NAND flash."

So what is the overall state of the semiconductor market? Figure 4 shows the global semiconductor market on a monthly basis (three-month moving average) and year-over-year growth since 1991, based on data from WSTS (World Semiconductor Trade Statistics).

Figure 4: Monthly global semiconductor market data and year-over-year growth rate. Source: Author, based on WSTS data. Note: Monthly global semiconductor market data are three-month moving averages.

The semiconductor market has experienced cyclical fluctuations of growth and downturn. During the 2000 IT bubble, the growth rate reached about 40%, but the bursting of the bubble caused a decline of more than 40% in 2001. The 2009 financial crisis caused the market to fall by about 30%, followed by a rebound of about 60% in 2010. Since then, despite events such as the 2019 memory market downturn and the 2023 inventory adjustment, the semiconductor market's growth rate has generally fluctuated between +40% and -20%.

However, since generative AI truly took off in 2024, the semiconductor market has undergone unprecedented expansion. Recent monthly semiconductor shipments reached $159.7 billion, up 144.3% year over year. This means the market size swelled to about 2.4 times its previous level in just one year. Compared with the 2022 peak (about $50 billion per month), when the market size was driven jointly by the COVID-19 pandemic and the memory market, today's market size is about three times that. As shown in Figure 4, in the data of the past 35 years, an annual growth rate exceeding 100% has never appeared.

This shows that the current AI boom is completely different from past semiconductor booms. In the past, the key driver of semiconductor market growth was increased shipments of end products such as personal computers and smartphones. Growth around 2000 was driven by the spread of personal computers and the Internet, while growth in the early 2010s was driven by the surge in smartphones.

However, the situation is very different today. By the third quarter of 2026, PC shipments are expected to fall 20.1%, and smartphone shipments are expected to fall 22.0%. Nevertheless, the semiconductor market size grew about 2.4 times year over year. The main factors driving this growth are explosive demand for AI-oriented GPUs and HBM, as well as soaring prices of DRAM and NAND flash memory.

In other words, it is not that "the semiconductor market grew because PCs and smartphones sold well." Rather, we now face a paradox: AI data centers consume large amounts of semiconductors, causing memory shortages and soaring prices for PCs and smartphones. As a result, even though PC and smartphone shipments are declining, only the value of the semiconductor market is growing.

How long will the "winter" of the consumer goods market last?

IDC warns that the worst is yet to come for the smartphone market, forecasting that market growth will remain sluggish until 2027 and that recovery will not come until 2028 at the earliest. As for PCs, prices are expected to remain high for the foreseeable future, and conditions may even worsen further in the coming quarters. As long as investment in AI data centers continues to grow, the memory shortage and weak growth in consumer products are likely to persist.

On the other hand, there are other hidden concerns in the semiconductor market. Semiconductor shipments grew 144.3% year over year, and a large part of that growth did not come from increased shipment volumes but from rising memory prices. With sales of PCs and smartphones (the main drivers of final demand for memory) falling 20%, is there still room for memory prices to continue rising? If the pace of investment in AI data centers slows even slightly, there is a risk that memory prices will fall sharply.

AI data centers are like a giant black hole, expanding their market size by devouring semiconductors while also cannibalizing semiconductor resources from the surrounding consumer electronics sector. The sharp decline in PC and smartphone sales in 2026 can be seen as a symbolic phenomenon of the "light and shadow" that the AI boom has brought to the global semiconductor industry.

This article is from the WeChat public account "Semiconductor Industry Panorama" (ID: ICViews), author: Takashi Yunogami, published by 36Kr with authorization.