How long can the boom in storage last?

The semiconductor industry, especially the memory market, is currently experiencing an unprecedented expansion—only “explosive growth” can adequately describe it. Phrases like “thriving” simply can’t do justice to the scale of what’s happening. Just take a glance at the chart, and anyone would be shocked by how massive it is.

Figure 1 shows the three-month moving average shipment volumes of various semiconductor categories, based on World Semiconductor Trade Statistics (WSTS) data, from 1991 to May 2026. Over the past 30 years, despite fluctuations in the economic environment, the four major categories—microdevices, memory, logic devices, and analog devices—have overall shown a steady upward trend. This represents the semiconductor industry’s “normal development trajectory.”

Figure 1. Three-month moving average shipment value by semiconductor product type (through April 2026) Source: Created by the author based on WSTS data

However, since around 2024, logic circuits—especially the rapidly emerging memory—have been growing in a nearly vertical surge, completely overturning people’s traditional understanding. In particular, the growth in memory is especially astonishing, nearly breaking through the peak of the chart. The speed of this growth is so fast that the growth history accumulated over the past 30 years seems like nothing more than warm-up.

Based on my many years of closely observing the semiconductor market, I can say with certainty that the memory market has never seen growth this steep. This is nothing short of a historic anomaly that is shaking the semiconductor industry’s very foundation.

In roughly 10 years, it grew by 10x, with a year-over-year growth rate as high as 285%—truly shocking. Figure 2 illustrates this explosive growth, focusing on the period of 2016 and beyond. In April 2016, the monthly shipments of MOS memory were about $5.6 billion. After a brief boom during the 2017–2018 “memory bubble,” shipment volumes plunged to only $5.8 billion by the economic downturn in early 2023. This marks the lowest point of the semiconductor industry downturn.

Figure 2. Three-month moving average shipment value of semiconductor products (by type) (April 2016 to April 2026) Source: Created by the author based on WSTS data

However, since then, there has been a recovery—or more precisely, a surge. As of May 2026, monthly memory shipments have skyrocketed to $63.3 billion. Compared with 2016, they have grown by more than 11x in just a decade. Compared with the low point in early 2023, they have grown by about 10.7x within just a few years.

Likewise, Figure 2 also shows that the scale of the logic business—including NVIDIA GPUs—has grown from $13.3 billion to $31.6 billion, clearly reflecting the “NVIDIA GPU effect” and the “AI effect.” However, even logic’s growth, compared with the explosive surge in the memory business, looks relatively mild.

From the perspective of growth rates, this anomaly becomes even more evident. Figure 3 shows the year-over-year growth rates of the same data. Memory’s recent year-over-year growth rate has reached as high as 285%, setting an all-time high.

Figure 3. Year-over-year growth rates of three-month moving average shipment volumes by semiconductor type (through April 2026) Source: Created by the author based on WSTS data

For reference, even at the peak of the previous “memory bubble” period (around 2017), the annual growth rate reached about 60%. By contrast, you can see just how astonishing the current growth rate is. Looking at the other three categories in Figure 3, even logic circuits are only around 40%, while microcontrollers and analog circuits are only around 14%–19%. In other words, only memory’s growth has truly reached “another level.”

DRAM and NAND flash are experiencing explosive growth

From the perspective of the memory market, the main drivers behind its explosive growth come from two memory products: DRAM and NAND flash (hereinafter referred to as NAND). Figure 4 shows a quarterly overview of the global DRAM and NAND markets based on TrendForce data.

Figure 4. Global DRAM and NAND markets (TrendForce forecast for Q2 2026) Source: Created by the author using TrendForce DataTrack data.

In early 2023, during the semiconductor industry downturn, DRAM prices fell to a bottom level of just $9.7 billion, and NAND prices also fell to $8.7 billion. This was a dark period when all memory manufacturers fell into losses and were forced to cut production.

However, by Q2 2026, the DRAM market is expected to jump to $145 billion, and the NAND market is expected to reach $81 billion. Compared with the low point in early 2023, the DRAM market size will grow by about 15x, while the NAND market size will grow by about 9x. If you combine the two, the quarterly total will reach $226 billion, and the annual total will exceed $900 billion—an astonishing figure at first glance.

Products once considered “cheap standards” or “commodities”: memory is now preparing to seize leadership in the semiconductor industry from microcomputers such as logic and MPUs.

The reason is abnormal surges in storage and wafer prices

Why is the memory market expanding to such a scale? The key is that this explosive growth is not simply driven by higher shipment volumes. The biggest factor is the abnormal surge in memory prices themselves.

Figure 5 shows the trends in DRAM (DDR516Gb2Gx8) spot prices and NAND (1TbTLC wafer) wafer prices. In early 2025, the DRAM spot price was only $4.70. Recently, the price surged to $46.00—an increase of about 10x. NAND wafer prices also jumped from $2.40 to $25.00, again an increase of about 10x.

Figure 5. Explosive rise in spot prices of DRAM (DDR516Gb2Gx8) and NAND (1TbTLC wafer) Source: Created by the author based on TrendForce DataTrack data

In other words, the main reason the market size grows by 10x is not that shipments grow by 10x, but because the “unit price” surges by nearly 10x. Even if memory shipment volumes stay the same, revenue will grow by 10x. This is the mechanism behind the explosive growth in the memory market size.

For memory manufacturers, this is an ideal environment. After all, prices can rise automatically without requiring a major increase in capital investment, thereby significantly improving profit margins. As we will discuss later, the stock prices of memory manufacturers have also surged because profits have been growing rapidly as a result of this abnormal price increase.

The reason for the price surge

So why have memory prices surged? It is because demand far exceeds supply—and if we trace the root of that demand, we find that it is due to the staggering investments made by hyperscale data center operators.

Figure 6 shows the capital expenditure trends of four major hyperscale data center operators (Amazon, Google, Microsoft, and Meta). In 2015, the combined total capital expenditure of these four companies was only $21 billion. Even then, that was considered a “huge sum.”

Figure 6. Capital expenditure of the top 4 hyperscale data center operators Source: Compiled by the author based on each company’s financial reports and TrendForce data, etc.

However, since OpenAI released ChatGPT in November 2022, the generative AI field has been swept by a wave of enthusiasm, and the capital investment curve has risen sharply as well. It is expected that by 2025, the total investment from these four companies will reach $355 billion, and by 2026 it is expected to reach an astonishing $755 billion. This represents an unprecedented investment boom—within just over a decade since 2015, investment has grown by about 36x.

$755 billion far exceeds 1.2 million billion yen. Just these four companies alone plan to invest within a single year an amount comparable to Japan’s national budget (general account) for building data centers and AI infrastructure. This makes it clear how extraordinary the situation is.

AI data centers are like “black holes”

The ultimate destination of this massive investment is AI data centers. Figure 7 presents the current situation in schematic form.

Figure 7. Why did DRAM and NAND flash prices surge?

As hyperscale data centers race to invest in AI data centers, the semiconductors required for AI learning and inference—namely GPUs from companies such as NVIDIA, high-bandwidth memory (HBM) used inside GPUs, and high-capacity storage SSDs equipped with NAND flash—are being pulled into the “black hole” of AI data centers one after another.

Memory manufacturers prioritize producing high-margin HBM and high-performance DRAM and NAND flash for data centers. This is simply a natural business decision. As a result, production line capacity is shifted toward the AI sector, causing a significant reduction in capacity available for other applications.

The industry hit the hardest is the DRAM and NAND flash used in digital consumer electronics such as personal computers, smartphones, and game consoles. These products are facing severe storage shortages, to the point of “complete inadequacy.”

Because supply capacity is limited, if demand becomes concentrated only in AI data centers, storage for consumer electronics will inevitably run out. This will lead to fierce competition for limited supply, which in turn will push prices up. This is directly related to the abnormal surge in storage prices mentioned earlier.

In fact, PC and smartphone manufacturers have already begun issuing distress signals, saying they can’t obtain the storage they need, procurement costs are rising, and they are being forced to pass these costs on to product prices. Ironically, under the shadow of the AI boom, the digital devices we use every day are becoming more expensive—and even more scarce.

Semiconductor market forecasts were “completely wrong”

In 2023, I made a prediction about the development trend of the global semiconductor market through 2032, as shown in Figure 8. By analyzing the semiconductor industry’s history and combining the “PC effect,” “internet effect,” “smartphone effect,” and the upcoming “AI semiconductor effect,” I predicted that the semiconductor market would grow at a pace of about “doubling every decade.”

This forecast expected that by 2032, the market size would reach about $1.2 trillion. At the time, I considered this to be a fairly optimistic prediction.

Figure 8. 2023 forecast for the global semiconductor market. Data source: Based on WSTS data and the author’s forecast.

However, it turns out that this prediction was completely wrong. In fact, it was wrong because it was too conservative. Please refer to Figure 9. According to the WSTS (World Semiconductor Trade Statistics) 2026 spring forecast, the global semiconductor market size will reach $630.5 billion in 2024, $795.6 billion in 2025, and will break through $1.5 trillion in 2026 to reach $1.5112 trillion. In addition, it is expected to rapidly grow to $1.9137 trillion by 2027, approaching $2 trillion.

Figure 9. Rapid expansion of the global semiconductor market due to the flourishing of AI—forecasts were far off Source: Created by the author based on WSTS data and the WSTS 2026 spring semiconductor forecast.

In just a few years, the market size easily surpassed the author’s earlier 2032 forecast level of $1.2 trillion. The 2032 forecast was already achieved in 2026. This does not mean the earlier forecast was “too optimistic”—rather, the disruptive power of the AI boom has completely overturned all previous conventions in the semiconductor industry.

Figure 10 clearly shows that this rapid growth is driven by memory (including DRAM and NAND) and logic circuits (including GPUs). The memory market is expected to break through $1 trillion in 2027, reflecting the explosive growth of DRAM+NAND described above. The logic circuits market is also expected to exceed $500 billion.

Figure 10. Rapid growth of memory (including DRAM and NAND) and logic (including GPUs) Source: Created by the author based on WSTS data and the WSTS 2026 spring semiconductor forecast.

On the other hand, the analog circuits and micro-semiconductor segments are almost flat. In other words, overall growth in the semiconductor market is not balanced. Instead, only these two AI-related segments are growing at unprecedented speed, pulling the entire market’s growth along with them. This creates an extremely distorted growth structure.

How long can this boom last?

Now, let’s address the question everyone cares about most: “How long will the flourishing of AI and the growth it brings last?”

To understand this, the key is to know the long history of the memory market. Figure 11 shows the year-over-year growth rates of the memory market over nearly 35 years since 1991. The chart reflects countless ups and downs that the semiconductor industry has experienced.

Figure 11. Year-over-year growth rates of the memory market Source: Created by the author based on WSTS data and the WSTS 2026 spring semiconductor forecast.

From the boom following the release of Windows 95 in 1995, to the IT bubble and its subsequent burst in 2000, to the memory bubble in 2017–2018, the Lehman Brothers collapse in 2008, and then the boom brought by the COVID-19 pandemic since 2020—the semiconductor industry has experienced roller-coaster booms and slumps.

What matters here is not the height of any single peak, but the “duration” of sustained positive growth. Careful analysis of Figure 11 reveals a key fact: in any historical period, the longest stretch during which the memory market maintained average annual positive growth has only been five years. Over the past 35 years, there has never been an instance of six or seven consecutive years of positive growth.

Why does the market tend to stabilize after five years? Because the memory market is fundamentally influenced by the “silicon cycle,” which includes demand surges, price increases, companies increasing production investment, supply oversupply, and then price crashes.

If the economy continues to prosper, companies inevitably compete to invest in equipment, which ultimately leads to supply exceeding demand and causes prices to crash. This is how memory operates as a product—its very nature determines this mechanism.

The current AI boom is expected to drive substantial growth in the memory market, starting from the 2023 low point and reaching a peak in 2024. Following historical patterns, this boom should end no later than 2028, and it may even reach its peak as early as 2027.

Although it is inevitable that someone will say “this time is different” or “AI is special,” we must not forget that every time a bubble economy appeared in the past, the same kind of claims were made.

The higher the mountain, the deeper the valley

Semiconductor industry history also tells us another undeniable rule: “The higher the peak, the deeper the following valley.”

If you look again at Figure 11, this rule becomes obvious. The IT bubble reached a peak annual growth rate of over 50% in 2000, but in 2001 immediately afterward, it plunged to a deep drop of -49.5%. Similarly, the 2017–2018 memory bubble saw a peak above 60%, followed by a sharp decline of 33% in 2019. In this industry, the higher the prosperity peak, the more severe and longer-lasting the subsequent downturn tends to be.

Let’s examine the current AI boom through this lens as well. A year-over-year growth rate of 285% is an unprecedented peak, far exceeding any previous bubble. The increase is so large that the peak of the IT bubble and the memory bubble looks insignificant in comparison.

Applying this principle, the conclusion is clear. After this unprecedented prosperity, the “valley” that follows is likely to be deeper and more severe than any downturn the semiconductor industry has experienced before. We should prepare for an extremely serious economic recession expected to begin in 2027–2028, and one that cannot be overcome with half-hearted efforts.

Memory manufacturers’ stock surges and “billionaires”

Right now, the stock prices of memory manufacturers are surging across the board. This is entirely predictable, because price increases lead to a major jump in their profits.

A symbolic example is Kioxia. The company’s stock price rise means that 600 investors—so to speak, “billionaires”—have gained more than 1 billion yen in unrealized gains from the shares they hold. The entire industry is immersed in celebrations like never before.

However, we now need to stay calm. Stock prices are like mirrors reflecting people’s expectations for the future. When those expectations become too high, even if reality falls slightly short of expectations, the stock price will collapse instantly.

The higher the mountain, the deeper the valley. The same logic applies to stock prices, not just market size. In every previous bubble burst, this scenario played out: at the peak of the bubble, “billionaires” who became rich were forced to significantly reduce their asset holdings when the recession arrived.

Now is a period of economic prosperity—precisely the time to prepare practically for a recession. This prosperity will not last forever. History shows that positive growth in the memory market can last at most five years; after that, a downturn period will arrive (at least on the surface, that’s how it looks). The higher the peak, the deeper the valley. Therefore, I strongly urge the companies currently enjoying prosperity: “While prosperity is still continuing, you should prepare yourselves practically for a recession.”