The more frantic AI develops, the more expensive your computer will be: A 40-year war among the "Big Three" storage giants

If you plan to buy a computer this year, you might be startled by the price first.

It's not because the screens have suddenly become more expensive, nor is it because the CPU manufacturers have managed to squeeze out a bit more performance. The real culprits behind the price increase are two things that you usually don't pay much attention to: memory and hard drives.

In the past, when buying a computer, people would focus on the chip, screen, graphics card, and camera. Few people would seriously calculate the cost of a single memory stick. Memory was a supporting role, just a line in the configuration list that was casually glanced over. But in 2026, this supporting role suddenly started to take center stage.

Apple, Microsoft, Sony, Nintendo, Xiaomi, vivo, and many others can't hold up anymore.

PC manufacturers, game console manufacturers, and mobile phone manufacturers are all facing the same problem: the components that they could previously pressure suppliers on are no longer negotiable. In the past, when consumers thought the products were too expensive, brands could ask the supply chain to absorb part of the cost. Now, the supply chain can only shrug and say they can't take it anymore because those further upstream have already raised the prices.

Who is raising the prices?

The answer is simple: the Big Three in storage.

Samsung, SK Hynix, and Micron.

These three names rarely appeared in the purchasing decisions of ordinary consumers in the past. You might be using an Apple computer, a Dell computer, a Xiaomi phone, a PlayStation, or an Xbox, but they all rely on DRAM and NAND. The former determines how many tasks a device can handle simultaneously during operation, while the latter determines how many photos, games, videos, and files you can store.

For the past few decades, the storage industry has always had a tough reputation.

When prices rise, manufacturers frantically expand production. When prices fall, the entire industry suffers losses. One moment they're making huge profits, and the next they're laying off employees. Storage manufacturers are like miners in a cycle; the more they dig, the more likely they are to bury themselves.

But AI has rewritten this tough business.

AI servers are not ordinary computers. They consume a large amount of GPUs and memory, and they do so voraciously. An ordinary consumer might struggle to decide between 8GB and 12GB of memory when buying a phone. However, when an AI cluster purchases memory, the units are thousands of accelerator cards and hundreds of terabytes of bandwidth. More importantly, what AI needs is not ordinary memory but HBM - High Bandwidth Memory.

You can think of HBM as a "high - speed supply station" next to the AI chip.

The GPU is responsible for calculations, but calculations can't happen out of thin air; it needs to constantly read data. Ordinary memory is like an ordinary road, which gets congested when there are too many cars. HBM is like a highway built right next to the factory, with a short distance, large bandwidth, and high speed. Without enough HBM, even the most expensive GPU will starve when training large models, running inferences, or doing AI ASIC work.

As a result, a very unusual phenomenon has emerged:

The hotter AI gets, the more expensive consumer electronics become.

On the surface, these two things seem unrelated. What does your purchase of a MacBook have to do with the construction of data centers by OpenAI, Google, Microsoft, and Meta? The connection lies in the wafer production capacity.

The production capacity of storage manufacturers is not like a faucet that can be turned on at will. Advanced DRAM production lines, NAND production lines, packaging production lines, and HBM stacking capabilities all require a construction cycle of several years. Now that cloud providers are coming with long - term contracts to secure supplies, and AI servers offer higher profits and greater certainty, the three major storage manufacturers will naturally prioritize supplying data centers.

The result is that AI servers take up more advanced memory production capacity, and consumer electronics manufacturers can only scramble for the remaining supplies.

This is not just a simple price increase.

This is a re - distribution of power in the electronics industry chain.

In the past, terminal brands like Apple, Microsoft, Dell, and Lenovo were more dominant. They had large orders and strong bargaining power. Suppliers were often willing to accept low profit margins in order to get on the list of major customers. Now the situation has changed. Storage manufacturers no longer need to focus only on mobile phones and PCs. They now have a bigger customer in front of them: AI data centers, such as large cloud providers like Microsoft, Google, and Amazon.

When the downstream customers shift from "consumer electronics" to "AI capital expenditure", the fate of the storage industry also changes.

This war has been going on for forty years.

01 Forty years ago, storage was not a Korean business

Looking at the storage industry today, Samsung and SK Hynix seem to be the natural protagonists. But in the 1980s, the center of this industry was not South Korea, but Japan and the United States.

Japanese manufacturers were once the kings of DRAM.

NEC, Toshiba, Hitachi, Fujitsu, and Mitsubishi Electric. These names might sound like a list of old - fashioned companies to young investors today, but in the 1980s and 1990s, they were the real stars of the global storage industry. Japanese companies were good at precision manufacturing, yield management, and long - term supply chain collaboration. DRAM, as a product, rewarded these capabilities.

Americans were initially very reluctant to accept this.

Because semiconductors were originally a core industry in the United States. Intel, TI, Motorola, and IBM all once held important positions in storage and logic chips. However, Japanese manufacturers caught up with their manufacturing capabilities and price competition, and American companies were becoming increasingly uncomfortable in the DRAM market. The semiconductor friction between the United States and Japan in the 1980s was essentially a collective anxiety after an industrial power shift.

South Korea stepped in at this point.

Samsung developed the 64Kb DRAM in 1983. Hyundai Electronics, which later became Hynix, started mass - producing DRAM in the mid - 1980s. At that time, South Korean storage companies had neither the technological accumulation of Japan nor the industrial reputation of the United States. What they had was the national will to catch up, the financing ability of chaebols, and a tenacious spirit that was not afraid of losses.

The storage industry is both suitable for late - comers to make a comeback and for them to fail.

It's not like consumer brands, which can gradually cultivate users through marketing. Nor is it like software, which can be launched in one version and then iterated. The storage war is very straightforward: a slight difference in process means higher costs; a lower yield means no profit; and a wrong step in the cycle means a break in cash flow.

Companies that can survive are not ordinary ones.

02 Samsung learned a cruel rule first: don't die in the trough

Samsung became the leader in the storage industry because it survived every trough.

There is an anti - human aspect to the storage industry: the time when you should cut investment the most is often also the time when you should continue investing.

When prices fall and profits decline, the first reaction of a normal company is to stop production, reduce production, and cut capital expenditure. But the problem is that semiconductor technology waits for no one. If you don't invest in this cycle, you'll fall behind in the next technological node. Once you fall behind in the technological node, costs won't come down. And if costs don't come down, you'll lose faster in a price war.

So storage is a business that "poor people shouldn't enter".

Behind Samsung is the South Korean chaebol system, its electronic terminal business, national industrial policies, and a large enough balance sheet. It can continue to build factories, buy equipment, and upgrade processes while others are struggling. When the cycle turns upward again, Samsung not only regains profits but also takes a larger market share.

This is Samsung's strategy:

Survive the cycle with scale.

Eliminate competitors through the cycle.

Use the supply contraction after competitors exit to drive up prices in the next cycle.

After the 1990s, Samsung gradually ascended to the throne of the global DRAM market. It won by outlasting its competitors. There are few fairy tales in this industry; most victories look like funerals.

However, Samsung's strength also planted the seeds of trouble in the later HBM era.

When a company is used to solving problems with scale, manufacturing, and vertical integration for a long time, it tends to believe that all problems can ultimately be solved by throwing resources at them. But HBM is not ordinary DRAM. It's not just about making the memory larger; it requires stacking, packaging, verification, and it also needs to be coordinated with the design rhythms of NVIDIA, Google TPU, and various AI ASICs.

In other words, what HBM sells is not a standard component.

It sells a system position.

This time, Samsung was a step behind.

03 SK Hynix's comeback is a window of opportunity for the second - place player

The story of SK Hynix is completely different from that of Samsung.

It's not a born king; it's more like a second - place player who has long lived in the shadow of others.

SK Hynix's predecessor was Hyundai Electronics. Around 2000, the Asian financial crisis and the downward cycle of the semiconductor industry hit South Korean companies hard. After Hyundai Electronics changed its name to Hynix, it went through a series of troubles such as debt, restructuring, and sales. It wasn't until 2012 when the SK Group took over Hynix that the company really entered a new stage.

At that time, this deal didn't seem very attractive.

Because the storage industry still had the same old problems: the cycle was too harsh, capital expenditure was too heavy, and Samsung was too strong. Buying a company that had been long suppressed by the leader didn't sound like a bargain - hunting; it sounded more like taking over a troubled asset.

But SK Hynix later waited for a window of opportunity: HBM.

HBM wasn't originally developed for ChatGPT; it has existed for a long time. But before the explosion of AI large models, it was never the most prominent market. What really turned HBM into a strategic resource was the explosive expansion of GPU clusters.

NVIDIA needs it.

Cloud providers need it.

Model companies indirectly need it.

The capital market finally realized that what is truly scarce in the AI era is not just computing power but also the bandwidth next to it.

SK Hynix seized this opportunity. It locked in key customers earlier in the HBM market, increased production earlier, and established supply relationships earlier. In the past, when people talked about South Korean semiconductors, the first thing that came to mind was Samsung. Now, when talking about AI memory, the first thing that often comes to mind is SK Hynix.

This is the most interesting reversal in the industry history:

The second - place player didn't defeat the first - place player in the old battlefield but re - defined victory and defeat in a new battlefield.

If Samsung won with scale discipline in the DRAM era, SK Hynix won by binding customers in the HBM era. The former is like building a city wall, while the latter is like seizing a wharf. When the AI cargo ships arrive, whoever occupies the wharf gets to collect the money first.

04 Micron is not the strongest, but it's the best at surviving

Micron's story follows a different path.

It doesn't have the South Korean chaebol system, nor does it have a vertical electronics empire like Samsung. It was founded in 1978 in Boise, Idaho, USA, and was initially just a small semiconductor design company. In the Silicon Valley narrative, Micron isn't even a typical protagonist.

But it survived.

These four words are very valuable in the storage industry.

There were many storage players in the United States in the past, but the only one that really remained in the first - tier is Micron. It has experienced the impact of Japan, the rise of South Korea, the PC cycle, the smartphone cycle, the prosperity of Chinese mobile phones, and the global semiconductor downturn.

Its crucial leap was the acquisition of Elpida in 2013.

Elpida was the last important legacy of Japanese DRAM. Japanese storage was once so strong, but in the end, its core assets were taken over by Micron. This deal gave Micron a larger DRAM scale, mobile DRAM capabilities, and Asian manufacturing assets, and it also further consolidated the global DRAM market.

Since then, the storage market has become more like a three - way battle.

Samsung, SK Hynix, and Micron control the world's most critical DRAM supply. There are more competitors in the NAND market, but in high - end DRAM and HBM, these three companies are the ones with real global influence.

The more concentrated the industry structure is, the more different the cycle becomes.

In the past, when prices rose, a lot of manufacturers would expand production, and in the end, they would kill themselves. Now that there are fewer players, everyone knows better. All three companies have experienced severe downturns and know the consequences of blind expansion. So the storage industry has changed from "everyone wants to increase production" to "no one wants to be the first to disrupt the market".

This doesn't mean the cycle has disappeared.

It just means the cycle has been tamed a bit.

05 After 2013, the war changed from a price war to production capacity discipline

The most tragic time in the storage industry is not when prices fall but when everyone thinks they can outlast others.

This is a typical prisoner's dilemma.

Every company knows that an oversupply will lead to price drops, but every company is also afraid that after they reduce production, their competitors will secretly expand production to grab market share. So they all expand together, prices fall together, and profits disappear together. In the end, it's not about who wins but who is the last to fall.

Around 2013, after the industry consolidation was completed, the logic began to change.

Elpida was merged into Micron, and Japanese DRAM exited the main stage. South Korea formed a dual - strong situation with Samsung and SK Hynix. The United States was left with only Micron. With fewer players, everyone finally began to cherish profits. The storage industry moved from a brutal price war to an oligopoly discipline era.

This change is very important.

Because it explains why the prices have risen so sharply under the current AI - driven demand.

If there were still seven or eight major players in the storage industry today, when the AI demand came, everyone might expand production regardless, and the price increase would be quickly offset by the new supply. But now it's different. The three major manufacturers all know that HBM, server DRAM, and enterprise - level SSDs have high profits. So why should they allocate their most precious production capacity to low - margin consumer electronics?

As a result, the production capacity began to shift.

The wafers are going to servers.

Capital expenditure is going towards HBM.

Customer priority is going to cloud providers.

Consumer electronics are being pushed to the back.

This is the real logic behind this year's price increase.

It's not that mobile phone manufacturers have suddenly become greedy, nor is it that computer manufacturers suddenly want to make a few hundred more dollars from you. It's that the upstream has re - allocated the supply. AI infrastructure is taking the goods first, and consumer electronics are queuing up at the back. If they can't get the supply, the prices will rise. If the prices can't be raised, the configurations will be cut. If that still doesn't work, the launch of new products will be postponed.

To put it simply, consumers are not just paying for Apple, Microsoft, and Sony now.

They're also paying for AI data centers.

06 HBM has reshuffled the seats of the Big Three in storage

If we only look at traditional DRAM, Samsung is still very strong.

But if we look at the narrative of the AI era, SK Hynix is more like the new protagonist.

This is the kind of plot that the capital market loves: the old king loses speed, the second - place player takes the throne, and the third - place player makes a comeback.

SK Hynix has obtained the best label with HBM. It is no longer just the second - largest South Korean storage company next to Samsung; it has become a core player in NVIDIA's AI supply chain. The tighter the AI computing power, the tighter the HBM supply. The tighter the HBM supply, the stronger SK Hynix's pricing power.

Samsung's problem is that it's too big.

Being big has its advantages, but it also comes with a price. Samsung is involved in mobile phones, TVs, home appliances, storage, found