The global computer market is entering a phase of tension that could increase the price of computers, smartphones and storage systems. The US manufacturer HP warns that the shortage of memory chips is at an « unprecedented » level and believes that prices will continue to rise. Behind this surge is a much deeper transformation of the semiconductor industry: artificial intelligence absorbs an increasing share of production capacity, while memory manufacturers favour components for data centres, which are more cost-effective than those used in consumer electronics.
The situation therefore goes beyond a simple cycle of rising AMR prices. For PC manufacturers, the problem is simultaneously becoming industrial, commercial and strategic. They must secure memory volumes that have become scarce, accept significantly higher costs and decide how much of this increase can be absorbed in their margins. The rest will necessarily reach the consumer either by higher prices or by less generous configurations.
HP warns: memory prices will rise again
HP’s warning, reported on August 21, is particularly clear. The U.S. group describes an exceptional shortage and does not anticipate rapid price decline.
This is in line with the findings of several months ago by specialists in the sector. Two categories of components are particularly concerned: DRAM, which is used mainly as live memory in computers and servers, and NAND, which is the core of SSDs and a large part of flash storage.
These components have become indispensable for almost all the digital economy. A laptop uses several gigabytes, a smartphone too. But a server intended for artificial intelligence consumes it in considerably higher proportions.
It is this scale difference that currently unbalances the market.
The development of the major data centers dedicated to AI leads Microsoft, Google, Amazon, Meta and other groups to order massive volumes of memory. At the same time, Nvidia and other specialized processor providers need high-speed memory to power their accelerators.
As a result, flea manufacturers are faced with an obvious economic trade-off: dedicating their limited capacities to relatively low-paying standard products or directing them to high-end memories sold to data centres at significantly higher prices.
Artificial intelligence changes factory economy
The current shortage is not only due to a sharp increase in demand. It also results from the decisions taken by the producers themselves.
Samsung Electronics, SK Hynix and Micron dominate most of DRAM’s global market. These groups are increasingly investing in memories for artificial intelligence infrastructure, including HBM, or broadband memory.
The HBM has become one of the strategic components of accelerators used to train and execute large artificial intelligence models. Its production requires significant investments and mobilizes capacities that cannot be immediately re-allocated to traditional memory.
Industry therefore faces a form of internal competition for the same industrial resources.
A slice of silicon used to manufacture certain components intended for a d Clean rooms, manufacturing equipment and capital investment are also limited.
Producers naturally favour markets with the highest margins.
This shift in investment explains why an increase in demand for artificial intelligence can ultimately increase the price of a desktop computer that does not itself use any particularly advanced AI service.
A crisis different from previous shortages
The memory sector is historically very cyclical. Periods of shortage lead to higher prices, manufacturers invest, production increases, and then excess supply leads to a decline.
This mechanism has been repeated several times.
But the current crisis could be different in scope and especially in the nature of demand.
Investments in data centres do not correspond to a one-time phenomenon comparable to the launch of a new generation of smartphones. They are based on a transformation of global IT infrastructure involving hundreds of billions of dollars.
Major technological groups continue to increase their spending on artificial intelligence. Each new generation of servers requires not only more computing capacity, but also more memory and storage.
IDC already estimated that the growth in global supply of DRAM and NAND would remain below traditional levels in 2026, despite the increase in demand.
This means that the market cannot simply rely on a rapid acceleration of production to restore equilibrium.
Building or expanding a semiconductor plant takes several years and billions of dollars. Even when producers decide today to increase their capacity, the additional supply does not immediately reach the market.
The price of PCs becomes directly exposed
For HP, Dell, Lenovo, Acer, Asus and other major manufacturers, memory is now a much larger component of the cost of a computer.
When prices were low, increasing a PC from 8 to 16 GB, or 16 to 32 GB of RAM, represented a relatively limited cost. Manufacturers could offer more memory without upsetting the final price.
This logic becomes much more difficult when the DRAM price increases.
IDC considered several scenarios for 2026 where the average price of computers could increase from 4% to 8%, depending on the duration and severity of the shortage. In the worst case scenario studied by the firm, global PC sales could at the same time decline by almost 9%.
These projections are not a definite forecast. However, they illustrate the economic mechanism at work.
A more expensive computer causes some purchases to be postponed. Individuals keep their machine longer. Companies are spreading their renewals. The decline in volumes can then affect manufacturers, even as their costs increase.
This could result in both component inflation and low sales.
Consumers could pay otherwise than in cash
The increase in the catalogue price is not the only way to pass on the increase in the components.
A manufacturer can also maintain a relatively stable price while reducing the characteristics of the machine.
Instead of offering 32 GB of memory in a certain range, it can maintain 16 GB. It can choose a 512 GB SSD rather than 1TB or reserve the capabilities superior to the most expensive models.
This strategy avoids a too visible increase in the price displayed, but nevertheless constitutes a form of inflation: for the same sum, the buyer gets less memory or storage than he might have received before.
The phenomenon could be particularly sensitive on entry and mid-range PCs.
Margins are often low. An increase of a few tens of dollars in the cost of components therefore represents a much larger proportion of the profit achieved on each machine.
High-end models offer more latitude as the cost of memory accounts for a lower share of their total price.
The paradox of « PC IA »
The shortage also comes at the worst moment for one of the IT business strategies: the development of computers capable of performing artificial intelligence functions locally.
The new « PC IA » generally incorporate more memory than traditional machines.
Microsoft imposes a minimum of 16 GB of RAM for certain categories of Copilot+ PCs, while more advanced uses can push manufacturers to 32 GB or more.
The industry is therefore looking to sell machines richer in memory precisely when this memory becomes much more expensive.
The paradox is important.
The growth of artificial intelligence increases memory demand in data centres. This demand raises prices. These prices then make it more expensive to manufacture personal computers supposed to democratize artificial intelligence.
Manufacturers must therefore choose between three options: increase prices, reduce margins, or limit the amount of integrated memory.
In most cases, they will probably use a combination of the three.
Large manufacturers have an advantage
The crisis will not affect all manufacturers in the same way.
HP, Lenovo or Dell have considerable buying volumes. They negotiate directly with large memory producers and can enter into supply contracts over several months or more.
This size gives them an advantage when components become scarce.
A large manufacturer can guarantee a supplier large and regular orders. It also has more liquidity to build up stocks or accept certain contractual conditions.
Small regional assemblers and manufacturers do not have the same bargaining power.
They often buy their components in markets where prices respond much faster to shortages. When RAM or SSD become scarce, their cost increases immediately.
The crisis could thus accelerate market concentration for the benefit of major manufacturers.
Even the individually assembled PC sector, especially by players, may be affected. When the price of memory kits and SSDs increases very strongly, the cost gap between a self-mounted machine and a computer pre-assembled by a large manufacturer can be reduced.
Storage is also under pressure
Attention often focuses on RAM, but NAND could become an equally important problem.
This memory is used in SSDs, smartphones, servers and countless electronic devices. Again, the demand for data centres is increasing rapidly.
Artificial intelligence systems do not just calculate. They need to store huge volumes of data, keep models, manage databases and record the results produced.
This explosion in storage gives NAND manufacturers new and highly remunerative outlets.
Data released this week by TrendForce show that demand for artificial intelligence servers continues to keep the market under pressure despite a less dynamic demand for PCs and smartphones.
NAND’s five largest manufacturers recorded a strong increase in sales in the second quarter of 2026, driven by price increases.
For memory producers, therefore, the shortage represents an extremely favourable period.
For their customers, it is exactly the opposite.
Samsung, SK Hynix and Micron come out reinforced
This reversal of the balance of power is one of the major economic elements of the crisis.
During the overproduction phases, memory manufacturers are forced to reduce their prices to sell their stocks. Their margins can collapse rapidly.
Today, the situation is opposite.
Demand exceeds supply and customers seek to secure volumes several quarters in advance. Producers can increase prices and favour the most profitable contracts.
Samsung, SK Hynix and Micron thus become direct beneficiaries of the artificial intelligence boom, even when they do not themselves manufacture the processors that execute the models.
An increasing share of the economic value of the AI moves towards the suppliers of components essential to its infrastructure.
This development reminds us that the artificial intelligence boom does not only concern Nvidia or companies developing models. It modifies the entire industrial chain of semiconductors.
A quick lull now seems unlikely
The central question remains the question of duration.
The traditional cycles of the memory market would normally suggest that very high prices lead to new investments, then an increase in supply and finally a correction.
This mechanism will probably still work.
But his calendar becomes particularly uncertain.
Several players in the sector are now talking about tensions that may continue beyond 2027. In NAND, some industrialists believe that the new capacities needed to completely rebalance the market could take several years.
It is important to be cautious with forecasts up to 2030: the semiconductor industry can change sharply if data centre demand slows down or if producers accelerate their investments further.
However, no indicators available as at 21 August suggest a rapid return to the very low prices observed before the outbreak of this shortage.
For HP, the immediate problem is therefore much simpler: building the next generations of PCs with components that continue to rise in price. For consumers, the consequence is likely to be gradually read on labels, but also in the proposed configurations. Memory, long regarded as one of the easiest and least expensive components to increase in a computer, has become one of the most disputed strategic resources of the artificial intelligence economy.
