Memory shortage forces companies to rethink budgets and IT roadmaps

The cause is a severe global shortage of silicon wafers. They are used everywhere – from wristwatches and refrigerators to large data centers. Substantial increases in electronics prices are ahead, he warns.

Grzegorz Nawacki, Cezary Szczepański, XYZ: Do companies buying computer hardware really have to rewrite their budgets right now?

Dariusz Piotrowski, General Manager, Dell Technologies Poland: Budgets – and in some cases even entire financial plans for the year.

Why is that?

Because of a significant increase in memory chip prices. This is not a temporary demand shock. We are dealing with severe shortages of key components that are used everywhere – servers, computers, cars, and even smart refrigerators.

What is driving this?

More than 85 percent of the market is controlled by three major players. When the COVID-19 pandemic hit, they sharply increased production capacity, as demand surged. After the pandemic, demand fell significantly, and production followed. Then, rather unexpectedly, starting around mid-2025, there was a sudden and very strong surge in demand for memory chips. This is a textbook “perfect storm” hitting the IT industry. The situation is genuinely abrupt and severe.

Memory chips in short supply, prices set to rise

What is the essence of the problem?

Essentially, suppliers like us – and many other companies – have begun receiving orders to build massive data centers dedicated to artificial intelligence. These are projects worth tens or even hundreds of billions of dollars. A key component of modern data centers – alongside Nvidia GPUs and chips from other manufacturers – is high-bandwidth memory (HBM).

So what happened? Manufacturers, faced with firm and extremely large orders for the most advanced – and most profitable – memory chips, decided to redirect production capacity toward that segment. That is where margins are higher, and where prices can be raised.

The result is growing shortages in the mid- and lower-end segments of the market. We started seeing this around late October and November. By December, the situation had already become difficult.

And is it still ongoing?

Yes, this is not a temporary anomaly. Over the past 30 years, the memory market has of course gone through cycles. There were even periods of price declines so steep that computer hardware became cheaper overall.

Today, the situation is the exact opposite. And memory is now everywhere: in phones, tablets, cars, smart refrigerators. Modern electronics are literally packed with it. As a result, this is not just an IT industry problem. Essentially every sector is now lining up for memory chips.

The problem is global in scale

How large is the issue?

Estimates from industry analysts suggest that, given current demand, the market is able to meet only around 70 percent of total requirements. Importantly, this gap has long-term consequences, as the order backlog continues to grow and unmet demand from previous months keeps accumulating. Everything is beginning to compound.

What are the effects of this situation?

First and foremost, sharp increases in memory prices across all market segments.

We are seeing price increases of tens of percent in some types of memory chips compared with the previous month. I cannot provide precise figures, as the situation is changing literally from day to day.

It must also be clearly stated: if a memory chip becomes more expensive in any IT device, then the entire device becomes more expensive. Everything depends on how large a share memory represents in the overall component set.

By how much will electronics prices increase?

What is the share of memory in a standard computer?

On average, it accounts for roughly 10 to 20 percent of the total value of a device. In servers, the share is even higher – around 30 percent, and sometimes more. Memory is one of the key components in that segment.

There is also a second consequence: delays in delivery schedules. At present, some companies are reporting delays of six to nine months.

Our supply chain is resilient and globally diversified. We designed it to ensure the necessary flexibility in the face of multiple macroeconomic, regulatory, and trade-related challenges.

So, does that mean that memory chips still end up at Dell regardless?

In a sense, yes. And that is the key difference compared with smaller suppliers, which offer one or a few products and suddenly find themselves at the back of the queue.

We plan our demand well in advance. That is why, despite the fact that the situation is difficult for the entire market, we feel relatively secure. Of course, such market disruptions are never positive. We would prefer a stable, predictable growth trajectory. But the current situation will take some time to normalize.

Two years before the situation changes

When could this happen?

Even if someone decided today to invest in a new memory chip factory, it typically takes 18 to 24 months from the start of construction to reaching full production capacity. This means the current structural market disruption will last at least two years – assuming, of course, that investors are willing to build new capacity at all and compete with the three giants that have dominated this market for decades.

This is not like producing plastic chairs, where production can be ramped up quickly. We are dealing with cutting-edge, highly advanced technology.

That is why one of the most important messages for companies is that this situation cannot simply be waited out. I have met many board members and IT directors who said: “We’ll wait three months and then resume our projects.” Unfortunately, it does not work that way.

What can be done?

This is the moment for a serious reassessment of IT plans – both in terms of budgets and project timelines.

Previously, companies assumed that hardware would arrive within two months, then software would be deployed, and within six months the project would be in production. Today, it may turn out that even the first stage is delayed, because hardware will arrive only after six or eight months. As a result, the entire project may not be completed until next year.

Tenders under pressure from new pricing

How does this affect larger procurement processes?

The new reality has to be taken into account. For example, it is worth shortening bid evaluation periods, and optionally designing more flexible hardware configurations, as certain types of memory are in short supply. It is also advisable to engage in dialogue with suppliers and ask about possible substitutes for a given configuration. And, of course, budgets must be planned differently.

How much further can prices rise? Given that building a new factory takes around two years, won’t delays keep accumulating for a long time?

Together with our board in Austin and senior leadership across the company, we answer this question very honestly: we don’t know. If they don’t know, then I certainly don’t know – and the entire market does not know either.

Of course, one can imagine a scenario in which the world reacts to higher prices and demand begins to fall. Economists would say that is exactly what should happen.

For 30 years, the market operated relatively stable, with only occasional short-term demand spikes. Today, we are in an unprecedented situation. I have been in this industry for over 30 years, and I have never seen such dynamics of change or such difficult negotiations. Some clients do not fully understand the situation. Some suspect that this is a deliberate move by manufacturers to raise prices. In reality, this is a structural shift in the IT market – and it will affect everyone.

A challenge for major data center investors

I remember a few years ago when companies building data centers were bidding aggressively for production capacity. Will we see the same now, and could strong competition – also backed by long-term contracts – start to squeeze you more in the construction of large data centers?

We are now seeing this in sharp focus. The most difficult task will be delivering large data centers, because the orders are enormous.

Delivering ten servers to a private customer is something many players will still be able to manage. But delivering 50,000 GPUs in liquid-cooled servers, fully assembled and shipped, all of which require memory – that is already a challenge that very few will be able to handle.

When it comes to pricing, will the biggest problem be in cheaper consumer electronics?

No, this issue affects all orders – both servers and personal computers. However, where margins are higher, production will first be redirected toward those segments. As a result, the biggest shortages will be visible in the cheapest products. On the other hand, the share of memory in the total cost of such solutions is relatively smaller there, so price increases – relative to the most advanced systems – will be somewhat less pronounced.

We are already observing some interesting phenomena. For example, the number of people buying old computers on second-hand marketplaces has increased.

To obtain memory chips?

Yes. When buying equipment, you now have to pay attention to whether there is a note with an asterisk saying: “memory not included.”

Dell hardware pricing

Does this situation also affect your pricing policy?

Of course. We also, where justified, take targeted pricing actions. Everything today has to be constantly recalculated. It is a continuous process. That is why internally implemented automation at Dell is so important. Those who fail to do this will always be a week, two weeks, or even a month behind in their response.

But doesn’t this create something of a self-reinforcing bubble? If you expect prices to rise by 20 percent next month, are you already selling at higher prices today?

No. We have guidelines regarding today’s margin and today’s quote validity. It has to align within the current month. If a price increase comes, then a new calculation is made.

Margins are not increasing – costs are. We maintain the same margin level, but the final price for the customer goes up.

Can the situation be changed?

Why don’t the three memory manufacturers simply increase production capacity?

There are two factors here. First, increasing capacity is neither simple nor quick – it requires time and significant capital investment. Second, the current situation has a market dimension for the producers themselves. During the pandemic, they recorded strong revenues, then reduced production as demand fell, and today they are trying to rebuild results. But this is primarily a question for them.

So they simply failed to anticipate the trend?

Perhaps.

Do you see any other potential pressure points of this scale?

At this point, no. This is the biggest disruption in the industry in 30 years, at least in terms of sales and component availability. It is difficult to identify any previous situation that simultaneously affected all segments of the market. I am setting aside the dot-com bubble burst or the crisis following the collapse of Lehman Brothers – they were of a different nature.

So there is no way to bypass this?

No.

AI computers and the future of the economy

Looking more broadly – could these issues slow down trends we have been discussing for some time, such as the shift toward AI computers?

That is possible. Longer project delivery times and higher costs may lead to postponements or scaling back of initiatives. That is why it is important to carefully assess actual needs. In organizations with larger infrastructure resources, it is reasonable to examine utilization levels – both during the day and at night. In many companies, there is still room for optimization that has not yet been used.

For example: during the day we work on laptops, and at night we can train models on them?

A laptop offers limited capabilities. More potential lies in workstations, which remain unused outside working hours and can be used for other tasks.

At the same time, there is a risk of extending hardware life cycles. That is not beneficial – replacing devices is linked to access to newer, more secure solutions that are better prepared for emerging types of threats.

Delays on the horizon

If we assume that data center capacity must grow dramatically to train models or build sovereign AI at the level of states and companies – and it suddenly turns out that hardware is scarce or very expensive – could some projects be delayed?

It could also simultaneously accelerate the development of edge AI. In many use cases, smaller, local models are sufficient and can be run even on relatively modest devices, without the need to rely on large data centers. This opens up space for the development of smaller models, such as Poland’s Bielik. Data remains local, and implementation costs can be relatively low.

So could the current situation also have positive consequences?

Yes. Large data centers may be built more slowly, but at the same time the importance of local solutions is growing – solutions that require fewer resources and can be cheaper. Hardware is purchased once and used locally. Edge AI is becoming an important direction of development. The key principle is to process data as close as possible to where it is generated.

Administration and new challenges

Is this shift also visible in public administration when it comes to the AI debate?

Absolutely.

I’m asking because recently there was a press conference where ministers from both the current and previous governments spoke about sovereign AI and the need to build it. Yet the event itself was essentially funded by Google, which sparked some controversy.

Sovereignty comes in many forms. It is a very interesting narrative. In this context, the development of local solutions is crucial, as they increase control over data and democratize access to technology. Projects such as Bielik are an example of this trend. The scale of interest is significant – in one competition, we received 630 submissions, far more than expected. This points to substantial potential for the development of the edge AI ecosystem.

We ourselves are now considering what to do with this ecosystem – how to analyze it, categorize it, and develop it further, because the scale of responses genuinely surprised us.

Dell’s strategic shift

Could this lead to the creation of an AI services platform?

Yes. It demonstrates the potential of local solutions and their importance in building technological sovereignty. If these projects are properly developed and funded, they could find applications in public administration, healthcare, and business.

Does Dell have ambitions to go beyond being a hardware manufacturer?

These efforts have been underway for years. We are building an ecosystem of solutions. We do not create end-user applications, but we do develop software for our devices, which are becoming increasingly advanced. We also have an R&D base and plan to further expand it.

End of cloud?

Does this mean a departure from the cloud model?

No. Cloud solutions remain an important part of the market. At the same time, the importance of local and edge solutions is growing. This direction was already visible several years ago, even before the term “edge AI” became popular. It is now also being confirmed by the strategies of the largest cloud service providers.

The target model is the coexistence of different approaches: cloud, edge, and hybrid solutions. The idea that one of them will fully dominate the market is not supported by practice.

There was a time when anyone questioning the idea that everything would move to the cloud was labeled a boomer.

Many people said that, and it was a flawed narrative because it relied on an absolute statement: “everything will move to the cloud.” And such absolutes are, by definition, almost always wrong. I cannot recall any technology in my lifetime where one could reasonably say “100 percent will move there.”

Remember the slogan “PC is dead”? It was used for years to predict the demise of personal computers. Yet smartphones are still here, tablets are still here, and PCs are doing just fine. I believe in coexistence. The market grows like a cake. It is not a zero-sum game of “either this or that.”

Dell and the Ministry of National Defence (MON). Where is the cooperation heading?

You recently entered into cooperation with the Ministry of National Defense. Is this purely about providing hardware?

This is an agreement in which hardware sales come at the very end. At the beginning, the most important element is knowledge sharing. That is what we agreed on: exchanging information about ongoing projects and use cases, as well as data on vulnerabilities – so that we can respond faster and communicate more effectively both with the Cyber Defense Forces Command and with the market.

In the current environment, even trends such as memory shortages are important signals for such a partner. They affect procurement planning and the broader situation in critical infrastructure. We also learn from our partner, as the Cyber Defence Forces Command is running very interesting projects. This is a real two-way cooperation. One example is the Cyber Legion project, which brings together IT specialists. Among our engineers and partners, there is clear interest in participating in such initiatives.

The Cyber Defense Forces Command has also signed similar agreements with other companies, including Microsoft. Where are the boundaries of such agreements? In Dell’s case, is this mainly about knowledge exchange, or do you also provide concrete solutions, tools, and hardware for the military?

We are entering the area of commercial confidentiality here – what exactly we deliver and to whom. However, it is no secret that Dell is a significant infrastructure supplier for many NATO countries, across a wide range of applications.

Our equipment can be appropriately hardened, adapted to demanding environments, analyzed for electromagnetic resilience, and prepared for specialized use cases. We have such solutions. Therefore, this cooperation also includes the delivery of infrastructure.

Cyber threats and Poland’s public sector

Is there awareness of cyber threats in Poland’s public sector? After all, we have had a war on our eastern border for four years, and Poland remains a constant target of cyberattacks.

What exactly do we mean by “the Polish public sector”? It consists of tens, if not hundreds of thousands of people. There are institutions – such as model city offices – that are fully aware and do as much as they can. But it is difficult to expect the same level of preparedness from every entity.

As Dell, we undertake various supporting initiatives. For example, we run a project called “Future Made in Poland.” We support initiatives that strengthen our technological sovereignty. And we will continue to do so. We will hire people for R&D, develop a Polish factory, cooperate with Bielik, and produce hardware in Poland.

This is an interesting case, because from the perspective of a global company with strong American roots, one might ask whether there is a risk that one day someone will call and say: “Dell has made a global deal with Sam Altman, so forget about Bielik”?

No, that is not how it works. Dell provides a platform. What runs on that platform in a given country depends on the customers. I cannot imagine a situation in which someone would say: “only one solution is allowed on this platform.” That would be contrary to our strategy since the company’s inception. It would be like saying that a computer can only run one application. Dell has never done that, and never will.

In fact, that is precisely our strength. For decades, we have been technologically agnostic. We are present both in private data centers, hybrid environments, and with the largest cloud service providers. We have never claimed that one approach excludes another. These models are complementary.

Geopolitics and Dell’s business in Poland

Can geopolitics influence how technology operates—for example through political decisions?

It is a bit like assuming that a president could force a global cloud provider to shut down a service in a given country. We do not operate in that model. A server purchased in Poland and installed in a bank will function as long as the bank uses it. We do not have the ability to remotely switch it off.

Does what is happening in geopolitics affect your development plans in Poland – for example investments in a factory or R&D?

Dell has for years relied on a distributed network of production sites and R&D centers. There have been situations requiring rapid response, even unrelated to politics. I remember a case of a tornado in the United States that damaged a logistics center. At that time, the factory near Łódź ramped up production within a matter of days, launched additional lines, and shipped equipment to the US by air.

Dell’s investments in Łódź

To what extent is production capacity in Łódź being utilized today?

That changes almost day by day. Toward the end of quarters, production usually increases as more orders come in. It is a recurring pattern. In general, the utilization level depends on current demand.

Do you plan further expansion?

We have already launched production lines for AI servers and complete liquid-cooled systems. This is a completely different level of infrastructure compared with traditional air-cooled servers. It requires the installation of advanced cooling systems and specialized testing stations. These investments are preparing the factory for further growth. A logistics center has also been built directly next to the facility.

Is it already operational?

Yes, it is operational, although the ramp-up is gradual.

Was there an official opening?

We do not organize large-scale events. We simply expand the infrastructure and continue working. Information about the start of construction was announced earlier, but it was communicated by the investment partner responsible for the project.

What Dell is planning next

And further investments?

If Michael Dell decides so, there will be.

So there are already projects waiting on the desk?

The company is efficiently managed from its headquarters in Austin. We focus on demonstrating Poland’s potential. Without developing operations and building local competencies, it would be difficult to justify further investments.

I ask because there are market voices suggesting that American companies are not always free to publicize foreign investments.

I have not come across any such restrictions.

Are Michael Dell and Donald Trump friends?

I am not aware of their personal relationship. What I do know is this: Michael has been running the company for over 40 years and has worked with many Presidents of the United States. He has maintained relationships with all of them, because the US government is one of the company’s largest customers. And that remains the case today.

The Trump administration and its impact on the world

Recently there was a major conflict between Anthropic and public authorities regarding the use of AI models for military purposes. Is this even an issue for Dell?

There is a boundary to the responsibility of a hardware manufacturer. This question should be addressed to the creators of models – such as OpenAI or Anthropic – regarding their values, ethical boundaries, and what they want or do not want to support. This is a very important debate in today’s world, but it does not directly concern us.

I also ask because in the case of Starlink and Elon Musk, it became clear that hardware and infrastructure issues can also raise concerns.

I understand, but we do not see that here. We have clear guidelines. It is known to which countries we do not sell equipment because embargoes apply – and that is our boundary.

Record-breaking Dell results and layoffs

Dell recently reported financial results – record-breaking on a global scale. However, I recall announcements from a year ago about major layoffs at the company, involving tens of thousands of people. How do these two phenomena connect, and where is the company heading?

I would separate what is a fairly natural process – and not only at Dell but across the entire IT industry – namely workforce reductions following a period of rapid growth. During the COVID-19 pandemic, there was a very strong increase in hiring, followed by a slowdown, and today we are seeing more of a correction. Looking at it more broadly, all industries are becoming more automated. They are increasingly using algorithms that streamline processes. This is a natural direction – efficiency simply improves.

As for Dell’s results, several points are worth noting. Revenue growth year over year was 19 percent. For our industry and the level of competition, that is very strong. It was a very good year for Dell, and we expect the next one to be good as well.

The business model is changing.

For 42 years since its founding, Dell has been associated primarily with PCs. For the first time in the company’s history, revenue from the PC segment was lower than from other areas – namely servers, data centers, storage, and similar solutions. Growth in these segments was around 40 percent, while the PC segment declined by approximately 5 percent.

Is AI the new engine of Dell?

Will this become Dell’s main growth engine in the future?

Growth in the PC market may come at the expense of competitors, but the situation in the server segment and data storage solutions looks different. We are processing ever larger volumes of information, artificial intelligence has emerged, and we are seeing a clear acceleration. We have positioned ourselves well in this space. In the fourth quarter alone, we received more orders for servers and AI-related solutions than in the previous three quarters combined. The scale of growth is very significant.

That is why Dell’s results are currently very strong. We expect this trend to continue, also driven by internal projects that are transforming how the company operates and leveraging automation and advanced algorithms. This process will continue.

How does this look in Poland?

We do not publish country-level data. However, I can say that we are seeing stronger growth in the infrastructure segment – similar to the global Dell trend – than in the personal computer segment. With one caveat: last year we executed large orders from the education sector. These were one-off in nature and therefore do not repeat annually. Excluding them, PC segment growth would have been stable.

Poland vs. the world

In Poland and in the server market – do you also see stronger growth compared with other European countries?

Poland is certainly benefiting from the funds currently being spent – both from the National Recovery Plan (KPO) and other programs. This has stimulated the public sector, but it is not the only driver. If we compare ourselves with other countries, it has been a good year for us. From what I know, other large European markets have also recorded very strong results, so we are not an exceptional outlier here. The strongest growth is visible in the server segment.

Key Takeaways

1. We are facing a sharp, structural shortage of computer memory chips driven by a sudden surge in demand fueled by the rapid development of artificial intelligence and the construction of hyperscale data centers. Manufacturers, by focusing on the most profitable segments – such as HBM memory used in AI applications – have reduced the availability of components for other uses. As a result, the market is currently able to meet only around 70 percent of total demand, leading to accumulating backlogs and a deepening disruption across multiple industries.
2. Rising memory costs are translating into higher prices for entire devices, with monthly increases in some cases reaching tens of percent. At the same time, delivery lead times are lengthening and may extend to several months. The scale of the issue is forcing companies to revise budgets, timelines, and procurement strategies, as previously assumed project schedules are no longer valid.
3. Supply chain disruptions may slow the development of large-scale AI projects and data centers, while at the same time favoring the growth of local solutions such as edge AI. Companies are being pushed to optimize the use of existing infrastructure and adopt more flexible investment planning. At the same time, experts emphasize that the current situation may persist for at least two years, as increasing semiconductor production capacity is a long and capital-intensive process.