Estimated reading time: 5 minutes
Why memory prices are soaring, what AI data centres have to do with it, and why businesses and consumers may be feeling the effects for years to come.
This article is a non-technical view on the subject. If you want to understand a bit more regarding the numbers and technology behind this?
Read our full analysis of the RAM crisis.
Artificial intelligence is changing the technology industry in ways most of us can see.
ChatGPT. Copilot. AI assistants. Automation.
But there is another consequence of the AI boom that is much less visible.
It is helping to drive up the cost of some of the basic components inside servers, laptops, phones and other devices.
At IntraLAN, we have seen like-for-like server quotations rise to around three times the level we were seeing approximately 18 months ago. Some server memory and SSD configurations have increased by around four times.
The reason is not simply inflation.
AI is consuming an extraordinary amount of physical computing infrastructure, and the effects are starting to reach the rest of the technology market.
🧠 AI needs a lot more than software
When we use an AI service, it is easy to think of it as something happening somewhere in “the cloud”.
Behind the scenes are huge data centres filled with powerful computers.
Those computers need processors, memory and storage.
Lots of it.
A typical business laptop might contain 16GB of memory.
A single modern AI computing rack can contain around 30TB.
That is roughly the same amount of memory capacity as around 1,900 ordinary 16GB laptops.
And that is one rack.
xAI’s original Colossus AI supercomputer deployment contained around 100,000 specialist processors and approximately 8 petabytes of high-speed memory.
In simple terms, that is equivalent in raw memory capacity to around 500,000 laptops fitted with 16GB each.
That gives some idea of the scale involved.
💾 It isn’t only RAM
AI systems also need enormous amounts of fast storage.
The same wider flash-memory industry that supplies SSDs for business servers and laptops is increasingly supplying giant AI data centres.
Industry research suggests enterprise SSDs accounted for almost half of global NAND memory shipments in the second quarter of 2026, up sharply from the previous year.
Manufacturers only have so much production capacity.
If more of that capacity is being absorbed by enormous AI projects, the pressure can eventually show up in the price of an ordinary server, laptop or smartphone.
🌍 And the build-out is still accelerating
There are few signs that the companies involved are preparing to slow down.
OpenAI has been securing infrastructure for its Stargate programme on a scale measured in gigawatts of power.
Meta is investing more than $50 billion in its Louisiana data-centre development.
xAI has discussed expanding towards one million AI processors.
And in Utah, the proposed Stratos development covers around 40,000 acres, with plans that could eventually involve up to 9GW of power generation.
For context, that is more than twice Utah’s recent average electricity consumption.
That does not mean the full project will necessarily be built.
But it demonstrates how dramatically the scale of AI infrastructure has changed.
Large data centres used to be discussed primarily in megawatts.
Today, some proposed AI developments are being discussed in gigawatts.
💰 The biggest technology companies appear determined not to fall behind
Microsoft, Alphabet, Meta and Amazon are all investing enormous sums in data centres, processors and related infrastructure.
In some recent quarters, that investment has consumed most, or even more than all, of the cash generated by their underlying businesses.
That does not mean these companies are in financial difficulty. They remain some of the richest and most profitable organisations in the world.
But their behaviour suggests something important.
They currently appear to regard falling behind in AI as a greater competitive risk than spending too much to stay in the race.
As long as that remains true, the demand for the hardware underneath AI is unlikely to disappear quickly.
🏭 Why can’t manufacturers simply make more?
This seems like the obvious answer.
If prices rise, build more factories.
Unfortunately, semiconductor factories cost billions and can take years to construct and bring into production.
Samsung, SK hynix and Micron dominate the global market for computer memory, and all three are investing heavily.
But some of the additional manufacturing capacity being planned today will not arrive until 2028, 2029 or beyond.
There is another difficulty.
The technology creating much of today’s demand barely existed in its current form four years ago.
Manufacturers are therefore being asked to make multi-billion-pound decisions about factories that may take years to become productive, based partly on demand from a market that can transform completely in less time than it takes to build them.
AI demand can accelerate in months.
Manufacturing capacity takes years.
🎈 What happens if the AI boom slows?
This is why manufacturers cannot simply build unlimited new capacity.
If they build too little, shortages and high prices continue.
If they build too much and AI demand cools, they could be left with billions invested in factories producing more memory than the market needs.
Perhaps AI becomes more efficient.
Perhaps businesses use less of it than expected.
Perhaps new technology changes the kind of components required.
The memory industry has experienced boom-and-bust cycles before, so manufacturers have good reason to be cautious.
⏳ How long could this last?
Nobody can say with certainty.
Some industry forecasts suggest supply could begin improving during 2027 and 2028 as new manufacturing capacity becomes available.
Other reports suggest some additional production may not reach meaningful scale until closer to 2030.
That does not mean prices will remain at today’s levels for another four years.
It does mean this is unlikely to be a problem that simply disappears over the next few months.
The industry is thinking in years rather than weeks.
🛒 Why should businesses and consumers care?
Because memory is inside almost everything.
Servers use it.
Laptops and desktop PCs use it.
Phones and tablets use it.
Games consoles use it.
Modern cars increasingly use large amounts of memory and storage for entertainment, connectivity, driver-assistance systems and other software features.
So even if you never buy an AI system, you can still end up paying for the demand AI creates.
For businesses, a server or PC replacement project that looked straightforward 18 months ago may now require a very different budget.
For consumers, the impact could appear as higher prices, smaller base specifications, more expensive storage upgrades or manufacturers keeping existing products on sale for longer.
🧭 What should businesses do?
The answer is not to panic buy.
But it does mean planning matters more than it used to.
If you know a server, storage system or fleet of PCs is likely to need replacing over the next 12 to 24 months, it is worth starting that conversation earlier.
Check warranties and end-of-support dates.
Understand how much useful life remains in your existing equipment.
And do not automatically assume that delaying a purchase means the technology will be cheaper when you eventually buy it.
Sometimes waiting will make sense.
Sometimes buying earlier will.
The important thing is to make that decision deliberately.
The bigger picture
The current memory crisis is not really a story about RAM.
It is a story about the physical scale of artificial intelligence.
AI may look like software on a screen, but behind it are data centres, processors, memory, storage, factories, energy infrastructure and billions of pounds of investment.
When technology is being built on that scale, the effects do not stay inside the AI industry.
Eventually they reach the rest of us.
That is why the price of your next server, laptop, phone, console or even car may be influenced by an AI race taking place in data centres thousands of miles away.
If your business is planning a hardware refresh over the next 12 to 24 months, we are happy to share what we are seeing in the market and help work through whether acting now, waiting or taking a different approach makes the most sense.
Ready to start the conversation?
At IntraLAN, we’re happy to share what we’re seeing in the market and help work out whether acting now, waiting, moving some workloads elsewhere or taking a different approach makes the most sense for your business.
