If you've been shopping for a PC upgrade recently, you may have noticed something unusual: RAM, SSDs and graphics cards aren't necessarily becoming cheaper in the way we've traditionally expected technology to.

In parts of the market, prices are actually increasing.

There isn't one single reason behind it. Instead, the global technology industry is being reshaped by a combination of artificial intelligence, enormous data-centre investment, changing semiconductor manufacturing priorities and constrained memory production capacity.

And although your gaming PC isn't directly competing with an AI supercomputer for the exact same RAM module, both ultimately depend on many of the same semiconductor manufacturers and production resources.

The AI Boom Is Changing the Memory Industry

Modern AI systems require extraordinary amounts of memory.

The GPUs and specialised processors used to train and run AI models rely heavily on High Bandwidth Memory (HBM), while the servers surrounding those processors require large amounts of conventional server DRAM and increasingly large quantities of high-performance storage.

Cloud providers, technology companies and data-centre operators are consequently investing heavily in AI infrastructure.

For memory manufacturers, this represents one of the fastest-growing and most valuable markets in the technology industry.

That changes manufacturing priorities.

Companies producing DRAM must decide how available fabrication capacity is divided between products including:

• High Bandwidth Memory for AI accelerators
• Server DRAM
• Desktop and notebook RAM
• Graphics memory
• Mobile memory
• Specialised memory products

Producing more memory for one rapidly expanding market can therefore reduce the manufacturing capacity available elsewhere.

Industry research from TrendForce indicates that the growing allocation of DRAM wafer capacity toward HBM and AI-related products is contributing to tight conventional DRAM supply.

Why Is RAM Getting More Expensive?

Desktop and notebook memory sit directly inside this changing supply environment.

AI servers aren't buying the same retail DIMMs you would install in your gaming PC. The connection happens much further upstream.

The world's major memory manufacturers use their fabrication plants to produce several different types of DRAM. When more production is allocated to high-demand server and AI applications, less manufacturing capacity is available for other markets.

HBM also requires significant wafer capacity compared with conventional DRAM, making the situation more complicated than simply producing additional chips.

There are also physical limits to how quickly manufacturers can respond.

Semiconductor fabrication plants cost billions to build. Construction, specialised equipment installation, process qualification and production ramp-up can take years.

This means manufacturers cannot simply switch on another production line whenever demand suddenly increases.

Current industry projections suggest some additional DRAM production will begin coming online during 2027, but substantial increases in available output may take considerably longer.

SSDs Are Facing Their Own Supply Pressures

SSDs primarily use NAND Flash rather than DRAM for storage, so the NAND market operates somewhat differently.

But AI is affecting storage as well.

AI data centres don't only need enormous processing power. They also need enormous amounts of fast storage.

Training datasets, AI models, databases, inference workloads and cached information all have to be stored somewhere.

That is driving strong demand for enterprise SSDs and high-capacity NAND Flash.

Industry forecasts currently suggest NAND Flash contract pricing could increase by approximately 10–15% quarter-on-quarter during Q3 2026.

AI inference and large-scale data-centre deployments remain important drivers of this demand.

There is, however, potentially some good news further ahead.

Unlike DRAM, additional NAND manufacturing capacity could eventually push the market toward a healthier supply position. Current industry projections suggest NAND supply conditions could loosen during the second half of 2027.

That could eventually create downward pressure on storage pricing, although future retail prices will also depend on exchange rates, shipping costs, component demand and local market conditions.

What Does This Have to Do With Graphics Cards?

Graphics cards add another layer to the situation.

Modern graphics cards use high-speed graphics memory such as GDDR6 and GDDR7, meaning wider memory-market conditions can influence the cost of producing GPUs.

But there's an even bigger connection.

Many companies involved in supplying consumer graphics hardware are also heavily involved in the rapidly expanding AI accelerator market.

AI infrastructure has become one of the most valuable areas of the semiconductor industry.

Advanced semiconductor manufacturing capacity, packaging technologies, memory supply and engineering resources are consequently being pulled toward AI products.

A graphics card also contains considerably more than the GPU itself.

• GPU silicon
• GDDR graphics memory
• Advanced semiconductor packaging
• Power delivery components
• Cooling systems
• PCB manufacturing
• Shipping and distribution costs

Pressure anywhere along this supply chain can influence the eventual retail price.

That helps create a GPU market where the traditional assumption that "last year's performance will automatically become much cheaper this year" doesn't necessarily hold.

Watch: The Gaming Market Is Starting to Break

The following video takes a broader look at how rising memory prices, hardware costs and AI-related demand are beginning to affect the gaming and consumer hardware market.

Why Can't Manufacturers Just Make More Chips?

Semiconductor manufacturing operates on a completely different scale from most industries.

A new advanced semiconductor fabrication facility can require billions in investment and several years before reaching meaningful production volumes.

Manufacturers also have to predict what demand will look like years into the future.

Build too little capacity and shortages occur.

Build too much and the industry can experience oversupply, collapsing prices and extremely expensive factories operating below capacity.

The memory industry has experienced both sides of that cycle repeatedly.

That makes manufacturers cautious about rapidly expanding capacity simply because current prices are high.

Everyone Is Competing for Advanced Manufacturing

The AI boom also isn't happening in isolation.

The semiconductor industry simultaneously needs to supply:

• Gaming PCs
• Business computers
• Notebooks
• Smartphones
• Servers
• AI accelerators
• Networking equipment
• Vehicles
• Industrial systems
• Consumer electronics

Many of these industries ultimately depend on a relatively small group of advanced semiconductor manufacturers and memory suppliers.

The result is global competition for manufacturing capacity.

AI companies and cloud providers simply happen to be among the customers currently investing the most aggressively.

Are RAM and SSD Prices Going to Keep Increasing?

Not necessarily forever.

Technology markets are famously cyclical.

Higher prices encourage manufacturers to invest in additional production capacity. Eventually supply can catch up with demand, inventories increase and prices begin falling again.

But different components may recover at different times.

Current industry forecasts suggest DRAM could remain structurally tight throughout 2027, particularly as AI servers continue increasing their requirements for HBM and conventional server memory.

NAND Flash could potentially loosen sooner as additional manufacturing output reaches the market.

Graphics card pricing is even harder to predict because GPU prices depend on considerably more than memory alone.

What Does This Mean If You're Planning a PC Upgrade?

Waiting for PC components to automatically become cheaper may not always work in today's market.

If your computer is regularly running out of memory, upgrading RAM can improve multitasking and memory-heavy applications.

If storage is your bottleneck, upgrading from a traditional hard drive or older SATA SSD to an NVMe SSD can dramatically improve system responsiveness and loading times.

If gaming or GPU-accelerated applications are the priority, compare graphics cards according to the performance you actually need rather than waiting indefinitely for a particular model to reach a historical price point.

The Bigger Picture

Artificial intelligence might feel like something happening inside enormous data centres thousands of kilometres away.

But its impact is increasingly reaching the hardware sitting on our desks.

The unprecedented investment going into AI infrastructure is changing where semiconductor manufacturers allocate production, which types of memory receive priority and how quickly supply can respond to demand.

It affects RAM.

It affects SSDs.

It affects graphics cards.

And ultimately, it can affect the cost of building or upgrading a PC.

The PC hardware market isn't disappearing, but the forces determining what our components cost are changing.

For the foreseeable future, AI infrastructure is likely to remain one of the biggest forces shaping the global memory and semiconductor market.