Why Your Next Laptop Costs More: The Memory Shortage Explained

If you priced a laptop two years ago, walked away, and came back this summer, you probably had the same reaction most people have had: the machine you remember is not the machine you can afford. Same brand, same screen size, a configuration that used to be unremarkable, and a number at the end that is meaningfully higher.

This is not inflation in the general sense, and it is not manufacturers being opportunistic, although some certainly are. It traces back to a specific bottleneck: memory.

Where the squeeze comes from

AI accelerators need high-bandwidth memory, and a great deal of it. HBM is manufactured by a small number of companies on production lines that took years and enormous capital to build. You cannot spin up more of it in a quarter because demand spiked.

When those lines get committed to the highest-margin customers, and data centre buyers are the highest-margin customers by a wide margin, everything else queues behind them. Conventional DRAM and NAND compete for related capacity, engineering attention and packaging. The knock-on effect reaches the parts that go into phones, laptops and desktops.

Why it shows up unevenly

Device makers have some room to absorb cost, and they use it selectively. The pattern that emerges looks like this:

• Entry configurations get thinner. The base model keeps its headline price but ships with less memory or slower storage than its predecessor.
• Upgrade tiers get steeper. The step from the base configuration to the sensible one carries a larger premium than it used to.
• Soldered components become the norm. When a part is expensive, making it non-upgradeable turns a one-time purchase decision into a permanent one.

That last point is the one that costs people the most money over time, and it is the easiest to miss at the checkout page.

How to buy sensibly while this lasts

Buy the memory you need up front

The traditional advice was to buy the cheapest configuration and add RAM yourself later. On most current machines you cannot, and where you can, the aftermarket price is no longer the bargain it once was. Decide what you need for the life of the machine and pay for it once.

Be honest about how much you need

The counterweight to the advice above is that most people overestimate. If your work is documents, browser tabs, video calls and the occasional photo edit, a mid-tier configuration will still feel fast in five years. The people who genuinely need more, running virtual machines, editing long-form video, compiling large projects or working with local models, usually know it already.

Consider last generation deliberately

Previous-generation machines are frequently the best value on the market right now, because they were specified and priced before the squeeze tightened. The performance difference is usually smaller than the price difference.

Take refurbished seriously

Manufacturer-refurbished stock with a real warranty is a different proposition from a marketplace listing. When component prices are elevated, the gap between refurbished and new widens in the buyer’s favour.

How long does this go on?

Nobody selling you a laptop knows, and anyone who claims certainty is guessing. Memory has always been cyclical: shortage drives investment, investment eventually produces oversupply, prices fall, investment stops, and the cycle repeats. The current cycle is unusual mainly because AI demand arrived faster than the industry’s normal planning horizon.

What is reasonably safe to say is that new capacity takes years, not months, to come online. If your current machine works, waiting is a legitimate strategy. If it does not, buy carefully rather than buying big.