The RAM Crisis: What My Search for DDR5 Revealed About the AI Hardware Boom
What began as a search for a reasonably priced 32GB DDR5 kit became a lesson in how AI infrastructure, constrained memory supply and a difficult resale market are changing the experience of building a PC.
For years, memory was one of the more predictable parts of building a PC. Decide how much capacity you need, check the speed and latency, confirm motherboard compatibility, and compare a few reputable kits. It was rarely the component that held up an entire build.
That has not been my recent experience.
I have been searching for a reasonably priced 32GB DDR5 kit for a future system. My shortlist has included G.Skill Trident Z5, Corsair Vengeance DDR5-6400 and Kingston Fury Beast DDR5-6000 CL36. None of those choices is unusually rare or exotic. Yet the process has involved repeatedly checking retail prices, following eBay auctions, contacting Gumtree sellers and weighing whether a used kit offers enough value to justify the additional risk.
Some promising listings have been collection only, which does not work for me. Others have lacked the delivery or payment protection I would want for a private purchase. A few have looked attractive until their price was compared with the warranty and return rights available from a retailer. Even when the specification is right, the overall transaction often is not.
I am still searching.
What initially felt like an inconvenient personal buying experience is part of a much larger change. The memory market is under pressure from the rapid expansion of artificial-intelligence infrastructure, and the effects are moving beyond data centres into the products that ordinary consumers and independent PC builders buy.
More Than a Temporary Price Spike
Memory has always been a cyclical industry. Periods of oversupply push prices down; production cuts and rising demand send them back up. It would therefore be easy to dismiss today's increases as another short-lived phase.
The present pressure appears more complicated.
Modern AI systems require enormous amounts of memory. High-bandwidth memory, or HBM, is especially important because it feeds data to AI accelerators at far greater rates than conventional desktop memory. The DDR5 modules used in a home PC are not the same products as the HBM packages used beside advanced accelerators, but they belong to the same wider semiconductor industry. Manufacturers must decide where to direct investment, engineering resources and production capacity.
When the strongest demand and margins are in AI infrastructure, advanced memory receives priority. That can tighten other parts of the market and increase costs throughout the supply chain.
This is no longer only an enthusiast complaint. In June 2026, Reuters reported that organisations representing automakers, retailers, electronics firms and other industries had warned of an urgent imbalance in memory supply. Their concern was that AI data centres were consuming a substantial share of available capacity, contributing to higher costs and reduced availability for consumer-facing industries.
In September, the Financial Times described the wider effect across consumer electronics, including computers, smartphones and games consoles. Separate Reuters reporting showed how constrained HBM supply was also raising the cost of AI accelerators themselves.
These reports do not mean that every expensive DDR5 listing has a single cause. Retail prices are influenced by inventory, exchange rates, distribution, product positioning and local demand. They do show, however, that the pressure I am seeing as a buyer exists within a genuine global supply problem.
What the Crisis Looks Like to an Individual Buyer
Industry analysis tends to focus on fabrication capacity, contract prices and global demand. At the level of one PC build, the consequences are more ordinary: more time spent searching, fewer convincing options and a higher cost for making the wrong decision.
My requirement is straightforward. I want a matched 2 × 16GB DDR5 kit from a reputable manufacturer, with sensible speed and latency, dependable compatibility and a realistic price. The memory may begin in one system and later move into another, so long-term usefulness matters more than chasing the highest number on a product page.
That is why a listing cannot be judged by price alone.
A used kit must be sufficiently cheaper than a new one to account for reduced certainty. The exact part number matters because products sold under the same family name can have different timings, voltages or internal revisions. The seller's history matters. So do proof of operation, remaining warranty, delivery arrangements, return rights and the protection offered by the payment method.
The result is a strange contradiction. RAM is a standard, mass-produced component, yet finding the right kit can feel like sourcing a specialist part. Search results create the impression of abundant choice, but the practical market becomes much smaller once unsuitable specifications, inflated prices, collection-only listings and weak buyer protection are removed.
For a UK buyer who cannot travel across the country to collect a component, location matters too. A bargain hundreds of miles away is not really available if the seller will not post it. An informal delivery arrangement may solve the distance problem while creating a much larger problem if the item never arrives or is not as described.
Why Faster Is Not Automatically Better
A difficult market makes technical judgement more important. When prices are elevated, an unsuitable purchase is more costly to correct.
DDR5 is commonly advertised through its transfer rate: 6000MT/s, 6400MT/s or higher. That figure is useful, but it is not a complete measure of quality or real-world performance. Timings, voltage, the processor's memory controller, motherboard layout and firmware support all influence whether a kit can run reliably at its advertised profile.
The advertised XMP or EXPO speed is also an overclocked profile rather than an unconditional guarantee for every hardware combination. A faster-rated kit is not automatically the better purchase if the platform cannot use that speed reliably or the workload gains little from it.
Capacity and upgrade strategy can matter more. For a gaming PC, 32GB is a strong practical target. In a virtualisation or cybersecurity lab, where several Windows and Linux virtual machines may run at once, having a clear path to 64GB can be more valuable than a small improvement in a synthetic benchmark.
This is why I have resisted buying the first kit that merely looks impressive. Scarcity creates urgency, and urgency encourages compromise. It becomes easy to pay too much, accept weaker protection or buy specifications that do not match the actual workload.
Patience is frustrating, but an expensive mistake would be worse.
The Hidden Cost of Building Independently
Large technology companies secure supply through high-volume, long-term agreements. Independent builders meet the market at the opposite end of the chain, after changes have passed through chip manufacturers, module vendors, distributors, retailers and resellers.
The cost is measured in more than pounds. It includes the time spent monitoring listings, checking part numbers, messaging sellers, comparing warranties and deciding whether a deal is safe. Higher memory costs can also reshape the rest of a build: an SSD upgrade may be delayed, a cooling budget reduced or an entire platform change postponed.
The second-hand market would normally offer relief, but it reacts to the same conditions. Sellers compare their asking prices with higher retail prices. Desirable kits attract attention quickly. Genuine value becomes harder to distinguish from unnecessary risk.
In that environment, the cheapest listing is not always the cheapest purchase.
Buying Carefully Without Panic-Buying
The current market rewards preparation rather than haste. These are the principles guiding my own search:
Define the workload first. Decide the required capacity, acceptable speed range, upgrade path and maximum budget before browsing listings.
Check the complete part number. Do not rely only on a product-family name or headline speed. Confirm the capacity, module count, timings, voltage and supported memory profile.
Check platform compatibility. Review the motherboard documentation and memory support list where available, while remembering that no qualification list can include every kit on the market.
Calculate the complete cost. Include delivery, warranty, return rights and payment protection—not only the advertised price.
Keep a flexible shortlist. Several reputable alternatives create more opportunities than waiting for one exact model.
Do not let market anxiety make the decision. Predictions of further increases do not turn every available listing into good value.
These principles cannot fix global supply, but they reduce the chance of responding to a difficult market with a poor purchase.
A Small Component With a Much Larger Story
My DDR5 search remains unfinished, but it has changed how I view the PC hardware market.
The AI boom is usually discussed in terms of large language models, accelerators and billion-pound data-centre projects. Its effects are also visible in much smaller decisions: a manufacturer adjusting a specification, a retailer increasing a price, an upgrade being delayed or a PC builder refreshing second-hand listings in search of a fair deal.
That connection is what makes the current RAM crisis worth examining. It is not simply a story about one component becoming expensive. It shows how rapidly infrastructure demand at the top of the technology industry can influence individual buyers at the other end.
I will eventually find the right kit. My goal is not to buy the most impressive specification on the box or win a benchmark. It is to secure reliable, compatible memory at a defensible price, through a transaction that offers appropriate protection.
Until then, the search continues—and so does a market shift far larger than any single PC.
Editorial note: Product availability and pricing can change quickly. This article reflects my personal buying experience and the market conditions reported as of 11 September 2026.