If you’ve tried buying a laptop, server, or even a simple RAM upgrade recently, you’ve likely felt the pinch: prices are rising, availability is dropping, and once‑easy upgrades now take weeks to source. This isn’t a coincidence or a temporary blip, it’s the result of massive structural changes in the tech industry, driven largely by the global AI boom.
Memory and storage manufacturers have shifted focus, supply chains are stretched thin, and older standards are disappearing faster than anyone predicted. Let’s break down exactly what’s going on, why it’s costing you more, and how long the pressure is likely to last.
The AI Gold Rush is Consuming Global Supply
The meteoric rise of AI, from training models to deploying data centres globally, has created an unprecedented surge in demand for high‑performance memory and storage.
Why AI needs so much memory
Modern AI workloads rely heavily on HBM (High Bandwidth Memory) and DDR5 server memory, both of which are significantly more complex and expensive to produce than standard PC RAM. Training large AI models requires:
- Massive parallel processing
- Extremely high memory bandwidth
- Large amounts of VRAM and system RAM
- Fast storage layers for datasets
Data centres that used to expand gradually are now scaling aggressively. Some hyperscaler AI clusters consume millions of gigabytes of memory at a time, effectively absorbing a huge chunk of global production.
The impact on consumers
When major AI companies reserve long‑term supply directly with manufacturers, the downstream effect is simple:
- Less memory and storage enters the open market
- Retail prices climb
- Consumer and SMB products are deprioritised
In short, AI is getting first pick, and the rest of the world must make do with what’s left.
Manufacturers are Chasing Profits, and Enterprise Hardware Pays Better
Samsung, Micron and SK Hynix, the giants responsible for most of the world’s memory, are businesses first. And right now, the biggest profits lie in supplying:
- AI data centres
- Cloud hyperscalers
- Enterprise‑grade server platforms
These customers pay significantly more per GB for advanced memory types like HBM and DDR5 RDIMMs. So, manufacturers are shifting their limited wafer capacity accordingly.
The ripple effect
This shift means:
- Reduced output of DDR4 and consumer‑focused DDR5
- Lower availability of budget and mid‑range SSDs
- Higher unit costs, as reduced mass production raises per‑unit pricing
- Longer lead times for laptop and desktop manufacturers
OEMs, even big names, are now competing for smaller allocations, and that scarcity trickles down to retail buyers.
The Move from DDR4 to DDR5 is Straining Supply
Every generational shift creates tension, but the switch from DDR4 to DDR5 has been particularly disruptive.
Why the transition is bumpy
1. DDR5 uses different power management (PMIC chips) which have faced their own supply shortages.
2. Manufacturers must retool factories, temporarily reducing output of both DDR4 and DDR5.
3. Older DDR4 machines are still widely used, especially in business fleets, and demand for replacement memory remains high, but production is dropping.
This creates a mismatch:
- Lots of ageing systems still need DDR4
- Much less DDR4 is being produced
The result?
DDR4 prices, instead of falling at end‑of‑life (as they normally would), are rising sharply.
Storage Prices are Rising, NAND Flash is in Short Supply
SSDs depend on NAND flash chips. And right now, NAND is being squeezed by the same forces:
- AI data centres need enormous amounts of high‑end NVMe storage
- Suppliers are prioritising data centre‑grade SSDs with better profit margins
- Consumer SSD production hasn’t kept up with demand
Analysts have already reported 5–10% increases, with premium SSDs climbing even faster.
Enterprise storage demand is exploding
AI data centres don’t just need memory, they need:
- Petabytes of NVMe drives
- Ultra‑fast PCIe Gen 4 and Gen 5 storage
- High endurance enterprise SSDs
This means manufacturers funnel more NAND into enterprise lines, leaving consumer SSDs under‑supplied.
Why Manufacturers can’t just “Make More”
It’s natural to ask: why not just increase production?
Because scaling semiconductor manufacturing is one of the hardest, most expensive tasks on Earth.
Building new fabs requires:
- Over £10 billion of investment
- 12–24 months of construction and calibration
- Highly specialised equipment
- Significant negotiation for raw materials (wafers, chemicals, photomasks)
- Long certification cycles with major customers
Memory fabs run at extreme precision, and even tiny variances can ruin entire batches. This means:
- Expanding too quickly risks quality problems
- Bringing new facilities online takes years
- Output cannot adjust rapidly to demand
Manufacturers are cautious because memory markets have historically been volatile, investing too aggressively can lead to oversupply and financial losses.
What This Means for You
Expect:
- Higher prices for laptops, servers, and upgrades
- Reduced availability of DDR4, budget SSDs, and older components
- Longer lead times for business hardware
- More expensive repairs, as replacement parts become rarer
- Increased cost of building your own PC
For businesses planning refresh cycles or capacity expansion, budgeting will need to adjust.
Practical advice
If you know you’ll need:
- Additional RAM
- New servers
- Storage expansion
- Laptop fleet upgrades
Start planning and purchasing sooner rather than later.
The Bottom Line
AI is fundamentally reshaping the tech supply chain. With data centres absorbing unprecedented amounts of memory and storage, manufacturers are prioritising high‑end, high‑profit components, leaving fewer parts available for everyday users.
Prices will likely continue rising until new fabrication plants come online and production catches up, a process that may take 12–24 months or more.
For now, whether you’re an IT manager, a business owner, or an everyday user:
buy early, plan ahead, and expect continued volatility in the hardware market.
