If you've priced out a server, laptop, or even a humble VPS lately, you've been a victim of the great RAM grab.

What used to be a boring, invisible component is suddenly one of the most fought-over resources in tech. AI giants are hoovering it up at an industrial scale, and everyone else is scrambling for what's left. The result is obvious.

Higher prices, thinner margins, delayed launches, and a growing sense that we're sliding into full-blown "RAMageddon."

This week made it painfully concrete. German hosting provider Hetzner announced price adjustments, with dedicated servers rising by an average of roughly 3%. That's not catastrophic, but it's noticeable. The bigger shock was cloud pricing, which in many cases is jumping 30–35%. That's the kind of increase that forces startups and small businesses to rethink their budgets overnight.

Meanwhile, a Reddit user reported receiving an OVH renewal offer quoting a 55.1% higher price starting in April. Others said VPS pricing had effectively doubled. When you see hikes north of 50%, it stops looking like routine cost adjustments and starts looking like structural stress in the supply chain. So, what's actually going on? Plain old supply and demand.

Supply Bottleneck

AI infrastructure is brutally memory-hungry. The servers powering copilots, chatbots, generative models, and training clusters don't just need GPUs — they need enormous pools of DRAM and high-bandwidth memory to keep those accelerators fed. Every new generation of AI hardware tends to consume even more RAM per system.

Companies like Alphabet (NASDAQ:GOOG), Microsoft (NASDAQ:MSFT), and OpenAI have the balance sheets to secure massive memory contracts at almost any price. When they build out data centers, they're buying at a scale that dwarfs traditional enterprise and consumer demand. That supply used to flow into laptops, smartphones, TVs, and budget cloud servers. Now it's being redirected into AI clusters.

The key issue is a highly concentrated supply market. Roughly 90–93% of global DRAM production comes from just three companies: Samsung Electronics (OTCPK: SSNLF), SK Hynix, and Micron Technology (NASDAQ:MU).

When demand surges, there aren't dozens of competitors ready to step in. And these firms are cautious, so expansions happen slowly and strategically.

Analyst firms are now sounding alarms. Counterpoint Research reported 40-50% surges in Q4 2025 and expects further 40-50% increases in Q1 2026 and 20% in Q2 2026.

Meanwhile, TrendForce has warned that DRAM price increases of 50% or more are becoming the base case. This isn't confined to exotic AI memory — it's spilling into PC RAM, NAND for SSDs, and practically every memory segment.

Fix, Not Buy

Pua Khein-Seng, CEO of Phison Electronics, recently spoke on the matter for Taiwanese media. The Verge reported his warning for H2 2026, as he predicted that some manufacturers might have to cut product lines, while those unable to secure enough RAM could shut down entirely.

He also believes that consumers might shift toward repairing devices rather than replacing them, simply because new hardware will cost more.

That's the real ripple effect. AI hyperscalers can absorb 80% or even 100% memory price spikes. Smaller PC builders, regional cloud providers, and niche hardware brands cannot. We're already seeing hardware price squeezes everywhere: laptops creeping up in price, SSD costs climbing, hosting plans getting repriced, and gaming hardware facing uncertainty.

The shelves probably won't go empty. This isn't a toilet paper panic. But the economics of computing are changing in real time. As long as the AI giants can afford all the memory they need, everyone else will keep paying for it.

Image: Shutterstock

• No comments yet. Be the first to comment!