If you priced new laptops or a server this year and wondered why the quote went up, AI is a big part of the answer. Not because your office bought AI, but because the data centers that run it are buying up the same kind of memory chips your computers need. It is one of the clearest examples of the AI boom showing up on an ordinary small-business invoice.
What happened
The squeeze was visible by January. The Register reported, citing the market research firm TrendForce, that conventional DRAM prices had already jumped 55 to 60% in a single quarter. Citing Korea Economic Daily, it said Samsung and SK hynix were planning to raise server memory prices by up to 70% that quarter, and it quoted the research firm IDC calling the shortage "unprecedented" and warning it may last into 2027.
The cause is a shift in what memory factories make. AI chips need a special, expensive type of memory called HBM (high-bandwidth memory), and AI servers also use large amounts of ordinary memory and storage. Bloomberg's analysis of the shortage reports that data centers grew to around half of global DRAM consumption in 2025, citing Bloomberg Intelligence, up sharply from five years earlier. Every production line moved to serve AI is one not making memory for laptops and phones.
The latest picture has not eased. TrendForce's September 2026 DRAM contract price report says contract prices held steady for the month but expectations for further increases were revised significantly higher. It describes major Korean and Taiwanese makers moving capacity toward server products, causing acute shortages of PC memory, and PC makers switching to older memory types in budget models to keep prices down.
Why it matters for your business
Memory is in almost everything you buy: laptops, desktops, servers, phones, network gear and storage. When it gets expensive, three things happen:
- Quotes go up or specs go down. You may pay more for the same machine, or get a machine at the old price with less memory than you expected. Read the spec sheet, not just the model name.
- Lead times stretch. Tight supply means some configurations are slow to arrive, especially higher-memory machines.
- Running AI locally costs more. If you were planning to run AI models on your own hardware for privacy reasons, memory is the most expensive part. That changes the math compared with using a cloud service.
The other side of the argument: memory has always run in boom and bust cycles, and high prices eventually draw new factories. Buyers who wait sometimes win. But new memory plants take years to build, and the forecasts above point to tight supply well into 2027.
What we'd do
- Do not wait for a crash on hardware you need now. If machines are failing or blocking work, buy. Waiting has not been rewarded this year.
- Check memory in every quote. For general office work, make sure you are not sold a cut-down configuration that will feel slow in two years.
- Extend the life of what works. Adding memory or replacing a drive in a healthy machine may beat buying new, though upgrade parts are pricier too.
- Rethink local AI hardware plans. Compare the full cost of buying servers for AI against a private cloud option before committing.
- Lock in quotes when you can. Ask suppliers how long a price is held, and get it in writing.
Our take
We think the memory shortage is the most direct way the AI boom touches businesses that never signed up for it. It is not a reason to panic-buy, but it is a reason to stop assuming hardware gets cheaper every year. Plan purchases a quarter ahead, read the specs carefully, and treat memory as the line item that decides whether a quote is a good deal. For now, the AI giants set the price, and the rest of us pay it.
Want a system like this in your business?
A discovery conversation is where we map it: what to automate first, where the human checkpoints go, and what it would take.
Book a discovery conversation →