Key Takeaways
- Between tariffs and supply chain issues, replacing servers every three to five years is looking a lot less realistic.
- Some companies — Big Tech, in particular — have already been practicing how long they can stretch their servers’ useful lives.
- One edge expert says companies may be able to buy less hardware by making better use of the servers they already have.
Has anyone priced a server lately? The landscape is…depressing. Memory prices have doubled — in some cases quadrupled — over the past six months.
Server orders from Dell, HPE, and Cisco now have wait times stretching from 10 to 24 weeks. Lenovo let outstanding hardware quotes expire in January and just decided to reprice from scratch.
Gartner warned a while ago that tariff shifts are making sourcing, pricing, and financial forecasting harder for companies. For server buyers, that might mean changing an old habit: replacing hardware every three to five years because, well, that is what companies have always done. But Susan Odle, CEO of enterprise edge provider StorMagic, says that’s looking a little outdated.
Server Orders Are Taking Longer to Fill
Dell, HPE, and Cisco server wait times are reportedly stretching from 10 to 24 weeks
“The global economic environment has become fundamentally unpredictable,” Odle said. “Tariffs are a major factor, but they are far from the only driver — policy shifts, supply chain shocks, regional instability, and the disruptive pace of AI adoption all compound the uncertainty.”
This of course doesn’t mean companies can avoid buying hardware forever, but it is enough of a nudge to go from “We’re replacing this in three years,” to “Does this actually have to be replaced yet?”
“Hardware is living longer than ever,” Odle said. “The default 3-5 year refresh cycle has been the norm for a long time, but it’s time to challenge that assumption.”
Hyperscalers Are Already Changing Server Lifespans
A lot of enterprises are still sticking to shorter refresh timelines, with a 2025 IDC white paper found that 44% of enterprises refresh server infrastructure every three years or less. But take a look at Big Tech and they’ve been planning for longer server life for the past several years:
- Amazon: Servers went from a three-year useful life to four years in 2020, then to five years in 2022
- Microsoft: Server and network equipment went from four years to six years, effective fiscal 2023
- Alphabet: Servers went from four years to six years, while some network equipment went from five years to six years, effective 2023
- Meta: Most servers and network assets were extended to 5.5 years, effective Jan. 1, 2025
Most enterprises are not Amazon or Microsoft, obviously. But that is kind of the point — and frankly, the way the cycle usually goes.
Remember when free SSL was not the norm? The big names changed that, and then the smaller providers followed suit as soon as customers began expecting it to come standard with most hosting plans. S3 compatibility followed a similar path in storage, with smaller providers building around the API expectations AWS helped set.
Once the largest platforms normalize a better, cheaper, or easier way to operate, the rest of the market eventually has to answer for why it is still doing things the old way. And that’s exactly why they follow suit.
The Answer Is Not “Keep Old Servers Alive Forever”
Odle is not arguing that companies should run old hardware until it wheezes itself to death. Instead, use software to make smaller infrastructure setups more resilient, so companies are not forced to buy more hardware every time they need more reliability.
As the CEO of StorMagic, an enterprise edge provider, Odle is a big fan of virtualization. Virtualization platforms that can deliver true high availability on just two nodes are, in her words, “low-hanging fruit, not just to cut costs, but to also blunt the impact of future cost increases.”
But saving money on servers is only helpful if it doesn’t also create a downtime problem, which is entirely possible. That’s where Odle says virtualization comes in: If companies can keep workloads running reliably with fewer machines, then why would they want to go back to replacing on the old schedule?
“Hardware is living longer than ever. The default 3-5 year refresh cycle has been the norm for a long time, but it’s time to challenge that assumption.”
At the edge, companies are often running infrastructure in smaller, spread-out locations — stores, branches, clinics, warehouses, factories, etc. So if each location needs fewer servers to stay reliable, the savings multiply fast. It means fewer machines to buy, ship, install, power, cool, fix, and eventually replace.
As Odle put it, the goal is “fewer boxes, smarter software, and designs that increase the return on investment in modernization efforts.” So while the three-to-five-year refresh cycle is not dying (yet), it’s definitely one of those things that some providers and managers are more than happy to bend.
