There is a category of risk that most product reviews do not mention, most warranties do not cover, and most buyers do not think about until it is too late. Call it deprecation risk: the chance that the hardware in your hands loses functionality through no physical fault of its own, on a timeline set entirely by someone else’s business decisions.
A Gransfors axe bought in 2005 works identically in 2025. It will work in 2045. Its only deprecation curve is physics: the edge dulls, the handle dries, and both are serviceable. The tool’s usefulness is not contingent on the continued operation of any external system.
Compare that to the Revolv smart home hub. Revolv was a well-reviewed product, and the people who bought it paid a one-off price for what appeared to be a piece of hardware. In 2014, Nest acquired the company. In 2016, Nest (by then owned by Google) shut down the servers the hub depended on to function. The date came, the servers went dark, and every Revolv unit in the world became inoperable. Not damaged. Not defective. Switched off remotely, at the discretion of a company the original buyer had never done business with.
The model at work here is not complicated. When you buy a connected device, you are not buying a product. You are buying access to a service hosted on someone else’s infrastructure, packaged as a product at a one-off price. The one-off price is what makes the transaction feel like ownership. The service dependency is what makes it something else.
This is invisible at the point of purchase. No connected device ships with a deprecation timeline on the box. No retailer discloses how long the manufacturer intends to maintain the servers. A product sold as “smart” may carry an implicit service contract of eighteen months or twelve years; neither figure appears anywhere in the purchase flow. You pay the product price and receive the service terms. The gap between those two things is the risk.
Sonos illustrated this in 2020, when it announced that certain older speakers would no longer receive software updates, and that running a mixed system of old and new hardware would hold back the newer units. The hardware was not broken. Sonos had simply decided that continuing to support it was not worth the engineering cost. Buyers who had paid full retail price years earlier found their speakers functionally downgraded by a firmware decision made in a boardroom they had no access to.
Printer manufacturers have turned the same dynamic into a revenue model. Epson and HP have both shipped firmware updates that disabled third-party ink on printers that users already owned and had paid for in full. The printer works. The authentication routine, added remotely via update, decides it does not. The constraint is not mechanical. It is a software policy, deployed onto hardware you are supposed to own, in service of a business objective you did not agree to.
The distinction that matters here is between tools that deprecate on physics timescales and tools that deprecate on business timescales. A hand tool, a mechanical watch, a cast iron pan: these degrade according to their materials and use. The timeline is measured in decades. The causes are friction, oxidation, and fatigue. They are understandable, gradual, and in most cases addressable. You can sharpen the knife. You cannot negotiate with a decommissioned API.
A connected device degrades according to the business decisions of the company that controls its backend. The timeline is unpredictable. A profitable product line may run servers for fifteen years. An acquired product may be deprecated in eighteen months. A solvent company may simply decide the maintenance cost no longer justifies the spend. None of these events are visible to you as a buyer, and none of them are covered by any warranty you have ever read.
The buying heuristic that follows is simple. If the tool requires a server call to function, you do not own it. You are licensing access to it, and the licence terms are set unilaterally, revised without your consent, and can be revoked at any time. The one-off price you paid is not purchasing the thing. It is purchasing admission.
This is not an argument against connectivity as such. There are connected tools with genuine long service records, and some of the risk is manageable through buying decisions: established companies with a track record, open firmware ecosystems, devices with local-only operation modes. But the risk class is real, it is growing as more physical goods acquire software dependencies, and it is almost entirely absent from the purchase conversation.
The tools that will hold their value over the next decade are the ones whose usefulness is not contingent on the continued co-operation of a server you do not control. That is not a sentimental claim about craftsmanship. It is a statement about where the deprecation risk sits, and who carries it.
For more on how service contracts have displaced product ownership, see The Subscription Trap. The right-to-repair debate approaches the same territory from the regulatory angle: Right to Repair: Not an Environment Argument.