Prices for memory used in routers and set-top boxes are surging nearly sevenfold thanks to AI, raising fresh fears that the industry’s silicon binge could leave telcos scrambling to get customers online.
Fresh analysis from Counterpoint Research suggests the cost of memory used in consumer hardware has shot up more than 600 percent over the past year as chipmakers quietly steer production towards higher-margin AI infrastructure. The knock-on effect is being felt most sharply in broadband kit, where DRAM and NAND prices are climbing far faster than in smartphones or PCs.
“Over the last nine months, smartphone memory prices jumped 3x, but the prices for ‘consumer memory’-based broadband products jumped almost 7x,” Counterpoint said. “Routers are hit the hardest, especially for OEMs with an unsecured supply and weaker negotiating power.”
That leaves some of the telecoms industry’s least glamorous hardware suddenly front and center in supply chain planning. Routers and TV boxes, once the sort of commodity gear telcos handed out like branded pens, are getting noticeably more expensive to build. In some low to mid-range models, memory alone now accounts for more than a fifth of total manufacturing costs, up from about 3 percent just a year ago.
None of this is happening in isolation. Memory shortages are already rippling through the PC market, with vendors warning of price increases and softer shipment volumes as suppliers prioritize chips for AI servers over everyday hardware.
Some of this squeeze is self-inflicted, according to Counterpoint. Telcos have spent the past couple of years adding AI-style smarts to home gateways, which means newer devices need noticeably more memory. Unfortunately, that pushes them into the same component market as hyperscale datacenters – and they don’t tend to win those bidding wars.
Counterpoint reckons the consequences could stretch beyond hardware margins. Broadband rollout plans could face delays or higher procurement costs, with tight memory supply expected to persist through at least mid-2026, even if prices stop climbing as aggressively.
For customers, the fallout may creep in gradually in the form of pricier routers or slower upgrades to newer connectivity tech. For the industry, though, the broader warning is already obvious. AI may be sold as the future of computing, but right now it is also quietly reshaping semiconductor economics – and making the unglamorous business of getting people online more expensive than anyone expected. ®