Ai’s silicon hunger: chip prices skyrocket, threatening ericsson’s profit margin
The relentless demand for artificial intelligence is triggering a semiconductor crisis, forcing networking giants like Ericsson to scramble for survival and renegotiate contracts. Prices for memory chips – the very lifeblood of AI data centers – have soared to unprecedented levels, fueled by a desperate race for processing power.
A node behind the curve
Ericsson’s head of mobile networks, Per Narvinger, reveals a stark reality: their equipment, critical for 5G infrastructure, utilizes chips lagging behind the cutting-edge performance seen in smartphones and some AI workloads. These chips, predominantly operating at a 5nm node, are significantly behind the 2nm processors powering devices like the Samsung Galaxy S26 series.
Tsmc’s bottleneck, ericsson’s wait
The bottleneck lies squarely with TSMC, the foundry dominating the production of these advanced chips. Demand is so intense that even Apple and Nvidia are facing lengthy lead times, relegated to a waiting list for access to TSMC’s 3nm capacity. Ericsson, reliant on ASICs for radio and baseband operations, is caught in the crossfire, grappling with higher silicon costs and a critical supply chain vulnerability.
Fighting for wafers
“Right now, many of the AI workloads are competing for the same wafers that we also are interested in,” Narvinger stated bluntly. The company is reportedly exploring aggressive renegotiations with customers, attempting to offset the escalating costs of these essential components. This proactive approach underscores the severity of the situation – Ericsson’s profit margins are squarely on the line.
A race to the 2nm
While TSMC works to alleviate lead times and potentially lower costs as AI firms transition to the 2nm process, Ericsson is preparing for the worst. The company has reportedly approached clients, seeking to adjust existing agreements and secure more favorable pricing. The situation mirrors that of Nokia, Ericsson’s primary competitor, who is also navigating similar challenges and engaging in contract discussions.
The cost of innovation
The underlying issue remains simple: AI’s explosive growth is creating a massive surge in demand for semiconductors. As BT CEO Allison Kirkby pointed out, AI data centers consume a staggering 70% of the world's memory chip capacity. This isn’t simply an inconvenience; it’s a fundamental reshaping of the technology landscape, forcing both equipment manufacturers and consumer device producers to adapt – or risk obsolescence. The price inflation isn't just affecting Ericsson; it’s a ripple effect across the entire industry.