A power restructuring, shaping the fate of the communications industry for the next 30 years, is quietly unfolding. News that chip giant NVIDIA (NVDA-US) plans a major push to replace traditional application-specific integrated circuits (ASICs) in 5G base stations with general-purpose GPUs has triggered significant volatility in capital markets.
Shares of the two major telecom equipment manufacturers on the Stockholm Stock Exchange plummeted, with Ericsson falling over 6% in a single day and Nokia dropping nearly 7%. This "software-defined networking" offensive, launched by the AI ??chip giant, is forcing traditional telecom equipment manufacturers to rethink their survival strategies.
The root of this panic lies in NVIDIA's precise deconstruction of the underlying logic of the Radio Access Network (RAN). Traditional base stations have long relied on highly customized but closed application-specific integrated circuits (ASICs), while NVIDIA argues that the essence of the communication physical layer is large-scale linear algebraic computation, which is highly consistent with the computational paradigm of AI deep learning.
Through the CUDA cores and Tensor Cores within its GPUs, NVIDIA not only efficiently handles these parallel computing tasks but also, through the evolution of its product matrix, successfully extends its technology from high-power cloud computing to low-power distributed sites.
NVIDIA's strategic layout is highly aggressive; its "AI-RAN Alliance" quickly attracted over a hundred members, including SoftBank and Microsoft.
This strategy of lowering the entry barrier with an open ecosystem and then establishing ecosystem lock-in through CUDA software is strikingly similar to the rise of the Android system in the smartphone era.
Its ultimate goal is to transform the original "hardware-agnostic" open network vision into a new market rule where "only NVIDIA hardware is acceptable."
Faced with NVIDIA's aggressive offensive, a clear division has emerged within traditional telecom giants. Ericsson chose to uphold its long-term investment in self-developed ASICs and allied with chip giant Intel to strongly support its CPU-centric "Cloud RAN" solution.
Ericsson believes that in the midst of a downturn in RAN spending by telecom operators, ASICs possess irreplaceable cost and temperature control advantages, and can prevent telecom operators from being locked into a single chip supplier.
Ericsson advocates that by adding neural network accelerators to hardware, real-time improvements in network performance can be achieved without blindly investing in expensive GPUs.
In contrast to Ericsson's defensive approach, another giant, Nokia, chose to "surrender," fully embracing Nvidia's GPU wave.
After receiving strategic investment from Nvidia, Nokia and Nvidia jointly launched the AIAerial platform and successfully completed Southeast Asia's first AI-driven 5G call trial, jointly demonstrating a forward-looking solution that uses GPUs to replace traditional CPUs as baseband processors.
This "GPU colonization" panic triggered by Nvidia essentially reflects the capital market's anxiety about the redistribution of the communications industry's value chain. Once GPUs become the default computing power foundation for wireless networks, traditional equipment manufacturers face the risk of being marginalized from "system integrators" to "hardware assemblers" like in the PC or smartphone era.
In this unavoidable paradigm shift, the question of whether Ericsson's ASIC moat or Nokia's new wave of GPUs will prevail in the future 6G era will determine the new landscape of the communications industry.
Tags：