@Lexus @Dibadoo Good points from both. Nokia may be moving to a kind of hybrid model where GPU and ASIC development coexist. AI-RAN would then serve not only as a potential new revenue stream but also as a development platform, while ReefShark remains an energy-efficient basic solution. The sustainability of this model ultimately depends on two things: either AI-RAN scales commercially, or the GPU-based approach genuinely alleviates cost pressure on ASIC development. Otherwise, maintaining two parallel paths could be burdensome in the long run.
One possible underlying factor here could also be the cost structure. It may be that scaling Nokia’s own high-level ASIC development for the AI era would be extremely capital-intensive for Nokia, and with already weak profitability and a smaller market share than Ericsson, the equation would have been challenging. The Nvidia collaboration is both a risk and an opportunity, but it can also be interpreted as a way to avoid the heaviest R&D arms race. At the same time, the focus may shift to basic SoC (System on Chip) solutions and, in demanding situations, to applying Nokia’s own expertise to Nvidia’s technology and the CUDA software platform.