Embedded World 2025 officially commenced this week in Nuremberg, Germany, with Sandra Rivera, CEO of FPGA company Altera, delivering the keynote address. In her presentation, Rivera discussed key developments in edge AI and highlighted the role that FPGAs can play in the rapidly evolving sector.
Rivera explained that FPGAs are particularly advantageous for AI applications due to their stable computational efficiency across various AI processing modes. While they may fall behind ASICs in certain specialized functions, FPGAs outperform traditional CPUs and GPUs in overall computing performance. Their flexibility and versatility provide a significant competitive edge in the AI market, making them a strong alternative to ASICs, which are designed for specific functions.
One of the most notable trends Rivera pointed out is the shift toward edge data centers, which will process approximately 75% of corporate data in the future, compared to traditional cloud data centers. This transition is driven by the need for faster computation, reduced latency, and heightened data security, particularly for real-time AI applications.
Rivera emphasized that edge AI systems prioritize these factors, with the added challenge of considering size, weight, and power consumption in constrained embedded systems. Reducing costs while meeting these strict requirements is essential for scaling AI deployment in edge environments.
The reprogrammable nature of FPGAs provides a distinct advantage in the AIoT market, where applications are diverse and constantly evolving. Unlike ASICs, which are designed for one specific task and often require new hardware for different applications, FPGAs can be continually reconfigured to meet changing needs. This flexibility allows for longer product lifecycles and the integration of various subsystems, supporting larger-scale computations while adapting to customer demands.
However, Rivera noted that while hardware advancements like FPGAs are crucial, software support is also necessary to fully unlock their potential in AI applications. To this end, Altera has developed an integrated AI platform that helps customers overcome technical barriers during the product design process, thereby accelerating development timelines. Rivera emphasized that whether clients bring their own data and models or require support in AI model-related technologies, Altera is now fully equipped to meet these demands.
Rivera also highlighted that Altera's FPGA platform offers comprehensive, ready-to-deploy solutions in areas ranging from smart factories to smart transportation systems. The company's success in expanding its AI inference capabilities from data centers to embedded systems illustrates its technological strength and positions it as a key player in the AI era. She added that FPGA technology has not faded over time but has instead been revitalized, thanks to the growth of AIoT applications.
Article edited by Jack Wu
