Enhancing the AI infrastructure ecosystem! Broadcom releases the next-generation network chip Thor Ultra, simultaneously launching the Wi-Fi 8 solution

According to reports, Broadcom (AVGO.US) is set to launch a new network chip called Thor Ultra. This chip can connect hundreds of thousands of data processing chips, helping enterprises build artificial intelligence (AI) computing systems. The report further states that Thor Ultra enables computing infrastructure operators to deploy a far greater number of chips than ever before, thereby allowing them to build and run large language models that support AI applications. Analysts say this will intensify competition between Broadcom and NVIDIA (NVDA.US). Currently, Broadcom has not responded immediately to requests for comment. The report mentions that the Thor Ultra chip will compete with NVIDIA's network interface chips, aiming to further consolidate Broadcom's dominant position in the internal network communication field of AI application-specific data centers. In the connection between AI systems and other parts of the data center, the Thor Ultra chip plays a key linking role. The core function of such network chips is to assist data center operators in achieving efficient information transmission within the facility. Broadcom Senior Vice President Ram Velaga stated, "In distributed computing systems, networking is crucial for building these large clusters. Therefore, I am not surprised that all companies involved in the GPU business would want to have a place in the networking field." The report states that compared to the previous generation of products, Broadcom engineers have doubled the bandwidth of Thor Ultra. Moreover, from the initial stages of the production process, they have conducted rigorous testing and review of this chip. Velaga revealed that to develop chips like Thor Ultra or to create the Tomahawk series flagship network switches, engineers build a complete system around the chip. The report points out that during this process, engineers discuss key issues such as the chip's packaging form, power consumption requirements, and heat dissipation performance with the hardware systems team. "For every dollar we invest in chip wafer development, ecosystem partners will invest at least $6 to $10," Velaga said. "Therefore, we put a lot of effort into the design phase as if this product is ready for mass production at any time." In another independent announcement, Broadcom released what it claims to be the industry's first silicon-based solution for the broadband wireless edge ecosystem, applicable to scenarios including home gateways, enterprise access points, and smart mobile terminals. Broadcom stated that this Wi-Fi 8 solution is tailored to meet the stringent demands for performance, reliability, intelligence, and energy efficiency in edge networks during the AI era. The company noted that the newly launched licensing options will help expand the scale of edge terminals such as smartphones, personal computers, automobiles, and Internet of Things (IoT) devices, thereby supporting "AI-first" access devices. Mark Gonikberg, Senior Vice President and General Manager of Broadcom's Wireless Communications and Connectivity Division, stated, "Broadcom's Wi-Fi 8 products are meticulously designed to provide the 'AI-ready' capabilities, high performance, low latency, and stability required by modern AI edges." Broadcom's Wi-Fi 8 ecosystem product portfolio specifically includes: the BCM6718 chip for home and operator access scenarios; the BCM43840 and BCM43820 chips for enterprise access scenarios; and the BCM43109 chip for edge wireless terminals such as smartphones, laptops, tablets, and automobiles