Nvidia has introduced a series of innovations and partnerships that signal its determination to remain a central force in the global artificial intelligence (AI) scene. Speaking at Computex 2025 in Taipei, Nvidia CEO Jensen Huang introduced NVLink Fusion, a new technology that signifies the company’s new approach to AI infrastructure.
Unlike previous versions of Nvidia’s proprietary NVLink, which only worked with its own chips, NVLink Fusion enables a more extensive range of processors—including those from external partners and even competitors—to connect seamlessly with Nvidia’s powerful GPUs. This move is widely seen as a response to the rising trend of custom silicon development among major cloud providers and technology firms.
“NV link fusion is so that you can build semi-custom AI infrastructure, not just semi-custom chips,” Huang announced at the event, emphasising the flexibility this new approach offers. He further elaborated that NVLink Fusion allows AI systems to combine Nvidia processors with a variety of CPUs and application-specific integrated circuits (ASICs), stating, “In any case, you have the benefit of using the NV link infrastructure and the NV link ecosystem.” This is expected to appeal to companies seeking to build bespoke AI solutions using a mix of hardware components.
Nvidia has already secured partnerships with several major chipmakers, including MediaTek, Marvell, Alchip, Astera Labs, Synopsys, and Cadence, who will be early adopters of the NVLink Fusion technology. The company also revealed that customers such as Fujitsu and Qualcomm Technologies will now be able to link their own third-party CPUs with Nvidia GPUs in AI data centres, expanding the possibilities for hybrid computing environments.
Industry analysts have pointed out that this strategy could help Nvidia capture a larger share of the data centre market, especially as more companies invest in ASICs and custom processors tailored for specific AI workloads.
However, the decision to open up NVLink does come with risks. Rolf Bulk, an equity research analyst at New Street Research, noted that by allowing customers to use alternative CPUs, Nvidia could see reduced demand for its own central processing units.
Nevertheless, he acknowledged that “at the system level, the added flexibility improves the competitiveness of Nvidia’s GPU-based solutions versus alternative emerging architectures, helping Nvidia to maintain its position at the centre of AI computing.” Notably, some of Nvidia’s most prominent rivals, such as AMD, Intel, and Broadcom, have yet to join the NVLink Fusion ecosystem, which could give Nvidia an early advantage in shaping industry standards.
Beyond NVLink Fusion, Nvidia also provided updates on its next-generation Grace Blackwell systems, designed to handle the increasing demands of generative AI and large language models. The forthcoming GB300 platform, set for release in the third quarter of this year, promises a boost in system performance. This is particularly relevant as AI models continue to grow in size and complexity, requiring ever-greater computational resources.
In addition, Nvidia launched the DGX Cloud Lepton, a new AI platform that serves as a global marketplace for high-performance GPU resources. This platform is intended to connect AI developers with tens of thousands of GPUs from a worldwide network of cloud providers. According to the company, “DGX Cloud Lepton helps address the critical challenge of securing reliable, high-performance GPU resources by unifying access to cloud AI services and GPU capacity across the NVIDIA compute ecosystem.”
Nvidia’s ambitions are not limited to technological innovation. The company announced plans to open a new office in Taiwan and to collaborate with Foxconn, the world’s largest electronics manufacturer, on the development of a cutting-edge AI supercomputer.
“We are delighted to partner with Foxconn and Taiwan to help build Taiwan’s AI infrastructure, and to support TSMC and other leading companies to advance innovation in the age of AI and robotics,” Huang said.