Le pari de 3,5 milliards de dollars de Nvidia sur MediaTek révèle son plan pour s'attaquer au développement de puces IA de Big Tech

Nvidia has agreed to invest $3.5 billion in Taiwan’s MediaTek, marking a strategic move to bolster its position in the rapidly evolving AI chip landscape. Under the terms of the agreement, MediaTek will incorporate Nvidia’s advanced technologies to develop custom chips tailored for AI firms and large cloud operators. These chips will be compatible with Nvidia’s data centers, allowing seamless integration within existing infrastructures. This partnership reflects Nvidia’s broader strategy to maintain dominance in the data center sector despite increasing competition from entities building their own specialized hardware. The deal follows a trend where prominent AI developers and cloud service providers such as Amazon, Google, Microsoft, OpenAI, and Anthropic are increasingly investing in proprietary silicon to reduce dependency on Nvidia’s GPUs. By collaborating with MediaTek, Nvidia aims to retain influence over the industry while enabling its partners to produce customized solutions that align with its technological framework. This approach allows Nvidia to remain a central player in the AI infrastructure space, even as others seek alternatives. Dion Harris, Nvidia’s senior director of high-performance computing and AI hyperscaler infrastructure solutions, emphasized during a recent media briefing that Nvidia operates as an AI infrastructure company rather than merely a semiconductor manufacturer. He noted that the company has significantly expanded its offerings beyond traditional computing chips, positioning itself as a key enabler for the future of artificial intelligence. This partnership exemplifies Nvidia’s ongoing strategy of funding and supporting companies whose products can integrate with its ecosystem. The deal grants MediaTek access to Nvidia’s NVLink Fusion ecosystem, which facilitates rapid communication between diverse chips, whether they are manufactured by Nvidia or other vendors. This capability is crucial for creating efficient, scalable AI environments. In addition to the investment, Nvidia has recently partnered with Amazon Web Services, although without the financial commitment. AWS plans to add 2 million Nvidia GPUs to its infrastructure and will implement NVLink Fusion. This collaboration highlights Nvidia’s efforts to expand its reach across different sectors, ensuring its technology remains integral to both enterprise and consumer applications. Through this alliance, MediaTek gains the ability to assist its clients in constructing custom chips optimized for specific tasks, utilizing Nvidia’s proven technology stack and rack-scale architecture. This enables organizations to standardize their AI infrastructure, streamlining operations and enhancing performance. Harris described the impact of this integration, stating that nearly all major cloud services and AI development teams utilize Nvidia’s platform in some capacity. MediaTek, known for its expertise in designing chips for smartphones, smart homes, vehicles, and telecommunications, will also continue working with Nvidia on projects like DGX Spark, a compact AI workstation aimed at developers. Additionally, the two companies will explore opportunities in AI-driven vehicles, leveraging Nvidia’s RTX graphics and Drive AGX platform for intelligent vehicle systems and autonomous driving capabilities. Jensen Huang, Nvidia’s founder and CEO, highlighted the transformative role of AI across all computing domains, from massive data centers to personal computers and vehicles. He stated that the collaboration with MediaTek is instrumental in expanding Nvidia’s presence into new markets and empowering users to create distinct AI systems on a large scale. This partnership underscores the dynamic evolution of the AI industry, driven by continuous innovation and strategic alliances.