Nvidia Invests $3.5 Billion in MediaTek and Launches NVLink Fusion

Key Takeaways

1Nvidia invests $3.5 billion in MediaTek and opens access to NVLink Fusion

2AWS plans to deploy 2 million Nvidia GPUs and integrate NVLink Fusion

3MediaTek aims for $2 billion in ASIC revenue by 2026 and continues its PC and automotive collaborations

Why it matters : The opening of NVLink Fusion to MediaTek combines custom chips with Nvidia infrastructure, while cloud deployments and PC/auto collaborations expand use cases.

Nvidia invests $3.5 billion in MediaTek and grants it access to NVLink Fusion. This agreement comes as clouds and AI labs design their own chips, while Nvidia multiplies partnerships, from AWS to PCs and automotive.

Automotive and PC: The MediaTek–Nvidia Collaboration Expands

MediaTek's automotive platforms rely on Nvidia's RTX graphics to equip smart vehicle cockpits and utilize the Nvidia Drive AGX platform for autonomous driving workloads. The two companies plan to continue developing platforms for software-defined, AI-powered vehicles. Their collaboration has also broadened with RTX Spark, an initiative aimed at integrating Nvidia's AI technology into consumer AI PCs, and they will continue to work together on DGX Spark, an AI desktop computer for developers. These projects build on MediaTek's experience in designing custom chips for smartphones, smart homes, automotive, and wireless communications.

Infrastructure Opening: MediaTek to Have NVLink Fusion

The agreement grants MediaTek access to Nvidia's NVLink Fusion ecosystem. NVLink allows different chips, including those not designed by Nvidia, to communicate quickly with each other. MediaTek plans to use Nvidia's technology to design custom chips for AI companies and hyperscalers, with direct integration into data centers based on Nvidia's infrastructure. These developments will leverage Nvidia's technology stack and rack-scale architecture. According to Dion Harris, enabling MediaTek to offer this extension should help its clients standardize on a common infrastructure and deploy their custom chips alongside other systems on the same platform, while also opening the ecosystem to MediaTek's entire customer base.

Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Hyperscalers and Custom Chips: The Competitive Context

An increasing number of AI companies and major cloud providers, including Amazon, Google, Microsoft, OpenAI, and Anthropic, are investing in designing their own chips to reduce their dependence on Nvidia's GPUs. Dion Harris notes that every cloud and model builder uses Nvidia's platform in one way or another. He emphasizes that Nvidia positions itself as an AI infrastructure company and has expanded its focus beyond pure compute chips in recent years.

MediaTek's Trajectory and Cloud Relay with AWS

Nvidia is dedicating $3.5 billion to MediaTek, which will enable it to continue developing custom chips designed by cloud companies or AI labs and optimized for their own workloads. MediaTek has not disclosed information regarding the customers for these chips but is gradually expanding its ASIC business aimed at data centers. In June, the company indicated it expects $2 billion in revenue by 2026 for this segment and aims for a larger market share in the coming years. Meanwhile, Nvidia announced a partnership with Amazon Web Services, without direct investment, under which AWS plans to deploy an additional 2 million Nvidia GPUs and integrate NVLink Fusion into its infrastructure. Jensen Huang, founder and CEO of Nvidia, believes that AI is transforming every computing platform, from AI factories to PCs and cars, and asserts that Nvidia and MediaTek are building platforms to bring Nvidia's accelerated computing to new markets and enable customers to create differentiated AI systems at scale.