Nvidia’s AI chip portfolio expands again. According to TechCrunch, Nvidia has made a$3.5 billion strategic investment in MediaTek—this deal reveals Jensen Huang’s deeper plan to counter the tech giants’ in-house AI chip wave.
Why MediaTek?
MediaTek is one of the world’s largest mobile chip design companies, with deep expertise inlow-power chip design, edge computing, and IoTAs cloud giants including Google, Amazon, and Microsoft develop their own AI chips (TPU, Trainium, Maia), Nvidia needs a new growth engine. MediaTek precisely fills Nvidia’s gap inon-device AI inference chipsThis investment addresses Nvidia’s shortfall in on-device AI inference chips.
The core logic behind this investment is:AI inference is shifting from the cloud to the edgeDemand for AI inference on smartphones, vehicles, and IoT devices is surging. MediaTek ships over 2 billion chips annually, covering the vast majority of non-Apple smartphones and smart devices worldwide. Through MediaTek, Nvidia effectively gains access toAI inference capabilities across 2 billion devices.
Countering tech giants’ “de-Nvidia-ification”
Over the past two years, Google’s TPU, Amazon’s Trainium, and Microsoft’s Maia chips have accelerated the replacement of Nvidia GPUs. These cloud giants are both Nvidia’s largest customers and its biggest threats. Nvidia’s strategy is clear:If the cloud pie may be carved up, seize the on-device market instead.
The $3.5 billion investment also includes joint development of next-generationAI PC chipsandIn-vehicle AI computing platformsThis complements Nvidia’s Grace-Hopper superchip roadmap—Nvidia’s own chips for data centers, and MediaTek+Nvidia IP combinations for end devices, achieving full coverage from cloud to edge.
What does it mean for the AI industry?
The most immediate impact of this deal isA sharp decline in on-device AI inference costsMediaTek’s mass-production capability means AI inference chips can cost just a few dollars per unit, enabling mid- and low-end smartphones to run large language models locally. Second, Nvidia’s technology IP—including the CUDA ecosystem and TensorRT inference engine—will penetrate more devices via MediaTek’s distribution channels, further solidifying Nvidia’s AI ecosystem moat.
For developers, the next few years may bring“Running Llama-class models locally”as a standard feature on mid-tier smartphones. This is a major boon for AI application developers—user base expansion will be exponential.
Summary: Nvidia’s “Plan B”
Nvidia’s investment in MediaTek is not merely a financial investment butA strategic hedge against shifting dynamics in the AI chip marketAmid intensifying competition from cloud giants’ in-house AI chips, on-device AI inference represents Nvidia’s next growth engine. For everyday users, this means AI will truly enter everyone’s pocket over the next few years—not via cloud APIs, but running directly on smartphones. To learn more about AI industry developments, visit AI Dash ——Discover the best AI tools.