On September 1st, shares of chip design giant MediaTek skyrocketed nearly 10% on the Taiwan stock exchange, hitting the daily price limit. What lit the fuse? A bombshell deal announced jointly by NVIDIA and MediaTek on August 31st (U.S. Eastern Time). NVIDIA is snapping up $3.5 billion worth of overseas convertible corporate bonds (ECBs) issued by MediaTek—the largest direct investment NVIDIA has ever made outside the United States.
This ECB issuance totals $3.9 billion, making it the biggest overseas convertible bond sale in Taiwan’s capital market history. NVIDIA is grabbing $3.5 billion of that—roughly 90% of the entire offering. Alphabet (Google’s parent company) also joined the same issuance, though it didn’t disclose its stake. One key thing to note: since the conversion price, maturity, and potential shareholding ratio haven’t been revealed yet, this should be seen as a convertible bond investment, not NVIDIA taking a direct equity stake in MediaTek. By choosing convertible bonds over plain stock, NVIDIA keeps its options open—it can convert to equity if MediaTek’s share price climbs, while still enjoying downside protection through fixed-income returns.
Jensen Huang, NVIDIA’s founder and CEO, told the media: “NVIDIA’s networking ecosystem is now part of MediaTek’s supply chain, and MediaTek’s XPU is now part of ours. In many ways, both of us have expanded our supply chains through this collaboration.”
According to the joint announcement, the partnership zeroes in on three main areas. In AI infrastructure, MediaTek will officially adopt NVIDIA’s newly launched NVLink Fusion platform, offering hyperscalers, cloud service providers, and cutting-edge AI model developers a pre-validated technical path to build custom XPU processors and seamlessly plug them into NVIDIA’s NVLink-connected, rack-scale AI factories.
What this really means: customers who need custom AI chips can hand the XPU design work to MediaTek and focus their own resources on optimizing the compute core architecture. No need to build the entire engineering ecosystem—interconnects, memory, packaging—from scratch. The NVLink connectivity tech, memory architecture, advanced packaging, and rack-level system integration needed for mass deployment can all lean on the combined NVIDIA-MediaTek tech stack.
In local AI computing, the two companies are doubling down on their existing partnership, continuing to co-develop multiple generations of NVIDIA RTX Spark and DGX Spark PC chips. NVIDIA’s GPU architecture is being deeply fused with MediaTek’s system-on-chip (SoC) design to power consumer PCs, AI developer supercomputers, and enterprise-grade workstations.
In the automotive intelligence space, MediaTek’s Dimensity Auto platform has already integrated NVIDIA technology, bringing advanced AI and RTX graphics to smart cockpits and working in tandem with NVIDIA DRIVE AGX. This collaboration is set to extend across multiple future product generations. NVIDIA previously only had the DRIVE AGX platform in the automotive space, lacking the full vehicle-grade SoC design and mass-production experience. MediaTek’s Dimensity Auto platform is already in use by multiple carmakers, so the two companies complement each other well in areas like automotive certification, functional safety, and cost-effective integration.
For MediaTek, this NVIDIA investment is essentially a golden ticket into the data center custom chip (ASIC) market. For years, MediaTek’s core business has been centered on smartphone SoCs and TV chips, with virtually zero presence in the AI data center arena. Lately, the company has also been wrestling with the growth ceiling of a saturated phone market. Data center ASICs, by contrast, are a fast-growing segment currently dominated by Broadcom and Marvell.
MediaTek is projecting $2 billion in revenue from custom AI chip business in 2026, and the company hopes to capture up to 15% of this market by 2027, translating to a revenue target of $7 billion to $12 billion. NVIDIA’s backing, without question, gives MediaTek a heavyweight stamp of credibility in the ASIC race.
Since the start of this year, NVIDIA’s investment strategy has been shifting toward strategic heavy-asset plays. On the compute infrastructure front, NVIDIA has joined OpenAI’s latest funding round of roughly $110 billion—media reports put its contribution around $30 billion—and has committed up to $10 billion to Anthropic. It’s also been pouring additional capital into AI cloud providers like CoreWeave and Nebius to support data center buildouts, and has inked a strategic partnership with the SK Group worth over $500 billion, planning gigawatt-scale AI factories.
On the upstream supply chain side, NVIDIA has invested up to $3.2 billion in Corning to expand optical fiber manufacturing, and has put billions into optical communications companies like Coherent, Lumentum, and Marvell to secure high-speed interconnect hardware for AI data centers. The $3.5 billion bet on MediaTek is NVIDIA’s key move in the chip design segment—by nurturing custom XPU partners, the company is diversifying its ecosystem’s compute supply to keep pace with the ever-growing and increasingly varied demands of AI workloads.
