Taiwan's MediaTek priced $3.9 billion in overseas convertible bonds on August 31, with NVIDIA purchasing $3.5 billion — roughly 90% of the offering — in a deal both companies have tied directly to a sweeping expansion of their AI technology partnership.
The issuance sets a record for the largest overseas convertible bond offering in Taiwan's capital market history. Alphabet joined NVIDIA as an investor in the remaining portion, though MediaTek has not disclosed Alphabet's allotment or the bond's full terms — including interest rate, maturity date, conversion price, or intended use of proceeds.
NVIDIA Anchors Its Data Center Strategy Through MediaTek
The financing comes packaged with a broad technology agreement. MediaTek will adopt NVIDIA's NVLink Fusion platform to offer hyperscalers, cloud service providers, and frontier AI model developers a pre-validated path to build custom XPUs — processors engineered for specific AI workloads — that integrate directly into NVIDIA-powered, rack-scale AI factories.
The structure reflects NVIDIA's strategic response to a growing wave of custom silicon development by Amazon, Google, Microsoft, OpenAI, and Anthropic, all of which have moved to reduce dependence on NVIDIA's general-purpose GPUs. By coupling its NVLink Fusion ecosystem with a direct investment in MediaTek, NVIDIA keeps custom chip designers anchored to its interconnect standards and rack-scale infrastructure even as customers pursue differentiated silicon.
"NVIDIA is an AI infrastructure company," said Dion Harris, NVIDIA's senior director of HPC and AI hyperscaler infrastructure solutions, in remarks reported by TechCrunch. "We expanded beyond pure computing chips years ago. By MediaTek being able to offer this extension to its customers, it allows them to standardize on the rack-scale infrastructure across their AI factories."
Last week, NVIDIA announced a similar arrangement with Amazon Web Services — granting access to NVLink Fusion without a direct investment — underscoring how the platform has become a recurring element in the chipmaker's ecosystem strategy.
NVLink Fusion Ties Custom XPUs To NVIDIA's Rack Architecture
At the center of the partnership is NVLink Fusion, a platform designed to shorten the path from custom chip development to production-ready AI factory deployment. Rather than validating every surrounding system element from scratch — interconnects, memory subsystems, packaging, and cabinet infrastructure — customers using NVLink Fusion can focus engineering resources on their core XPU design while relying on NVIDIA and MediaTek for the surrounding technology stack.
The platform includes three key components: NVLink Fusion Chiplets that connect custom XPUs to NVIDIA's NVLink scale-up fabric via optical or electrical interconnects; NVLink-C2C for high-bandwidth chip-to-chip connectivity; and NVHBM, a customized high-bandwidth memory technology designed to improve energy efficiency and free up additional silicon area for computation.
MediaTek brings custom ASIC design, system-on-chip engineering, advanced packaging, and connectivity expertise to the arrangement. Customers can tailor network connectivity, memory architecture, packaging, performance, and power characteristics to specific workload and infrastructure requirements — then deploy on NVIDIA's validated MGX rack-scale architecture.
"MediaTek and NVIDIA share a vision for making advanced AI computing pervasive across the technology landscape," said MediaTek Vice Chairman and CEO Rick Tsai in a joint press release. "By combining NVIDIA's leadership in accelerated computing and AI software ecosystem with MediaTek's expertise in a diverse AI technology portfolio from edge to cloud, and our leadership position in custom silicon, we can accelerate innovation for our customers."
MediaTek's custom data center ASIC business is targeting $2 billion in revenue for 2026, with the company aiming to capture between 15% and 20% of the custom AI accelerator market by 2027.
Three Fronts: Cloud, Edge, And The AI-Defined Car
Beyond cloud infrastructure, the partnership spans local AI computing and automotive. On the edge computing side, MediaTek and NVIDIA will collaborate on multiple generations of RTX Spark and DGX Spark chips — the processors powering NVIDIA's consumer AI PCs, developer-grade personal supercomputers, and enterprise workstations. The companies previously co-developed the GB10 Grace Blackwell Superchip, which combines an NVIDIA Blackwell GPU with a Grace CPU connected by NVLink-C2C, and serves as the core of NVIDIA's DGX Spark desktop system.
On the automotive side, MediaTek's Dimensity Auto platform will continue to integrate NVIDIA RTX graphics for intelligent vehicle cockpits and operate alongside NVIDIA Drive AGX, the company's in-vehicle computing platform for autonomous driving workloads. The two companies plan to extend this collaboration across multiple future product generations as the industry moves toward AI-powered, software-defined vehicles.
"AI is transforming every computing platform — from the world's largest AI factories to the PC and the car," said NVIDIA founder and CEO Jensen Huang in the joint statement. "Together, we're building platforms that bring NVIDIA accelerated computing to new markets and give customers the freedom to create differentiated AI systems at enormous scale."
With the $3.9 billion convertible bond setting a Taiwan capital market record and NVIDIA committing nearly all of its value, the deal marks a pivot in the MediaTek-NVIDIA relationship — from component-level collaboration to a shared stake in how AI infrastructure gets built, deployed, and scaled at rack level. How quickly MediaTek can translate those design wins into mass-produced XPUs will determine whether the investment pays off for both sides. （Related： AI Sends Chip Equipment Market Toward $220B as SEMI Raises Forecast ｜ Latest ）
