NVIDIA Puts $3.5 Billion Into MediaTek to Wire Custom AI Chips Into Its Empire

NVIDIA has invested $3.5 billion in convertible bonds issued by MediaTek, the companies announced Monday, deepening a partnership that now spans AI data center infrastructure, local AI computing, and automotive platforms. The deal, which lands just days after NVIDIA’s most recent earnings cycle, is the chipmaker’s latest and most direct move to extend its AI empire beyond its own GPUs — into the custom silicon that everyone else is building.

At the center of the announcement is MediaTek’s adoption of the NVIDIA NVLink Fusion platform, first introduced at Computex 2025. NVLink Fusion gives hyperscalers, cloud service providers, and frontier model developers a prevalidated path to develop custom XPUs — their own AI accelerators — and plug them into NVIDIA’s NVLink-connected, rack-scale “AI factories” rather than building standalone clusters.

What NVLink Fusion Actually Does

Building a custom AI accelerator is no longer the hard part. Getting it into a production rack-scale system is. A deployable XPU requires multi-die integration, advanced packaging, high-speed SerDes, HBM memory, I/O, and scale-up networking — each element demanding extensive chip-to-rack engineering, qualification, and supply-chain support before it can run in a live data center.

NVLink Fusion packages the critical technologies surrounding a custom XPU so customers don’t have to engineer every element from scratch:

• NVLink Fusion chiplet — connects XPUs to the NVIDIA NVLink scale-up fabric using NVIDIA photonics or electrical interconnects
• NVLink-C2C — high-bandwidth, energy-efficient connectivity between XPUs, NVIDIA Rosa CPUs, and other compatible processors
• NVHBM — customized memory capabilities that increase bandwidth and energy efficiency while dedicating more silicon area to compute

Customers bring their XPU designs to MediaTek and tailor connectivity, memory, packaging, performance, and power characteristics to their workloads. MediaTek handles the integration; NVIDIA supplies the fabric, the rack architecture (MGX), and the ecosystem. The pitch is simple: focus your resources on differentiated compute, and rely on NVIDIA and MediaTek for everything required to actually deploy it.

Three Pillars of the Expanded Partnership

The collaboration is structured around three major areas:

AI infrastructure. MediaTek joins the NVLink Fusion ecosystem, enabling customers to develop custom AI infrastructure designed to integrate with NVIDIA rack-scale systems. The platforms are designed to evolve alongside future NVIDIA architectures, so a customer’s custom accelerator doesn’t get orphaned when NVIDIA moves to its next generation.

Local AI computing. The companies will continue collaborating on multiple generations of NVIDIA RTX Spark and DGX Spark PC chips — products that combine NVIDIA GPUs with MediaTek SoCs for consumer PCs, AI developer supercomputers, and enterprise workstations. This builds on the GB10 Grace Blackwell Superchip that powers DGX Spark, which pairs a Blackwell GPU and Grace CPU connected by NVLink-C2C.

Automotive. MediaTek Dimensity Auto platforms will continue integrating NVIDIA technologies for AI-powered cockpits and software-defined vehicles, working alongside NVIDIA DRIVE AGX — what the companies call the “physical AI” era of increasingly intelligent cars.

Why This Deal Matters

For MediaTek, the investment and the platform deal cement its transformation from a mobile-chip house into a serious custom AI silicon player. The company has been raising its targets aggressively: it doubled its 2026 AI ASIC revenue target to roughly $2 billion, and in late July raised its ambition again to capture 15–20% of an estimated $80 billion custom AI chip market by 2027. NVIDIA’s capital and ecosystem access supercharge that trajectory.

For NVIDIA, the strategic logic is subtler — and arguably more consequential. Every hyperscaler is building its own AI chips to reduce dependence on NVIDIA. NVLink Fusion turns that threat into a subscription: even when customers design their own XPUs, they still need NVIDIA’s interconnect fabric, memory architecture, rack systems, and networking to make those chips competitive at scale. NVIDIA captures value from custom silicon it didn’t manufacture.

The $3.5 billion convertible bond adds a financial dimension to this pattern. NVIDIA has been financing companies across the AI ecosystem — from the Intel $5 billion equity investment to the sprawling AI Compute Partnership program — in deals critics describe as “circular financing”: NVIDIA funds partners who then buy or extend its stack. A convertible bond gives NVIDIA upside exposure to MediaTek’s equity appreciation while initially holding debt, aligning both companies’ incentives over a multi-year horizon. MediaTek explicitly cautioned in its forward-looking statements that the collaboration’s commercial outcome remains subject to competitive and execution risks — but the direction is clear.

Quotes from the Announcement

“AI is transforming every computing platform — from the world’s largest AI factories to the PC and the car,” said Jensen Huang, founder and CEO of NVIDIA. “MediaTek is one of the world’s great semiconductor companies, with exceptional expertise in system-on-chip design, connectivity, leading performance and power efficiency. 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.”

“MediaTek and NVIDIA share a vision for making advanced AI computing pervasive across the technology landscape,” said Rick Tsai, vice chairman and CEO of MediaTek. “NVIDIA’s investment strengthens a collaboration that spans cloud AI infrastructure, local AI computing and automotive in the era of physical AI.”

The Bigger Picture

The custom ASIC market is where the next phase of the AI hardware race is being fought. Broadcom — the current leader in custom AI silicon — sees its AI chip revenue reaching “significantly in excess of $100 billion” in 2027, and Marvell, Alchip, and Astera Labs are all fighting for position. MediaTek’s combination of NVIDIA’s ecosystem and its own power-efficient SoC design prowess makes it a credible challenger to Broadcom’s dominance.

Meanwhile, the deal extends a busy stretch for NVIDIA’s ecosystem strategy. Just this month it has navigated an internal pause on parts of its AI Compute Partnership program, faced scrutiny over “circular financing” concerns, and now deployed $3.5 billion to lock in one of Asia’s most important chip designers. The message to the market: whether customers buy NVIDIA GPUs or build their own XPUs, NVIDIA intends to be paid either way.

For developers and enterprises, the near-term practical effect is limited — NVLink Fusion products are still working their way through design cycles. But the strategic signal matters. The era of monolithic, single-vendor AI infrastructure is giving way to semi-custom AI factories, and NVIDIA has just moved to own the connective tissue of that new world.