Nvidia Puts $3.5 Billion Into MediaTek Bonds, Ties It to NVLink

What happened

Nvidia has agreed to invest $3.5 billion in convertible bonds issued by MediaTek, part of a broader partnership the two companies say spans AI infrastructure, local AI computing devices such as RTX Spark and DGX Spark-class PCs, and automotive software-defined vehicle platforms. Reuters independently confirmed the $3.5 billion figure and reported that it makes Nvidia the largest participant in MediaTek’s $3.9 billion overseas convertible-bond offering. The Taiwan-focused outlet BigGo, citing a joint company statement and an Nvidia press conference on August 31, called it Nvidia’s first direct investment in a Taiwanese IC design company and its largest single subscription in the bond offering.

As part of the arrangement, MediaTek will adopt Nvidia’s NVLink Fusion platform, giving hyperscalers, cloud providers and frontier model developers a pre-validated path to design custom XPUs that plug into Nvidia’s rack-scale, NVLink-connected AI systems. Reuters reported the bond investment follows a $5 billion financing plan MediaTek’s board approved in July 2026 to fund its expansion into AI data-center chips beyond its traditional smartphone-chip business, and that Alphabet also participated in the same bond offering without disclosing the size of its stake.

Jensen Huang, Nvidia’s founder and chief executive, framed the deal in sweeping terms: “AI is transforming every computing platform — from the world’s largest AI factories to the PC and the car.” MediaTek vice chairman and CEO Rick Tsai echoed that framing: “MediaTek and NVIDIA share a vision for making advanced AI computing pervasive across the technology landscape.”

MediaTek, according to its own guidance reported by TechCrunch, expects its custom data-center ASIC business to generate $2 billion in 2026 revenue — a figure that comes from the company itself, with no independent measurement found in the sources reviewed.

What this means (and what it does not)

TechCrunch placed the deal against a backdrop of Amazon, Google, Microsoft, OpenAI and Anthropic all building proprietary AI chips to reduce their reliance on Nvidia GPUs, framing the MediaTek investment as Nvidia’s response to that trend. Dion Harris, Nvidia’s senior director of HPC and AI hyperscaler infrastructure solutions, described the goal to TechCrunch this way: “This is really about opening up this ecosystem to the entire MediaTek customer base.” Nvidia frames the arrangement as opening an ecosystem, but the mechanics point the other way too: embedding NVLink Fusion into MediaTek’s custom-chip design flow keeps Nvidia’s interconnect central even as hyperscalers build their own silicon. MediaTek, for its part, gets a $3.5 billion cash infusion and a credibility boost as it enters the data-center ASIC market — which is why its executives echo Nvidia’s language of shared vision.

The deal does not establish that Nvidia is losing ground to in-house chip efforts. TechCrunch also reported that AWS separately announced plans to deploy 2 million additional Nvidia GPUs while integrating NVLink Fusion, alongside its own in-house chip programs — the two strategies are running in parallel, not one replacing the other. Nor does the investment prove circularity in the sense critics allege. Joe Tigay, a portfolio manager at Rational Equity Armor Fund, told Reuters: “Nvidia is financing MediaTek so MediaTek can develop products that extend Nvidia’s architecture…it is still Nvidia using its balance sheet to accelerate ecosystem growth.” Huang directly disputed that characterization on Bloomberg TV: “This is not circular because obviously they do their own business and we do our own business.” Alphabet’s own position complicates any simple reading further — it invested in the same MediaTek bond offering while Google continues developing its own in-house TPU chips, a tension no source reviewed addresses or reconciles.

What we still do not know

No source discloses whether MediaTek’s adoption of NVLink Fusion is exclusive, or whether MediaTek remains free to design chips on non-Nvidia interconnect stacks for other customers. Neither the Nvidia press release nor Reuters or TechCrunch’s reporting gives a production timeline or shipping dates for MediaTek-designed, NVLink Fusion-based XPUs. No source compares the $3.5 billion stake to Nvidia’s own internal R&D spending on custom silicon, so how large this bet is relative to Nvidia’s broader chip strategy is not established. The terms of the convertible bonds themselves — conversion price, maturity, coupon rate — were not disclosed anywhere in the sources reviewed. It is also unclear whether MediaTek’s cited $2 billion 2026 ASIC-revenue figure includes work under the new Nvidia partnership specifically, or its broader ASIC business including non-Nvidia customers. And the central strategic question the deal raises — whether Amazon, Google, Microsoft, OpenAI or Anthropic will actually route any custom-chip work through MediaTek and NVLink Fusion, rather than building fully independent of Nvidia’s stack — is not answered by any source reviewed.