Geo News
Community curated by people like you
LatestAICryptoHealthWorld AffairsUS Politics
Nvidia invests $3.5 billion in MediaTek through convertible bonds
00

Nvidia invests $3.5 billion in MediaTek through convertible bonds

Aug 31, 2026

Nvidia has announced a $3.5 billion investment in zero-coupon convertible bonds issued by MediaTek, representing roughly 90% of the Taiwanese chip designer's record $3.9 billion offshore offering. Under the agreement, MediaTek will adopt Nvidia's NVLink Fusion platform, allowing custom AI accelerators to plug directly into Nvidia's rack-scale infrastructure. While critics raise concerns over circular financing, the deal positions Nvidia to profit from the custom silicon boom by establishing its proprietary interconnect fabric as the industry standard.

Nvidia MediaTek convertible bonds

  • ▪Nvidia announced an investment of $3.5 billion in convertible bonds issued by MediaTek on August 31, 2026
  • ▪Alphabet also purchased bonds in MediaTek's $3.9 billion offering, though the exact investment amount was not disclosed
  • ▪Nvidia's $3.5 billion purchase covers approximately 90% of MediaTek's record $3.9 billion offshore convertible bond offering
  • ▪Nvidia purchased zero-coupon bonds, meaning MediaTek pays no interest unless the bondholders convert the debt into equity at a preset strike price

NVLink Fusion platform adoption

  • ▪The partnership between Nvidia and MediaTek extends to building RTX Spark and DGX Spark chips for AI personal computers and developer workstations
  • ▪MediaTek will adopt Nvidia's NVLink Fusion platform to connect custom-built chips directly into Nvidia's rack-scale computing systems

Custom AI chip interconnect

  • ▪Custom AI accelerators require interconnects, networking, and validated system designs to scale across server racks and run in production environments
  • ▪The NVLink Fusion platform provides hyperscalers, cloud providers, and model developers a prevalidated route to design custom AI accelerators that plug into Nvidia's infrastructure

MediaTek AI accelerator revenue

  • ▪MediaTek is targeting 15% to 20% of the custom AI accelerator market, which is projected to reach $80 billion by 2027
  • ▪MediaTek expects its custom AI accelerator chip business to clear $2 billion in revenue in 2026
  • ▪MediaTek's first AI accelerator application-specific integrated circuit is scheduled for mass production in the fourth quarter of 2026, with a second-generation chip slated for 2028

Circular financing criticism

  • ▪Nvidia CEO Jensen Huang rejected circular financing claims, stating that Nvidia and MediaTek run entirely separate businesses
  • ▪Critics characterized the deal as circular financing, arguing Nvidia bankrolls buyers who purchase its chips, boosting Nvidia's profits and stock value

2 sources

Decrypt
Nvidia Invests $3.5 Billion in MediaTek to Expand Beyond GPUs - Decrypt
View source article
Forbes
Nvidia Turns $3.5 Billion MediaTek Deal Into A Toll On Custom AI Chips
View source article

Featured stories

View more in Compute, chips & AI infrastructure

Marvell raises annual forecast but shares fall on Google AI deal timing concerns

Aug 27, 2026 · 3 sources

SB Energy files for IPO with Nvidia backing

Sep 1, 2026 · 2 sources

South Korea proposes record $597 billion 2027 budget to supercharge AI investment

Sep 1, 2026 · 2 sources

Nvidia unveils NVHBM memory architecture, moving controller into HBM stack to free 25% more compute

Aug 31, 2026 · 3 sources

Story comments

Loading comments…

Related entities

United States

Related Projects

Nvidia

Topics

Compute, chips & AI infrastructureConvertible bondsSemiconductor industry

Featured stories

View more in Compute, chips & AI infrastructure

Marvell raises annual forecast but shares fall on Google AI deal timing concerns

Aug 27, 2026 · 3 sources

SB Energy files for IPO with Nvidia backing

Sep 1, 2026 · 2 sources

South Korea proposes record $597 billion 2027 budget to supercharge AI investment

Sep 1, 2026 · 2 sources

Nvidia unveils NVHBM memory architecture, moving controller into HBM stack to free 25% more compute

Aug 31, 2026 · 3 sources