Nvidia Invests $3.5 Billion in MediaTek to Keep Its Grip on Custom AI Chips

By Carlos Montiel | Enterprise AI Specialist
Leer en español →
Published: 2026-09-05 | By: Carlos Montiel | Reading time: ~4 minutes

Amazon, Google, Microsoft, OpenAI, and Anthropic are increasingly building their own AI chips to cut their dependence on Nvidia. Nvidia's response wasn't to fight that trend — it was to buy a seat inside it: a $3.5 billion investment in Taiwanese chipmaker MediaTek to make sure even the competition's "custom" silicon keeps running on Nvidia infrastructure.

A convertible bond, not an acquisition

According to reports from TechCrunch, Forbes, and Yahoo Finance, Nvidia structured the deal as a $3.5 billion investment in a MediaTek convertible bond, focused on custom AI chips for data centers, PCs, and cars. As part of the agreement, MediaTek will adopt Nvidia's NVLink Fusion platform, which lets custom chips designed by third parties plug directly into Nvidia's rack-scale data center systems.

The model: you design the core, we handle the rest

The design of the deal makes the strategic goal clear: customers can design their own compute core while Nvidia and MediaTek handle packaging, memory, and rack-level integration. MediaTek projects its custom AI chip business will generate $2 billion in revenue during 2026, and it has raised its market share target in the data center segment to between 15% and 20% by 2027.

The strategic read: as major cloud customers design their own silicon to shrink their Nvidia bill, Nvidia responds by becoming the mandatory toll booth for that very trend — charging for the interconnect, packaging, and rack integration, regardless of whose chip is running inside.

What it means for companies planning AI infrastructure

For companies in Guatemala and Latin America that rely on cloud providers (AWS, Azure, GCP) rather than buying their own hardware, this deal doesn't change anything immediately — but it is a signal of where the underlying cost of the compute capacity those clouds resell is headed. If your cloud provider starts offering instances built on custom silicon (via MediaTek or other NVLink Fusion partners) as a cheaper alternative to traditional Nvidia GPUs, it's worth weighing the performance trade-off before migrating critical training or inference workloads purely on price.

Carlos Montiel
Enterprise AI Solutions Architect
Specialist in LLMs, Agents, and Orchestration
guatemalia.com/en/#contact · info@guatemalia.com

Need to implement AI at your company?

Carlos Montiel is an enterprise AI solutions architect. He implements LLMs, Agents, RAG, and orchestrators for companies in Guatemala and Latin America. Reach out for a consultation.

Contact Carlos Montiel

info@guatemalia.com