Nvidia is putting $3.5 billion into Taiwanese chipmaker MediaTek, a move that shows exactly how it plans to stay on top even as its biggest customers build chips to replace its GPUs. According to TechCrunch AI, the deal gives MediaTek access to Nvidia’s technology so it can design custom AI chips that plug straight into Nvidia-based data centers. It’s a smart hedge, and it tells you a lot about where the chip war is heading.
The deal in brief
- Who: Nvidia and MediaTek, the Taiwanese chipmaker known for smartphone, smart home, auto, and wireless silicon.
- What: A $3.5B investment plus access to Nvidia’s NVLink Fusion ecosystem, including NVLink, the tech that lets any chips (even non-Nvidia ones) talk to each other fast.
- When: Announced Monday, per TechCrunch AI.
- Why it matters: MediaTek can now build custom chips for cloud giants and AI labs that slot right into Nvidia’s rack-scale infrastructure.
Why Nvidia is doing this
Here’s the tension Nvidia is managing. Amazon, Google, Microsoft, OpenAI, and Anthropic are all investing in their own chips to depend less on Nvidia’s GPUs. That’s a direct threat to Nvidia’s core business.
Instead of fighting it, Nvidia is bending with it. By letting MediaTek design custom silicon that runs on NVLink Fusion, Nvidia cedes some ground on the chips themselves while keeping control of the thing that holds every AI data center together: the networking and rack architecture.
“Nvidia is an AI infrastructure company,” Dion Harris, Nvidia’s senior director of HPC and AI hyperscaler infrastructure solutions, told reporters, per TechCrunch AI. “We expanded beyond pure computing chips years ago.”
What stands out here is the strategy. Custom chips can win the workload, but they still ride Nvidia’s platform. As Harris put it, “every cloud, every model builder is deploying our platform in some shape, form, or fashion.”
The circular financing pattern
This deal fits a pattern TechCrunch AI flags directly: Nvidia keeps investing in companies whose business flows back into its own ecosystem. Money goes out, demand for Nvidia’s platform comes back.
It’s not the only recent example. Last week Nvidia announced a similar partnership with Amazon Web Services, minus the direct cash. AWS agreed to deploy an additional 2 million Nvidia GPUs and integrate NVLink Fusion. Different structure, same goal: make Nvidia’s stack the default even where custom chips show up.
What MediaTek gets
MediaTek has been quietly building its custom data center ASIC business. In June the company said it expects that unit to generate $2 billion in revenue in 2026, with plans to chase more of the market.
This partnership pours fuel on that. MediaTek can now offer its customers a standardized, rack-scale platform and deploy their custom chips right alongside it. “This is really about opening up this ecosystem to the entire MediaTek customer base,” Harris said.
The collaboration reaches past data centers too:
- DGX Spark and RTX Spark: MediaTek keeps working with Nvidia on the small developer-focused desktop AI computer, plus Nvidia’s push into consumer AI PCs.
- Software-defined vehicles: The two will keep building auto platforms, with MediaTek leaning on Nvidia’s RTX graphics for smart cockpits and Nvidia Drive AGX for autonomous driving.
Why it matters for the industry
The status quo was simple: if you wanted serious AI compute, you bought Nvidia GPUs. The hyperscalers hated that dependence and started building alternatives. This deal is Nvidia’s answer.
Rather than losing customers to custom silicon, Nvidia is positioning itself as the standard everything plugs into. Even a rival chip becomes a reason to buy Nvidia’s networking and rack architecture. That’s a defensible spot to hold.
“AI is transforming every computing platform, from the world’s largest AI factories to the PC and the car,” Nvidia CEO Jensen Huang said in a statement. “Together, we’re building platforms that bring NVIDIA accelerated computing to new markets.”
Watch for more deals shaped like this one. Expect Nvidia to keep trading GPU exclusivity for platform lock-in, and expect custom silicon players to keep building on top of it rather than around it. Full details are in the original TechCrunch AI report.