Who Makes NVIDIA's Chips? Does NVIDIA Manufacture Its Own?

NVIDIA does not manufacture its own chips. TSMC fabricates them, SK hynix and Micron supply the memory, and Foxconn assembles the boards. Here is exactly who makes NVIDIA chips and where they are made.

Make Sure I Know It

Sep 1, 2026 · 3 min read

NVIDIA is one of the most valuable companies on earth and it does not manufacture a single one of its own chips. That surprises people often enough that it's worth answering directly: NVIDIA designs the GPU, and someone else fabricates it, packages it, populates it with memory, and assembles it onto a board.

The short answer

No — NVIDIA does not make its own chips. It is a fabless company, meaning it owns the design and outsources every physical manufacturing step:

Who fabricates NVIDIA chips: TSMC, the Taiwanese foundry that prints NVIDIA's leading-edge GPUs on its most advanced process nodes.
Where NVIDIA chips are made: primarily TSMC's fabs in Taiwan, with advanced packaging done at TSMC facilities as well.
Who supplies the memory: SK hynix, Micron Technology, and Samsung Electronics provide the high-bandwidth memory that sits alongside each GPU die.
Who assembles the finished boards: Foxconn (Hon Hai), among other contract manufacturers.

The tier-by-tier breakdown with live tickers is on the NVIDIA supply chain map.

What "fabless" actually means

A fabless company owns three things: the chip architecture, the software ecosystem, and the customer relationships. It does not own a fab. Building a leading-edge fab costs tens of billions of dollars and takes years, so NVIDIA made the same choice most chip designers make — buy capacity from a foundry rather than operate one.

Who physically builds an NVIDIA GPU

TSMC does the part most people mean by "making the chip." It takes NVIDIA's design and prints it onto silicon wafers at its leading-edge nodes. For NVIDIA's most advanced data-center GPUs there is effectively no drop-in alternative at that performance level — Samsung and Intel both operate foundries, but the leading-edge AI parts run through TSMC.

TSMC also handles advanced packaging, the step that bonds the GPU die together with its memory stacks into a single module. This is a genuine bottleneck: packaging capacity, not wafer capacity, has repeatedly been the constraint on how many AI accelerators can ship in a given quarter.

SK hynix, Micron, and Samsung Electronics supply the high-bandwidth memory. HBM is stacked, tested, and sold as a finished component, so it's manufactured on entirely separate lines from the GPU itself and then brought together at packaging.

Foxconn and other contract manufacturers take the packaged module and build it into the finished board and server tray that a data center actually racks.

Why the answer matters

Ask "does NVIDIA make its own chips" and the honest answer reveals the real structure of the AI supply chain: an enormous amount of the world's compute depends on one foundry's leading-edge capacity, in one country, with a packaging step that only a handful of facilities can perform.

That concentration is what makes the chain efficient and fragile at the same time. A disruption at a single site in Taiwan doesn't slow NVIDIA down — it slows down every AI data center that was waiting on NVIDIA.