NVIDIA to Reject TSMC’s Price Hike on 5nm Wafers Used to Fab the GeForce RTX 40 Series GPUs

NVIDIA’s next-gen Ada Lovelace GPUs are set to launch the following quarter, starting with the GeForce RTX 4090 on the 12th of October. Fabbed on TSMC’s N4 node (optimized N5), these will be some of the most expensive chips the GPU-maker has ever designed. Pricey boards, substrates, and process technology are the primary factors driving up the BOM.

According to foundry sources, TSMC plans to raise wafer prices for all its clients by 3% to 6% next year. However, Apple, its largest client accounted for more than 25% of the foundry’s revenue in 2021, has refused to go along with the price hike.

NVIDIA, a significant client of TSMC’s N7 and N4 process nodes, is looking to take the same route, hoping to get an exemption from the price increase. If the latter complies, there’s a good chance that AMD and MediaTek will get the same treatment.

As of 2022, Apple is TSMC’s largest customer, followed by AMD at #2 and MediaTek at #3. NVIDIA and Qualcomm have a dual-foundry model wherein they leverage TSMC and Samsung foundries to get the best deal and capacity.

Source: Asia Tech Press