NVIDIA Considers Intel Collaboration, Igniting Manufacturing Competition

NVIDIA CEO Jensen Huang revealed in a global media and analyst Q&A conference yesterday that the company maintains an open attitude towards collaborating with Intel in wafer foundry partnerships. Currently, NVIDIA's wafer foundry partners include TSMC and Samsung, with TSMC being the largest partner. Huang hinted at the possibility of introducing Intel as a third-party partner, which has sparked industry speculation about a new wave of competitive battles in advanced wafer manufacturing processes. TSMC remains cautious in response to these developments.

As the AI boom continues to surge, NVIDIA's stock price soared nearly 5% in early trading on the US stock market on the 30th, pushing its market value past $1 trillion, making it the first chip manufacturer to join the trillion-dollar club. Intel also experienced a 2% surge in early trading, while TSMC ADR remained relatively stable.

The global wave of AI brought by NVIDIA has taken the market by storm, and their chips have become highly sought-after products. As a result, there is a synchronous increase in demand for wafer foundry capacity. It is understood that NVIDIA is currently evaluating Intel's manufacturing process and has recently received reports on test chip samples of Intel's next-generation process, which have shown promising results.

Huang stated that NVIDIA has reached a certain level of development and has many customers relying on its products. Therefore, the company hopes to diversify its supply chain as much as possible and have scheduling flexibility. This is crucial to maintain resilience. As a result, their products are produced in many different locations.

Regarding wafer foundry partnerships, Huang openly acknowledged NVIDIA's longstanding and deep collaboration with TSMC, as well as their partnership with Samsung. However, he also expressed an open attitude towards collaborating with Intel.

In response to Huang's comments, TSMC, as usual, declined to comment on specific customer information. Previous research by industry analysts estimated that NVIDIA accounted for approximately 5.8% of TSMC's revenue in 2021, which is expected to decrease to around 3% in 2022. It is speculated that it may rebound to over 6% in 2023, depending on the actual volume of customer shipments.

TSMC is currently NVIDIA's most relied-upon wafer foundry partner. Initially, the industry anticipated that as NVIDIA's AI chip demand continues to skyrocket, TSMC would need to increase production capacity. Given the gap between Samsung's process technology and yield rate compared to TSMC, TSMC was expected to be NVIDIA's preferred choice for additional orders in the future. However, if NVIDIA adds Intel as a wafer foundry partner, it will undoubtedly ignite a fierce competition among TSMC, Samsung, and Intel. The proportion of NVIDIA's order volume allocated to Intel will be worth watching.

NVIDIA's products mainly belong to the high-performance computing sector and adopt advanced process solutions from TSMC, primarily utilizing the 4-nanometer process from the 5-nanometer family. These products include the previously released GeForce RTX 40 series graphics cards and the A100 chip. In the future, NVIDIA will supply the Chinese market with downgraded versions, the A800 and H800, which will adopt the 7-nanometer process, in conjunction with TSMC's advanced CoWoS-S packaging.

Intel is actively investing in wafer foundry operations. Under the leadership of CEO Pat Gelsinger, the company has put forward the IDM 2.0 strategy, which includes expanding its in-house capacity, expanding wafer foundry business, and increasing the utilization of third-party wafer foundry capacity.