Intel may become a major manufacturer of GPUs in the coming years as part of its IDM 2.0 initiative. During a question and answer session with the press at Computex, NVIDIA CEO Jensen Huang revealed that the chipmaker has been in talks with Intel over a possible foundry contract. Huang was responding to concerns over the adoption of TSMC as the sole supplier of every present-gen GPU for the company.
You know that we also manufacture with Samsung, and we’re open to manufacturing with Intel. Pat has said in the past that we’re evaluating the process, and we recently received the test chip results of their next-generation process, and the results look good.Jensen Huang, NVIDIA CEO.
The NVIDIA CEO went on to confirm that Team Green had recently received the test chip results of Intel’s next-gen process node, and the results are promising. Intel was originally planning to take external foundry orders starting its 20A (2nm) node in 2024-25. However, in an early-2023 interview, Pat Gelsinger (Intel CEO) disclosed that the IDM 2.0 initiative had secured a leading data center solutions provider for the Intel 3 (3nm) node.
The Intel 3 node will power Intel’s 6th Gen Xeon Granite Rapids-SP processors. It’ll remain out of the cost-oriented client segment and will mainly be used for high-margin center products. In addition to Granite Rapids, Sierra Forest (E-core only cloud CPU) will also leverage the Intel 3 process. These Xeon lineups are slated to launch in late 2024 or early 2025.
Seeing as how TSMC’s 3nm (N3) pricing has gone through the roof, it is possible that NVIDIA may tap into Intel’s 3nm (or Intel 3) node for its next-gen data center GPUs. The RTX 5090 and Co. may be outsourced to a Samsung node as Team Blue may not have the capacity for it.
There is also a possibility that NVIDIA is planning to use the Intel 20A (2nm) in the future for its 2nd Gen Arm-based CPUs. The 1st Gen Grace CPU (5nm) has been received well and its successor would be a good fit for the Intel foundry.