Attention is focusing on Nvidia's latest artificial intelligence (AI) chip, "Feynman," which will be unveiled at the annual developer conference GTC 2026, to be held in California from the 16th to the 19th next month (local time). Nvidia has revealed next-generation AI chips every year through GTC, and for this year's GTC, CEO Jensen Huang has said he will "unveil a chip the world has never seen." Feynman is said to apply Taiwan Semiconductor Manufacturing Co.'s (TSMC) 1-nanometer (nm, one-billionth of a meter) class process.
According to the industry on the 25th, Nvidia is expected to unveil the next-generation AI chip Feynman at GTC 2026. In addition to the next-generation AI chip, Nvidia is expected to showcase a wide range of next-generation technologies, including PC chips entering a new market, software to implement physical AI, and Digital Twin.
Feynman is a next-generation data center graphics processing unit (GPU) that Nvidia is developing with a target launch in 2028. It is a next-generation product succeeding Blackwell, the latest AI chip currently supplied to the market, and Rubin, which is being released this year, and is understood to be the first to go into mass production using Nvidia's 1 nm class foundry (contract semiconductor manufacturing) process.
TSMC, which exclusively mass-produces Nvidia's AI chips, has moved to expand capacity for its 16A (1.6 nm class) process to offer Nvidia a 1 nm class process and is said to begin mass production in the second half of this year. Attention is also focusing on collaboration with Intel, in which Nvidia has made an equity investment, regarding this chip production. Nvidia is reportedly considering entrusting part of Feynman GPU production to Intel. Cooperation on packaging, as well as foundry manufacturing, is also understood to be under discussion.
A semiconductor industry official said, "Nvidia will focus on supplying a next-generation AI chip with overwhelming performance to cement its No. 1 position in the AI chip market," and added, "It is also expected to concentrate on expanding the ecosystem by unveiling Digital Twin, physical AI, and more powered by Nvidia GPUs."
At this year's GTC, Samsung Electronics and SK hynix will also take the stage together to share the results of their collaboration with Nvidia. Samsung Electronics will introduce the performance of its sixth-generation high bandwidth memory (HBM4), released to the market this year. On the 12th of this month, Samsung Electronics formalized HBM4 shipments, which are understood to be mass-produced products to be mounted on Nvidia's Rubin GPUs. Samsung Electronics is also expected to present analysis results on how the seventh-generation HBM (HBM4E), now in development as the next model, is compatible with Nvidia GPUs.
SK hynix is also set to introduce HBM4 and more on the GTC stage. SK hynix will speak under the theme "How HBM4 made large language models (LLMs) more efficient." SK hynix also said it is mass-producing HBM4 to be mounted on Nvidia's Rubin GPUs. This year, SK hynix is expected to supply about two-thirds of the total HBM4 volume requested by Nvidia.