Feynman architecture debut? NVIDIA GTC conference may launch 1.6nm chip

The current market is highly focused on the GTC conference. NVIDIA may unveil its next-generation chip codenamed Feynman at the GTC conference and publicly showcase products utilizing TSMC's A16, 1.6nm process for the first time, shifting market attention on its computing power roadmap from Vera Rubin to a further cycle.

According to Wccftech citing South Korean media Chosun Biz, NVIDIA's GTC 2026 presentation plan has "surpassed Vera Rubin," and this year's conference may mark the public debut of Feynman. GTC 2026 will open on March 15 and will return to San Jose, California, USA.

Jensen Huang previously stated that his keynote speech will showcase "previously undisclosed" technologies. For investors, such statements often indicate that a new round of product cadence and key supply chain choices are about to be confirmed, especially regarding the trade-offs between advanced processes and packaging forms.

If Feynman indeed adopts TSMC's A16, Wccftech believes that NVIDIA will be the first, and possibly the only customer in the early mass production phase of this node, further tying market expectations for advanced capacity and yield ramp-up to NVIDIA.

At the same time, the market is also assessing whether Feynman will incorporate Groq's LPU units to reduce latency, but this could significantly increase design and manufacturing complexity and affect the mass production timeline.

Focus of GTC 2026 may shift from Vera Rubin to Feynman

Chosun Biz's report points to a key signal that NVIDIA is preparing to shift the narrative focus at GTC 2026 from Vera Rubin to Feynman.

Similar to past announcements of new architectures at the conference, the presentation of Feynman may primarily focus on capability overviews, architectural outlines, and mass production timelines, rather than disclosing all details at once.

Currently, technical information about Feynman remains limited, but the "backward-looking generation" preview itself is enough to prompt the market to reprice its product iteration cadence for the next few years and its reliance on upstream advanced processes.

TSMC A16, 1.6nm node: Key variables for SPR and initial customer structure

According to Wccftech, Feynman may become one of the first chips to adopt TSMC's A16, 1.6nm process. The A16 is described as a significant leap in the semiconductor field, featuring Super Power Rail (SPR), and is referred to as "the world's smallest node technology." What is more noteworthy is the customer structure. Wccftech believes that NVIDIA will become the first customer in the early large-scale production phase of the A16 node, and "may be the only customer."

At the same time, mobile customers may adopt this standard at a later stage, as it requires architectural-level modifications. For the market, this means that the early capacity utilization and introduction rhythm of A16 may largely revolve around NVIDIA's product strategy.

Speculation on Groq LPU Packaging: Latency Becomes a New Battleground for GPU Manufacturers

In addition to the generational changes in process technology, Feynman has been assigned another potential clue: some analysts speculate that it may first integrate Groq's LPU hardware stack. The discussion starts from the point that latency is becoming one of the key optimization metrics for GPU manufacturers.

In terms of packaging and integration methods, the market speculates that NVIDIA may adopt a path similar to "hybrid bonding," using the LPU unit as an on-package option, with its implementation being compared to AMD's X3D processors.

However, Wccftech also points out that doing so would significantly increase design and production difficulty, meaning that even if the direction is clear, the pace of implementation may still rely more on engineering complexity and manufacturing maturity.

Mass Production Timeline: Expected to Start Production in 2028, Shipments Possibly in 2029-2030

In terms of commercialization rhythm, Wccftech expects: Feynman's production is expected to start in 2028, with customer shipments possibly falling between 2029 and 2030, depending on NVIDIA's strategic choices.

This also explains why GTC 2026 is more likely to be a "forward-looking" release, focusing on architectural outlines and roadmaps, establishing expectations for the next-generation platform first, and then gradually realizing mass production and delivery