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

The market is highly focused on GTC 2026. NVIDIA may unveil the next-generation chip codenamed Feynman at GTC, and publicly showcase products using TSMC’s A16, 1.6nm process for the first time, shifting market attention on its computing roadmap further out from Vera Rubin to a longer cycle.

According to Wccftech citing Korean media Chosun Biz, NVIDIA’s GTC 2026 keynote plans have “surpassed Vera Rubin,” and this year’s conference may feature Feynman’s first public appearance. GTC 2026 is scheduled to open on March 15 in San Jose, California.

Jensen Huang previously stated that his keynote will showcase “never-before-seen” technology. For investors, such statements often imply a new product cycle and key supply chain decisions are about to be confirmed, especially regarding advanced process nodes and packaging options.

If Feynman indeed adopts TSMC’s A16, Wccftech believes NVIDIA will be the first, and possibly the only, customer to mass produce at this node during its initial ramp-up phase, further tying market expectations for advanced capacity and yield improvements to NVIDIA.

Meanwhile, the market is also evaluating whether Feynman will incorporate Groq’s LPU units to reduce latency, though this could significantly increase design and manufacturing complexity and impact production timelines.

GTC 2026 Focus May Shift from Vera Rubin to Feynman

Chosun Biz’s report signals a key shift: NVIDIA is preparing to pivot its GTC 2026 narrative from Vera Rubin to Feynman.

Similar to past conferences where new architectures were announced, Feynman’s reveal is likely to focus on capability overviews, architecture outlines, and production timelines, rather than full disclosure of all details at once.

Technical details about Feynman remain limited, but the “next-generation” preview alone is enough for the market to reassess its product iteration pace over the coming years and its dependence on upstream advanced process nodes.

TSMC A16, 1.6nm Node: Key Variable for SPR and Early Customer Structure

According to Wccftech, Feynman may be among the first chips to adopt TSMC’s A16, 1.6nm process. A16 is described as a major breakthrough in semiconductor technology, featuring Super Power Rail (SPR), and being called the “world’s smallest node technology.”

More importantly, the customer structure is noteworthy. Wccftech believes NVIDIA will be the first to mass produce at the A16 node during its early ramp-up phase, and may be the only customer.

Meanwhile, mobile device manufacturers are likely to adopt this standard at a later stage, as it requires architectural modifications. For the market, this means early capacity utilization and deployment of A16 could largely revolve around NVIDIA’s product strategy.

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

Beyond process node generations, Feynman is also speculated to potentially integrate Groq’s LPU hardware stack for the first time. The discussion stems from the fact that latency is becoming a key optimization target for GPU vendors.

In terms of packaging and integration, market speculation suggests NVIDIA might adopt a “hybrid bonding” approach, placing the LPU as an on-package option, similar to AMD’s X3D processors.

However, Wccftech notes that this would significantly increase design and manufacturing complexity, meaning that even with a clear direction, the implementation pace may still depend heavily on engineering challenges and manufacturing maturity.

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

Regarding commercialization, Wccftech estimates: Feynman’s production could begin in 2028, with shipments to customers possibly starting in 2029 or 2030, depending on NVIDIA’s strategic choices.

This also explains why GTC 2026 is more likely to be a “forward-looking” announcement, focusing on architecture outlines and roadmaps, establishing expectations for the next-generation platform, and gradually translating into mass production and delivery.