NVIDIA locked down exclusive dibs on TSMC's A16 manufacturing process for their Feynman GPU lineup that drops after Rubin chips hit the market. The Taiwanese fab is basically building production capacity just to keep Jensen Huang happy, and Apple might skip this node entirely to jump straight to A14 tech when they move past their current two-nanometer stuff.
The A16 process brings speed bumps around eight to ten percent, with power cuts hitting fifteen to twenty percent compared to older nodes, plus it uses nanosheet transistors with backside power rails that work better for datacenter workloads. TSMC plans to fire up mass production by late next year at their Kaohsiung facility, and the green team gets first access to wafers before anyone else can even place orders.
Both companies are doubling down on their partnership because AI demand keeps exploding, and NVIDIA needs cutting-edge process tech to stay ahead of AMD and the hyperscalers building their own accelerators.
The A16 process brings speed bumps around eight to ten percent, with power cuts hitting fifteen to twenty percent compared to older nodes, plus it uses nanosheet transistors with backside power rails that work better for datacenter workloads. TSMC plans to fire up mass production by late next year at their Kaohsiung facility, and the green team gets first access to wafers before anyone else can even place orders.
Both companies are doubling down on their partnership because AI demand keeps exploding, and NVIDIA needs cutting-edge process tech to stay ahead of AMD and the hyperscalers building their own accelerators.
