NVIDIA CEO Jensen Huang Officially Confirms Vera Rubin Mass Production at GTC Taipei 2026
The system uses TSMC's 3nm process, with a supply chain scale twice that of Grace Blackwell. The assembly time for a single rack has been reduced from 2 hours to 5 minutes, with the entire machine containing 1.3 million components and over 60 trillion transistors.
Microsoft, Dell, and CoreWeave have already deployed the Vera Rubin NVL72 engineering machine, with large-scale shipments set to begin in the second half of 2026.
In market mechanisms, AI computing power infrastructure is entering a phase of large-scale mass production and delivery, with funds accelerating towards high-density liquid cooling systems and supply chains. NVIDIA benefits from capacity ramp-up advantages, while cloud service providers and AI training demand parties face pressure in the competition for computing power acquisition.
Source: Public Information
ABAB AI Insight
NVIDIA previously achieved million-level deliveries with Grace Blackwell, and this mass production of Vera Rubin continues its extreme speed path from announcement to production. The system's reliability is significantly enhanced through wireless cable PCB mid-boards, onboard ConnectX-9 SuperNIC, and BlueField-4 DPU, while the liquid cooling bus supports 5000 amps to meet extreme power consumption needs.
On the capital path, NVIDIA is concentrating resources on supply chain expansion and manufacturing optimization, motivated by the need to quickly respond to major clients like Microsoft and CoreWeave for next-generation AI factories. By doubling capacity, NVIDIA aims to achieve scalable revenue and simultaneously prepare millions of square feet of capacity for Vera Rubin's subsequent ramp-up.
Similar cases include the rapid transition of Grace Blackwell from announcement to mass production, as well as NVIDIA's generational delivery rhythm from Hopper to Blackwell in data centers; NVIDIA is currently in a critical transformation phase from experimental deployment to global factory-level mass production of AI infrastructure.
Essentially, this represents a technological replacement: AI computing systems are shifting from laboratory prototypes to efficient factory-level mass production. The mechanism is that mature supply chains and breakthroughs in manufacturing processes significantly reduce deployment costs, allowing leading manufacturers to quickly convert capital into actual computing power output, thereby consolidating their pricing power and delivery barriers in the global AI infrastructure market.
ABAB News · Cognitive Law
The speed of mass production determines the outcome of the computing power competition.
The stronger the supply chain, the larger the AI factory; the faster the assembly, the more aggressive the delivery.
Excellent companies sell capacity, ordinary companies sell PPT.