NVIDIA is locking in long-term manufacturing capacity to meet surging demand for AI systems, striking multiyear deals across the chip supply chain from substrates to advanced packaging. The company expanded commitments with Kinsus and Samsung Electro-Mechanics to boost ABF substrate output—Kinsus plans a 3.5x capacity increase at its Taoyuan, Taiwan site beginning in 2027—while deepening ties with SPIL and Amkor to scale advanced packaging in Taiwan, Arizona and South Korea through agreements totaling nearly $3 billion. NVIDIA also said volume shipments of its Vera Rubin NVL72 systems and Vera CPUs began in the third quarter, built on TSMC’s 5nm/3nm and CoWoS technologies, and highlighted Foxconn’s global footprint to support next‑gen GPU and networking platforms. The moves aim to reduce bottlenecks in critical inputs, align supply with multi‑year AI demand, and reinforce NVIDIA’s lead in accelerated computing, even as the company flags execution, geopolitical and supply‑chain risks.




























