In a nutshell: Nvidia is predicted to launch its Blackwell GPU structure for AI workloads quickly, and a number of other high-profile prospects have already positioned orders to safe the brand new chips. To satisfy anticipated demand, Nvidia is partnering with Taiwanese producer Foxconn to considerably enhance availability within the coming months and past.
Foxconn is increasing its manufacturing operations in Mexico with a brand new plant devoted to Nvidia’s Blackwell structure, the just lately unveiled GPU designed to speed up AI and high-performance computing workloads.
Foxconn Senior VP Benjamin Ting acknowledged that the corporate is setting up the most important GB200 manufacturing facility on the earth, aiming to fulfill the “large” demand for AI accelerators within the close to future.
The plant will reportedly be in-built Guadalajara, a Mexican authorities supply confirmed. Because the world’s largest electronics contract producer, Foxconn was chosen by Nvidia as a key companion in bringing the Blackwell chip design to fruition. Foxconn Chairman Younger Liu talked about that the manufacturing capability of the brand new plant will likely be “extraordinarily giant,” although he didn’t present particular particulars.
Foxconn is finest generally known as Apple’s major manufacturing companion, producing iPhones, iPads, and different gadgets. Its manufacturing strains additionally assemble different notable merchandise, together with PlayStation, Xbox, and Change gaming consoles, in addition to Amazon Kindle e-readers.
Foxconn’s subsequent step in its diversification plan focuses on AI. Chairman Younger Liu confirmed that the improved manufacturing capabilities on the new Mexico facility will help the manufacturing of superior cooling options required for the GB200 server infrastructure.
Nvidia has hailed Blackwell because the “spark” of a brand new industrial revolution, with expertise particularly designed to speed up AI workloads, scale back inefficiencies and downtimes, and optimize power consumption.
The GB200 NVL72 platform may even energy an upcoming supercomputer that Nvidia and Foxconn plan to construct in Taiwan. The Hon Hai Kaohsiung Supercomputing Middle challenge, unveiled throughout Hon Hai Tech Day earlier this week, goals to deliver on-line a high-performance computing system able to delivering 90 exaflops of “AI efficiency” by 2026.
Nvidia mentioned the brand new supercomputer will likely be Taiwan’s quickest computing infrastructure, with Foxconn managing AI-heavy HPC workloads. The system will likely be used for most cancers analysis, AI mannequin improvement, and sensible metropolis functions. Moreover, the challenge will help digital twins, robotic automation, sensible city infrastructure, and AI-assisted metropolis companies.
The brand new AI supercomputer “is without doubt one of the strongest on the earth, representing a major leap ahead in AI computing and effectivity,” Foxconn vice-president and spokesman James Wu mentioned.
The GB200 NVL72 information heart platform is a rack-based infrastructure that includes 36 Nvidia Grace (Arm) CPUs, 72 Blackwell GPUs, and an NVLink connection able to delivering 130TB/s of bandwidth.
