Foxconn will be the lead supplier for NVIDIA's Groq 3 LPX, a chip designed specifically to boost AI inferencing capabilities.
NVIDIA Vera Rubin Platforms To Benefit Greatly From Groq 3 LPX For Boosting AI Inferencing, Driving Foxconn's Share To Record Levels
At GTC 2026, NVIDIA announced its Groq 3 LPX chip, in addition to the Vera Rubin AI platform. The Groq 3 LPX is an LPU that is designed to boost AI inferencing by up to 35x. The Groq 3 LPX rack will feature a total of 256 chips with 128 GB of SRAM and 12 TB of DDR5 memory, designed for multi-trillion parameter AI models.
As per Taiwanese news outlets, the Groq 3 LPX is now shipping ahead of schedule in the third quarter of 2026. This shows how NVIDIA is fast-tracking its entire Vera Rubin platform to meet the growing demand for inferencing AI, a key enabler for the Agentic AI era.
Nvidia is aggressively targeting the AI inference market, with its latest AI inference cabinet, Groq 3 LPX, reportedly shipping ahead of schedule in the third quarter. Foxconn is the exclusive supplier of the computing tray and the main supplier of the LPX cabinet assembly, competing with Nvidia for a share of the AI inference market opportunity.
Originally, the industry had outlined limited shipments of Groq 3 LPX this year, but current supply chain reports state that LP30 and LP35 chips featured inside the LPX racks will reach a staggering 1.5 million units this year and 2.5 million units in 2027, respectively.
The reports suggest that Foxconn is the "exclusive" supplier of the Groq 3 LPX computing tray and the main supplier of the LPX cabinet assembly. The huge demand for NVIDIA's Vera Rubin platforms plus the Groq 3 LPX racks will drive Foxconn's share from 55% to 60% in the second half of this year, leading to huge growth in revenue.
Based on the initial estimates, Foxconn is set to deliver 6000 Groq 3 LPX racks this year and another 10,000 racks in 2027. This excludes the next-generation LPX racks based on the LP40 chips, which are expected to start shipping next year.
The primary Vera Rubin "NVL72" racks are expected to reach 12,000 units in 2026, with leading customers being Google, Amazon AWS, and Microsoft. The mass production of Vera Rubin VR200 NVL72 servers is expected to commence by the end of Q3 2026.
Foxconn is also expanding its production for NVIDIA's latest AI servers, given the huge demand. The CEO of Foxconn, Liu Yangwei, has already stated that the company can produce over 1000 cabinets per week, and is expected to increase the production capacity to 2000 cabinets by the end of 2026.
About the author: A Software Engineer by training and a PC enthusiast by passion, Hassan Mujtaba serves as Wccftech's Senior Editor for hardware section. With years of experience in the industry, he specializes in deep-dive technical analysis of next-generation CPU and GPU architectures, motherboards, and cooling solutions. His work involves not only breaking news on upcoming technologies but also extensive hands-on reviews and benchmarking.
Follow Wccftech on Google to get more of our news coverage in your feeds.
