Industry Rumor NVIDIA Drains Foxconn’s Entire CPO Rack Supply Ahead of Schedule, Forcing the Assembler to Skip Showcase at Computex This Year Hassan Mujtaba • at EDT Add on Google Foxconn has reportedly shipped the entirety of its Co-Packaged Optics racks to NVIDIA as it quintuples its shipment estimates for 2026-2027. You May Not Be Able To Find Any Foxconn Co-Packaged Optics Rack At Computex, As The Entire Supply Is Shipped To NVIDIA NVIDIA is accelerating its co-packaged optics development ahead of the Feynman launch in 2028. The next-gen CPO racks solve major infrastructure connectivity & scale-out bottlenecks with AMD all-in on the race too. Related Story AMD Rolls Out Six New Ryzen PRO 9000 Chips Headlined by the 9965X3D, Months After the SKU First Surfaced OnlineAs per rumors in the Taiwanese industry, Foxconn has apparently shipped all of its CPO racks to NVIDIA ahead of schedule. NVIDIA is Foxconn's major customer for AI infrastructure, and plays a crucial role in supplying Optical switch racks. The company had originally estimated over 10,000 unit shipments for 2026, but has since increased the estimate to 50,000 through 2026-2027, 5x the original requirements. Foxconn is expected to see double-digit gross margins and higher ASPs through Co-Packaged Optics solutions, contributing to over 15% of its revenue. United Daily News further states that Foxconn had initiated mass prototype shipments in the first quarter of 2026, followed by mass production and shipments in the third quarter. The racks are being made at the Vietnam plant, and as per industry sources, all prototype racks have now been shipped to NVIDIA. This signals tightness in the supply chain as display cabinets, usually reserved for major conventions or events, have also been shipped out. The company is expected to go CPO Rack-less at this year's Computex. Industry insiders say that Foxconn's all-optical CPO switch cabinets, manufactured at its Vietnam plant, have already been shipped to Nvidia, and the supply is extremely tight. Even the display cabinets have been given to Nvidia, with "not a single unit left." Foxconn's original plan to showcase CPO switch cabinets at this year's Taipei International Computer Show may not result in any actual demonstrations. United Daily News NVIDIA is planning to ship more than 50,000 CPO switches and racks between 2026 and 2027, and Foxconn, being the sole supplier of this solution, will benefit massively. Co-Packaged Optics solutions will mark the first arrival in 2027, followed by large-scale use by 2028 as next-generation AI platforms such as Rubin & Feynman enter the data-center segments. About the : A Software Engineer by training and a PC enthusiast by passion, Hassan Mujtaba serves as 's 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 on Google to get more of our news coverage in your feeds. Further Reading Zyphra & AMD Launch New Open AI Platform Powered By 15MW MI355X GPUs With Expansion Planned To MI450 & Beyond AMD’s vLLM-ATOM Plugin Supercharges DeepSeek-R1, Kimi-K2, and gpt-oss-120B AI LLM Inference on Instinct MI350 and MI400 Accelerators Intel Resurrects On-Package Memory With Razor Lake-AX, Loading Up LPDDR6 to Hunt Down AMD’s Medusa Halo by 2028 Elon Musk and Tim Cook Join US President Trump’s China Trip This Week, Yet NVIDIA’s Jensen Wasn’t On The List Read all on NVIDIA Drains Foxconn’s Entire CPO Rack Supply Ahead of Schedule, Forcing the Assembler to Skip Showcase at Computex This Year
Reading tools
Key takeaways
- Industry Rumor NVIDIA Drains Foxconn’s Entire CPO Rack Supply Ahead of Schedule, Forcing the Assembler to Skip Showcase...
- You May Not Be Able To Find Any Foxconn Co-Packaged Optics Rack At Computex, As The Entire Supply Is Shipped To NVIDIA N...
- The next-gen CPO racks solve major infrastructure connectivity & scale-out bottlenecks with AMD all-in on the race too.
Share this article